<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Onchain Divers on Medium]]></title>
        <description><![CDATA[Stories by Onchain Divers on Medium]]></description>
        <link>https://medium.com/@hello_89216?source=rss-1b5226afa4e7------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*rJXrErcPBYYZaLHN-5Mw5g.jpeg</url>
            <title>Stories by Onchain Divers on Medium</title>
            <link>https://medium.com/@hello_89216?source=rss-1b5226afa4e7------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 22:12:49 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@hello_89216/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Free on-chain dashboards with 90 days of history: Pump.fun, Hyperliquid, Polymarket and more]]></title>
            <link>https://medium.com/@hello_89216/free-on-chain-dashboards-with-90-days-of-history-pump-fun-hyperliquid-polymarket-and-more-fb612fd7bdb5?source=rss-1b5226afa4e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/fb612fd7bdb5</guid>
            <category><![CDATA[hyperliquid]]></category>
            <category><![CDATA[prediction-markets]]></category>
            <category><![CDATA[solana-network]]></category>
            <category><![CDATA[memecoins]]></category>
            <category><![CDATA[blockchain]]></category>
            <dc:creator><![CDATA[Onchain Divers]]></dc:creator>
            <pubDate>Fri, 13 Nov 2026 15:33:18 GMT</pubDate>
            <atom:updated>2026-11-13T15:33:18.086Z</atom:updated>
            <content:encoded><![CDATA[<h4>Write SQL, chart the results, fork public work, or ask an agent to build the page for you.</h4><p>How many pump.fun tokens launched each day last week? After a one-line setup, ask Claude, Cursor or ChatGPT, and get back a public chart you can send to anyone. You don’t need an account, an API key or credits.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*B44d0HJPxlHz6ulJ.png" /></figure><h3>Astrolabe is the workbench; the indexer is the database</h3><p>We built <a href="https://astrolabe.cartographer.club/">Astrolabe</a> so that anyone can work with the same on-chain data we use every day, for free. It’s a free workbench built on the <a href="https://onchaindivers.com/">Onchain Divers indexers</a>. The indexer is a full ClickHouse database with the complete history. Astrolabe is a Dune-style layer on top of it. You write SQL or ask the built-in AI assistant to write it, chart the results, build dashboards, share or embed them, and fork public work. There are three ways in: the site, the MCP server for agents, or the database itself.</p><h3>What’s inside</h3><p>The free workbench exposes the same raw tables as the indexers:</p><ul><li><strong>Solana:</strong> pump.fun, PumpSwap, Raydium, Meteora, Jito tips, token transfers and blocks</li><li><strong>Hyperliquid</strong> perps and <strong>Polymarket</strong> prediction markets</li><li><strong>Robinhood Chain, BNB Chain and Arc</strong>, including their launchpads, plus <strong>TON</strong></li><li><strong>Real-world assets on Solana:</strong> tokenized stocks, pre-IPO assets and commodities from xStocks and Backpack Securities</li></ul><p>Every table and column has a description, including units. With raw blockchain data you can easily lose an hour working out whether an amount is in lamports or SOL, or whether a token amount is raw or decimal-adjusted. The descriptions are what stop you from shipping a chart that looks plausible and is wrong.</p><h3>Who it is for</h3><ul><li><strong>Memecoin and copy traders:</strong> fork a public wallet dashboard, change the window or the threshold, and get a research shortlist.</li><li><strong>Quants and ML researchers:</strong> build features and backtests from real fills, with the SQL in full view.</li><li><strong>Analysts:</strong> compare Polymarket, Hyperliquid and Solana RWAs in one interface.</li><li><strong>Builders:</strong> embed live Astrolabe charts in your own site, docs or product.</li></ul><h3>Dashboard: Pump.fun Smart Money</h3><p><a href="https://astrolabe.cartographer.club/dashboards/71">Pump.fun Smart Money</a> is our most viewed and most forked dashboard. It takes every pump.fun and PumpSwap trader from the last three days. After each address’s first buy of a new token, it checks whether that token reached 2x within three hours. Then it ranks the addresses by hit rate.</p><p>Read it as a starting point for copy-trading research, not as a signal service. A historical hit rate does not predict the next trade, and three hours is only one definition of success. Fork it to test another window, threshold or ranking. Every public dashboard is a template.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*oyU6gf26tRKUE_4L.png" /></figure><h3>Dashboard: real-world assets on Solana</h3><p><a href="https://astrolabe.cartographer.club/dashboards/17">RWA Overview</a> tracks tokenized stocks, pre-IPO assets and commodities on Solana. Tokenized equities are a fresh narrative, and this is a narrative you can follow trade by trade on-chain. Look at the market, then fork the dashboard to focus on another set of assets.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*2tXD0Oslg4eW-M4r.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*COB98oeTjU5PnNFn.png" /></figure><p>There are also <a href="https://astrolabe.cartographer.club/dashboards/38">Hyperliquid Overview</a>, with volume, traders, coins and liquidations on perps, and <a href="https://astrolabe.cartographer.club/dashboards/33">Polymarket Overview</a>, with volume, trades, traders and market activity.</p><h3>Built on the same data</h3><p><a href="https://rwa.trenchscreener.ai/">RWA World</a> is an on-chain liquidity monitor for tokenized stocks and gold on Solana. It builds an implied order book from Jupiter, DFlow and OKX router quotes. It is a separate product, but it runs on the same Solana indexer as Astrolabe, which shows the data is enough to build real tools on.</p><h3>Give an agent the data, the rules and a way to publish</h3><p>The first piece of feedback from everyone was “make an MCP”, so we made one. The MCP server is free and needs no key and no sign-up:</p><pre><a href="https://astrolabe.cartographer.club/mcp">https://astrolabe.cartographer.club/mcp</a></pre><p>Claude Code setup is one line:</p><pre>claude mcp add --transport http astrolabe https://astrolabe.cartographer.club/mcp</pre><p>Setup for the Claude app, Cursor, Codex, ChatGPT and any other MCP client is at <a href="https://astrolabe.cartographer.club/agents">astrolabe.cartographer.club/agents</a>.</p><p>Paste this as your first prompt:</p><blockquote><em>Show me pump.fun token launches per day over the last 7 days, and give me a link I can share.</em></blockquote><p>Here is what happens next:</p><ol><li><strong>The agent finds the tables</strong> and reads the description of every column, including units.</li><li><strong>It reads our rules.</strong> A dedicated AGENTS database holds our notes on units, freshness, partition filters and known quirks, plus verified example queries. This is why the agent writes correct SQL instead of guessing.</li><li><strong>It writes and runs the query.</strong></li><li><strong>It saves the result as a public page</strong> with a chart and sends you the link.</li></ol><p>Ask for a single number and you get a number. Ask for a series, a breakdown or a comparison and you get a chart with a link. Ask for a dashboard and the agent builds one. Paste a link to someone else’s dashboard and ask for changes: the agent forks it and edits the copy, never the original.</p><p>A more involved example is <a href="https://astrolabe.cartographer.club/queries/573">Pump.fun / PumpSwap: best wallets for copy-trading, last 10 days</a>, built entirely by an agent. It reconstructs realized PnL per token for closed positions, then filters for organic win rates, minimum holding time and profit factor. It shows that an agent can do non-trivial research on its own. The result is still a starting point, not a signal.</p><p><em>For developers:</em> the data tools are the official mcp-clickhouse ones (list_databases, list_tables, run_query). Astrolabe adds save_query, create_dashboard, fork_dashboard, search and a few more. Everything agents save is public, in English, under the agents account.</p><h3>A free account adds your own data and alerts</h3><p>Registration is free, and it turns Astrolabe from a viewer into a working tool:</p><ul><li><strong>Your own data:</strong> upload a CSV, such as your wallet list, labels or watchlist, as a table and join it with the indexer data.</li><li><strong>Alerts:</strong> any saved query can run on a schedule and send an e-mail or a webhook to Slack, Discord or Telegram when its result meets a condition.</li><li><strong>Private work:</strong> private queries and dashboards, saved across devices, plus stars and a public profile page.</li></ul><p>Put those together: upload a list of wallets you follow, join it with pump.fun activity, and get a Telegram ping when one of them buys a fresh token.</p><h3>Free, registered and live</h3><p><strong>No account (free)</strong> — SQL editor, charts, dashboards, parameters, forks, embeds, sharing, AI assistant — MCP server for agents, with no key — About 90 days of history, loaded daily at 11:00 UTC for the previous UTC day — Everything you save is public</p><p><strong>Free account</strong> — Everything above — Private work, CSV uploads, alerts, stars and a profile page</p><p><strong>Live access (200 USDC per indexer per 30 days)</strong> — The indexers themselves, seconds behind the chain — Full history: Solana since 2024, Polymarket since 2022, Hyperliquid since July 2025 — Team research tables: wallet clusters, labels and derived views — One switch moves your existing queries and dashboards to live data, with nothing to rewrite — A direct read-only ClickHouse login with no API layer and no request limits, for Python, DBeaver, Grafana, Jupyter or your own code. The same login works with mcp-clickhouse on your machine, so your agent gets live data too.</p><p>Subscriptions are per indexer: Solana, Hyperliquid, Polymarket, or EVM chains (Robinhood Chain, BNB Chain, Arc). You pay a USDC invoice on Solana from any wallet, and access turns on within a minute or two. If you already have indexer access, log in on the site with the same credentials. A trial is available on request in the site’s help chat or on <a href="https://t.me/inventandchill">Telegram</a>.</p><h3>The limits, plainly</h3><p>Free data ends with yesterday and loads once a day. Most tables keep about 90 days. jito_tips keeps 14 days and token_transfers keeps 31. Without an account, everything you save is public.</p><h3>Start with one question</h3><ul><li><a href="https://astrolabe.cartographer.club/">Open Astrolabe</a> and browse the dashboards.</li><li><a href="https://astrolabe.cartographer.club/agents">Connect the MCP server</a> and try the seven-day launch prompt.</li><li><a href="https://astrolabe.cartographer.club/dashboards/71">Fork Pump.fun Smart Money</a> and change one assumption.</li></ul><p>Questions: <a href="https://t.me/inventandchill">t.me/inventandchill</a> or the help chat on the site.</p><p><em>OnChain Divers provides decoded Solana data: launch, swap and migration data for researchers building lifecycle-aware market models. Subscribe for schema updates, reproducible SQL and delayed research examples.</em></p><p><strong>Connect with us:</strong></p><ul><li>Website: <a href="https://onchaindivers.com/">onchaindivers.com</a></li><li>Twitter: <a href="https://x.com/onchain_divers">@onchain_divers</a></li><li>Discord: <a href="https://discord.com/invite/PRWyXUvhU3">Join our community</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fb612fd7bdb5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Memories from the old Jito auction room]]></title>
            <link>https://medium.datadriveninvestor.com/memories-from-the-old-jito-auction-room-cf65034beb5f?source=rss-1b5226afa4e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/cf65034beb5f</guid>
            <category><![CDATA[memecoins]]></category>
            <category><![CDATA[cryptocurrency-investment]]></category>
            <category><![CDATA[jito]]></category>
            <category><![CDATA[solana-network]]></category>
            <category><![CDATA[blockchain]]></category>
            <dc:creator><![CDATA[Onchain Divers]]></dc:creator>
            <pubDate>Mon, 09 Nov 2026 18:11:56 GMT</pubDate>
            <atom:updated>2026-11-10T11:50:52.522Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*t9uYYj8tqzcdEEO7.png" /></figure><p>Two old Solana tricks survived in different forms. One is preserved in a live program whose bytecode still carries a clock switch, two tip accounts, and a fixed XOR mask. The other survives as an operator’s recollection: make simulation read an old account state, then let live execution read the new one.</p><p>There is a particular kind of Solana story that begins with two bundles submitted one after another.</p><p>The details have softened. The slots and signatures are gone. What remains is the shape of the trick: Jito’s auction simulated one world, while the validator eventually executed in another.</p><p>One version used time. Another used state.</p><p>This article is an attempt to reopen both case files without polishing the uncertainty out of them. The first case has a surviving artifact. We pulled the program from chain, matched its hash, opened it in a disassembler, and found the machinery where the old story said it would be. The second case is oral history. It is technically plausible and useful, but it does not yet have the same paper trail.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*NzdPg-O9aFokqU7L.png" /></figure><h3>Case file 1: the program that kept two tips</h3><p>The rumor was simple enough to repeat and strange enough to remember. A bot could show Jito’s block engine a generous tip during simulation, then pay almost nothing when the transaction actually ran.</p><p>The program was known as sra4.</p><p>At first, all we had was a compiled binary and a name. Then the live account supplied the missing chain of custody:</p><pre>Program<br>sra4HHgwsn5Ts6V1TUdLykK1gRJwDQhgscuvXDHxSCx<br><br>ProgramData<br>BrfLQkZwde23FnZJsHud7MezWhtDxLoe8KUuSXxeibws</pre><p>We fetched the upgradeable-loader accounts through Solana RPC and cut the ELF out of the ProgramData payload. It was an 847,872-byte stripped SBF shared object.</p><pre>SHA-256<br>3f4de86d593ea8ff2d0670739dc298020e26def00d882f7f8e6b7efd04601453</pre><p>That hash matches the supplied investigation exactly. The file under the microscope is not merely “a copy attributed to sra4.” It is byte-for-byte identical to the ELF currently stored behind the <a href="https://solscan.io/account/sra4HHgwsn5Ts6V1TUdLykK1gRJwDQhgscuvXDHxSCx">live program account</a>.</p><p>The loader metadata records deployment slot 273,039,214 and upgrade authority arsc4jbDnzaqcCLByyGo7fg7S2SmcFsWUzQuDtLZh2y. The program can still be upgraded, so we froze the recovered ELF in the release archive. On-chain archaeology needs dated specimens.</p><h3>The labels no one removed</h3><p>Stripped binaries usually make you work for the story. This one left two labels in plain sight:</p><pre>jito_fake_tip_account<br>jito_real_tip_account</pre><p>They sit next to each other in the live ELF:</p><pre>0x629ab  jito_fake_tip_account<br>0x629c0  jito_real_tip_account</pre><p>The names are almost too convenient. They are also not enough on their own. A string does not prove which addresses were passed at runtime, whether Jito recognized either destination as a tip account, or whether the mechanism ever won an auction.</p><p>But it tells us where to keep digging.</p><h3>A little XOR stage magic</h3><p>The disassembler turns the old anecdote into something more physical. At SBF PC 32309 (byte VA 0x3f1a8), the program loads this signed constant:</p><pre>-6383738603697460409</pre><p>Read as an unsigned 64-bit value, it becomes:</p><pre>0xa768678ffbaa2b47</pre><p>Then the same constant is XORed into several loaded fields:</p><pre>32309: r1 = -6383738603697460409 ll<br>32311: r2 = *(u64 *)(r10 - 3488)<br>32312: r2 ^= r1<br>32313: r6 = *(u64 *)(r10 - 3480)<br>32314: r6 ^= r1<br>...<br>32318: r3 = *(u64 *)(r10 - 3496)<br>32319: r3 ^= r1<br>...<br>32324: r3 = *(u64 *)(r10 - 3504)<br>32325: r3 ^= r1</pre><p>This is the “XOR magic,” although it is better understood as stage makeup than cryptography. The mask makes raw values less obvious in memory and slows down casual inspection. It cannot keep a secret from an analyst because the mask and the instructions that remove it are shipped inside the program.</p><p>A few instructions later, the rest of the payment skeleton appears:</p><pre>32316: r9 /= 10000<br>32322: if r5 &gt;= r6 goto +1<br>...<br>32344: if r4 &gt; 1000 goto +1<br>32345: r4 = 1000</pre><p>The /10000 is consistent with basis-point-style rate scaling. The comparison at PC 32322 is the recovered boundary branch. The final pair enforces a minimum of 1,000 lamports on the path we tested.</p><p>The stripped ELF does not preserve a friendly source name for every register, so we should not pretend the assembly says current_slot in bright letters. The slot and rate interpretation comes from tracing the surrounding program flow and exercising the original instructions. The live disassembly independently confirms the fixed-mask decoding, the boundary choice, the rate denominator, and the 1,000-lamport floor.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*hvmcyNN03nvhe3CL.png" /></figure><h3>The clock switch</h3><p>Jito bundles execute sequentially and atomically. Before a validator includes them, Jito’s block engine simulates submitted bundles, compares profitable combinations, and forwards winners.[1]</p><p>That creates two moments:</p><ol><li>the block engine previews the proposed bundle;</li><li>a validator executes it later.</li></ol><p>sra4 did not need a syscall that said, “you are inside a simulation.” It used Solana’s clock as a proxy. The caller supplied a cutoff slot, two tip settings, and two possible recipients.</p><pre>observed slot &lt; cutoff<br>    early recipient<br>    early tip<br><br>observed slot &gt;= cutoff<br>    later recipient<br>    later tip</pre><p>If simulation happened before the cutoff and execution happened after it, the same instruction data could produce two different payments.</p><p>The isolated test held everything constant except the observed slot:</p><pre>profit input:        1 SOL<br>available balance:  10 SOL<br>early tip rate:     90% of profit<br>later tip rate:      0%</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*n_aCVQS3yHJfQexI.png" /></figure><p>The first result is 900,000 times the second.</p><p>The tiny payment is not one lamport, despite the way the story is sometimes told. It is 1,000 lamports. The later rate is zero, but the program raises the calculated transfer to that floor.</p><p>These are controlled test values. They are not amounts recovered from a historical auction. Seven illustrative cases and 1,000 generated cases exercised the recovered payment behavior; no live exploit was submitted.</p><h3>What the old auction would have seen</h3><p>The likely sequence reads like a timing diagram:</p><ol><li>The operator chooses a cutoff near the expected gap between simulation and execution.</li><li>The block engine simulates before that cutoff.</li><li>The early path calculates the large tip.</li><li>The auction evaluates the bundle using the simulated result.</li><li>Execution crosses the cutoff.</li><li>The later path pays the minimum.</li></ol><p>The binary and tests support the branch behavior. Jito’s documentation supports the simulation-then-validator pipeline.[1] The missing piece is the auction ledger: we do not have evidence showing that Jito credited the early result, ranked the bundle on that basis, and then accepted the smaller landed payment.</p><p>The program preserves the attempt. It does not preserve the scoreboard.</p><h3>Case file 2: the account that knew whether the world was live</h3><p>The second story comes from the same period, but its evidence is different. This is a recollection from the operator, not a recovered program claim.</p><p>This mechanism required two separate Jito bundles. The first changed a designated account. The second carried the trade and read that account before deciding which path to take:</p><pre>bundle A<br>    change the state of a designated account<br><br>bundle B<br>    read that account<br>    old state  -&gt; treat this as simulation<br>    new state  -&gt; treat this as live execution<br>    then trade accordingly</pre><p>Why could that work?</p><p>The block engine simulates bundle B against a bank snapshot where bundle A has not landed. The reader sees the old value. On the live path, bundle A lands first and commits the state change. When bundle B executes afterward, its trade sees the new value.</p><p>The account becomes a one-bit witness:</p><pre>old value = the earlier write is not visible here<br>new value = the earlier write has already happened here</pre><p>Unlike the clock switch, this version does not wait for a slot number to cross a threshold. It asks whether a prior side effect exists in the current view of the world.</p><p>That distinction is useful. Time is only one way to make simulation and execution diverge. Any input that can change between the two observations may become a discriminator: account state, balances, sequence counters, or the presence of a write from an earlier transaction.</p><p>The separation into two Jito bundles is essential. If the state-changing transaction and the trade sat inside one correctly simulated atomic bundle, the simulator should apply the first transaction’s write before running the second. There would be no old-state signal. The trick depends on bundle B being simulated before bundle A’s write is present, then executed after bundle A has landed.</p><h3>Two tricks, one invariant</h3><p>The clock switch and the state latch look different in code, but they pressure the same assumption.</p><p>The auction simulates a bundle because it wants to price what will happen. That works only if the properties used for ranking survive into execution.</p><p>For a tip, the invariant is:</p><pre>if simulation values a payment of x,<br>execution must pay x or fail</pre><p>For a state-dependent trade, it is broader:</p><pre>if simulation ranks a transaction using state s,<br>execution must not silently choose another economic path under state s&#39;</pre><p>Atomic execution is not enough. A bundle can be atomic and valid while still making its economically important behavior depend on a clock tick or state change that differs between preview and settlement.</p><h3>What an indexer can recover</h3><p>A chain indexer sees the world after execution. It can record:</p><pre>executed slot<br>account state before and after the landed transaction<br>actual recipient<br>actual lamports transferred<br>transaction order<br>success or failure</pre><p>It cannot reconstruct, from chain data alone:</p><pre>simulation slot<br>simulation bank snapshot<br>simulated account state<br>simulated recipient or transfer<br>auction score<br>competing bundles</pre><p>That is why these old stories are so difficult to prove years later. The chain preserved the ending, not the preview Jito used to rank the submission.</p><p>A serious detector needs both views. On-chain data can identify programs with slot branches, suspiciously small landed tips, or trades conditioned on freshly changed accounts. Auction telemetry must show what the simulator observed and what value it assigned to the bundle.</p><h3>What is proven, remembered, and still missing</h3><p>The first case now has a solid artifact trail:</p><ul><li>the live sra4 ProgramData account yields the exact ELF from the supplied investigation;</li><li>the ELF contains both Jito account labels;</li><li>the disassembly contains a fixed XOR mask, a boundary branch, /10000 rate scaling, and a 1,000-lamport floor;</li><li>isolated tests demonstrate the 0.9 SOL versus 1,000-lamport behavior across the cutoff.</li></ul><p>We still do not know which historical auctions, if any, were won this way or how much value was captured.</p><p>The second case is an operator’s recollection. It establishes the intended two-bundle design, but we do not yet have the program, signatures, bundle IDs, or a reproducible test. We are publishing it as oral history while keeping that evidence boundary intact.</p><p>The operator also recalls that Jito later pushed participants toward fixed-size tips. We have not found the original announcement, so that remains attributed. The rationale fits the first failure mode: a fixed payment committed in the submitted transaction cannot shrink merely because the clock advances.</p><p>The old auction room did not leave us a complete diary. It left a live ELF, two wonderfully indiscreet account labels, a hexadecimal mask, and a second story about an account changing state between preview and settlement.</p><p>Sometimes that is enough to reopen the case.</p><p><em>Research note: On 2026–11–05 UTC, we fetched program </em>sra4HHgwsn5Ts6V1TUdLykK1gRJwDQhgscuvXDHxSCx<em> through Solana RPC and extracted its ELF from ProgramData account </em>BrfLQkZwde23FnZJsHud7MezWhtDxLoe8KUuSXxeibws<em>. The SHA-256 is </em>3f4de86d593ea8ff2d0670739dc298020e26def00d882f7f8e6b7efd04601453<em>. The release package includes the RPC responses, deployed ELF, full disassembly, relocations, focused disassembly notes, deterministic string-offset report, and reproduction scripts. The original test-result files were not present in </em>/tmp<em>; the reported controlled-test outcomes are preserved from the supplied investigation.</em></p><p><em>OnChain Divers publishes evidence-led research on Solana execution, market structure, and indexer behavior. Subscribe for new technical investigations and reproducible data notes.</em></p><h3>Sources</h3><p>[1] <a href="https://docs.jito.wtf/lowlatencytxnsend">https://docs.jito.wtf/lowlatencytxnsend</a> — Jito low-latency transaction send documentation</p><p><em>OnChain Divers provides decoded Solana data: launch, swap and migration data for researchers building lifecycle-aware market models. Subscribe for schema updates, reproducible SQL and delayed research examples.</em></p><p><strong>Connect with us:</strong></p><ul><li>Website: <a href="https://onchaindivers.com/">onchaindivers.com</a></li><li>Twitter: <a href="https://x.com/onchain_divers">@onchain_divers</a></li><li>Discord: <a href="https://discord.com/invite/PRWyXUvhU3">Join our community</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cf65034beb5f" width="1" height="1" alt=""><hr><p><a href="https://medium.datadriveninvestor.com/memories-from-the-old-jito-auction-room-cf65034beb5f">Memories from the old Jito auction room</a> was originally published in <a href="https://medium.datadriveninvestor.com/">DataDrivenInvestor</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[One Buyer, One Slot, Then PumpSwap]]></title>
            <link>https://medium.datadriveninvestor.com/one-buyer-one-slot-then-pumpswap-3cd0c019b7a2?source=rss-1b5226afa4e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/3cd0c019b7a2</guid>
            <category><![CDATA[blockchain]]></category>
            <category><![CDATA[cryptocurrency-investment]]></category>
            <category><![CDATA[backtesting]]></category>
            <category><![CDATA[solana-network]]></category>
            <category><![CDATA[memecoins]]></category>
            <dc:creator><![CDATA[Onchain Divers]]></dc:creator>
            <pubDate>Mon, 09 Nov 2026 18:03:52 GMT</pubDate>
            <atom:updated>2026-11-10T11:51:28.224Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Instant versus organic-like PumpFun migrations</em></p><p><strong>Migration cohort: October 1–25, 2026 UTC. PumpSwap outcomes use the same 24-hour horizon after each migration.</strong></p><p>The fresh-token histogram raised a more interesting question than token age alone could answer. A token can reach PumpSwap in two very different ways:</p><ol><li>one wallet effectively clears the curve immediately; or</li><li>hundreds of wallets buy through the curve before migration.</li></ol><p>Those paths should not be pooled together. They create the same on-chain endpoint — a canonical PumpSwap pool — but the flow that gets them there is completely different.</p><p>The difference persists after migration. In our two clean cohorts, instant migrations produced <strong>74.3% of PumpSwap transactions but only 45.7% of priced volume</strong>. Organic-like migrations produced fewer transactions, larger average trades, and more 24-hour volume per token.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*mzf457AGRN0GSD9Z.jpeg" /></figure><h3>“Instant” needs a mechanical definition</h3><p>Creation-to-migration latency is not a smooth distribution. There is a distinct zero-second spike: <strong>10,748 of 28,727 observed migrations</strong> landed in the same indexed second as token creation, and <strong>10,321</strong> of those were also in the same slot.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*zEhNWE6UhMy-ecK5.jpeg" /></figure><p>They were not one atomic transaction. Creation and migration had different signatures in every strict instant case. The common pattern was a short ordered sequence inside one slot: <strong>create, one dominant curve buy, then migrate.</strong></p><p>For the primary comparison, we use two intentionally strict definitions.</p><h4>Instant migration</h4><p>A token is classified as instant when all four conditions hold:</p><pre>creation timestamp == migration timestamp<br>creation slot == migration slot<br>successful pre-migration buy transactions &lt;= 3<br>largest buyer supplied &gt;= 50% of pre-migration quote flow</pre><p>This produces <strong>9,138 tokens</strong>, or <strong>33.6%</strong> of the 27,208-token equal-horizon cohort.</p><p>The median instant token had:</p><ul><li><strong>0 seconds</strong> from creation to migration;</li><li><strong>1</strong> successful pre-migration buy transaction;</li><li><strong>1</strong> buying wallet;</li><li><strong>100%</strong> of pre-migration quote flow from the largest buyer.</li></ul><p>Among native-SOL launches, that median buy was <strong>85.0 SOL</strong>.</p><h4>Organic-like migration</h4><p>We call the second group “organic-like,” not organic. <strong>Wallet count does not prove independent people or independent capital.</strong></p><p>A token enters this cohort when:</p><pre>creation-to-migration latency &gt;= 10 seconds<br>successful pre-migration buy transactions &gt;= 20<br>pre-migration buying wallets &gt;= 20<br>largest buyer supplied &lt;= 30% of pre-migration quote flow</pre><p>This produces <strong>8,055 tokens</strong>, or <strong>29.6%</strong> of the cohort. Another <strong>10,015 tokens</strong> remain in a transition/other bucket and are not forced into either side of the headline comparison.</p><p>The median organic-like token had:</p><ul><li><strong>401 seconds</strong> from creation to migration;</li><li><strong>705</strong> successful pre-migration buys;</li><li><strong>436</strong> buying wallets;</li><li>only <strong>3.7%</strong> of quote flow from the largest buyer.</li></ul><p>For the native-SOL subset, the median buy was <strong>0.174 SOL</strong>. The contrast is therefore not merely “fast versus slow.” <strong>It is one very large curve-clearing order versus a distributed sequence of much smaller orders.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*VGURm3Pe3zXP5oRf.jpeg" /></figure><h3>What happened on PumpSwap</h3><p>Every token receives the same 24-hour observation horizon after its decoded migration. We join PumpSwap swaps to the exact migrated mint × pool, then price quote flow with the indexer’s historical daily quote-price series.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*CntZDcEZL6uTK08R.png" /></figure><p>Across all supported quote assets, organic-like tokens generated <strong>$3.04 billion</strong> of priced 24-hour PumpSwap volume versus <strong>$2.55 billion</strong> for instant tokens. That is <strong>19.0% more total volume despite 11.9% fewer tokens</strong>. Median volume per token was <strong>33.0% higher</strong>.</p><p>The transaction picture runs in the opposite direction. Instant tokens generated <strong>165.1 million</strong> PumpSwap transactions against <strong>57.0 million</strong> for organic-like tokens. The median instant token had 3.1 times as many PumpSwap transactions, but aggregate priced notional per transaction was 3.4 times larger in the organic-like cohort.</p><p>This is the central finding: <strong>instant migrations maximize event count; distributed migrations carry more dollars per token and per transaction.</strong></p><h3>Quote-asset mix explains part of the aggregate gap</h3><p>Instant migrations were almost entirely native-SOL quoted: <strong>99.5%</strong> of the cohort. The organic-like group was <strong>82.2%</strong> native SOL and included more tokens quoted in USDC or other Pump-created coins.</p><p>Restricting both groups to native-SOL launches removes the aggregate-volume advantage:</p><ul><li>instant: <strong>9,096 tokens</strong>, <strong>$2.543 billion</strong> of 24-hour volume;</li><li>organic-like: <strong>6,624 tokens</strong>, <strong>$2.561 billion</strong>;</li><li>organic-like total-volume advantage: only <strong>0.7%</strong>.</li></ul><p>But the per-token result becomes slightly stronger. Median 24-hour volume was <strong>$134,886</strong> for organic-like SOL launches versus <strong>$99,606</strong> for instant SOL launches, a <strong>35.4%</strong> difference. The organic-like subset produced roughly the same total volume with 27% fewer tokens.</p><p>So the careful reading is:</p><ul><li>quote mix contributes to the all-asset total-volume gap;</li><li>it does not explain the higher organic-like volume per token;</li><li>instant launches still dominate transaction count by a wide margin.</li></ul><h3>Why migration time is not enough</h3><p>A simple cutoff such as “under ten seconds” catches several different behaviors. The 1–2 second cohort had a median of seven buys and seven wallets. By 11–30 seconds, the median was already 125 buys and 84 wallets. This is why the classifier combines time, transaction count, buyer breadth, and concentration.</p><p>The result is stable around the chosen boundaries:</p><ul><li>raising the instant cap from three to five or ten pre-migration buys leaves median 24-hour volume near <strong>$100,000</strong>;</li><li>requiring organic-like latency of 10, 30, or 60 seconds leaves median volume between <strong>$131,000 and $134,000</strong>;</li><li>increasing the organic-like buyer floor from 20 to 50 produces the same ranking.</li></ul><p><strong>The exact thresholds determine cohort size, but they do not create the main result.</strong></p><h3>What this does — and does not — measure</h3><p>Pump’s documentation says canonical PumpSwap pools created through the Pump migration instruction use pool index zero.[1] Synthetic migration can let the completing buy trade against the future pool’s implied reserves before migrate_v2 creates the actual pool.[2] We therefore anchor the post-migration window to the decoded migration event, not to the final curve buy.</p><p>Several boundaries still matter:</p><ul><li><strong>Organic-like is an observed flow shape, not proof of organic demand.</strong> Hundreds of wallets can belong to coordinated operators.</li><li><strong>Transaction count is not economic volume.</strong> The instant cohort demonstrates exactly why the two should be reported separately.</li><li><strong>Volume does not establish unique users or market quality.</strong> Signing wallets are addresses, and repeated automated trades can inflate activity.</li><li><strong>The comparison is descriptive.</strong> Migration architecture may correlate with creator type, quote asset, distribution strategy, or promotion. We have not isolated a causal effect.</li><li><strong>The transition bucket remains large.</strong> It contains concentrated slow launches, Mayhem-mode tokens, and hybrids that fail one or more strict conditions.</li><li><strong>Pricing coverage is high in the primary cohorts.</strong> Historical quote prices cover 100% of instant swap rows and 99.1% of organic-like rows over the 24-hour horizon.</li></ul><h3>Takeaway</h3><p>“Fresh token” is too broad a category for PumpSwap.</p><p>One third of the equal-horizon migration cohort followed a strict instant pattern: same slot, one dominant buyer, almost immediate migration. These tokens generated enormous numbers of small PumpSwap transactions.</p><p>The organic-like group took several minutes to migrate, accumulated hundreds of pre-migration buyers, and produced fewer but larger PumpSwap trades. Its median 24-hour volume was roughly one third higher, including after controlling for native-SOL quote assets.</p><p>The follow-up question is no longer whether fresh tokens dominate PumpSwap. They do. The useful question is <strong>which migration architecture created the fresh token in the first place</strong>.</p><h3>Sources</h3><p>[1] PumpSwap program documentation — <a href="https://github.com/pump-fun/pump-public-docs/blob/639b87fb45c3adeffa6f7c4a1bf0cb85e392bf22/docs/PUMP_SWAP_README.md">https://github.com/pump-fun/pump-public-docs/blob/639b87fb45c3adeffa6f7c4a1bf0cb85e392bf22/docs/PUMP_SWAP_README.md</a></p><p>[2] Synthetic migration documentation — <a href="https://github.com/pump-fun/pump-public-docs/blob/639b87fb45c3adeffa6f7c4a1bf0cb85e392bf22/docs/SYNTHETIC_MIGRATION.md">https://github.com/pump-fun/pump-public-docs/blob/639b87fb45c3adeffa6f7c4a1bf0cb85e392bf22/docs/SYNTHETIC_MIGRATION.md</a></p><p><em>OnChain Divers provides decoded Solana data: launch, swap and migration data for researchers building lifecycle-aware market models. Subscribe for schema updates, reproducible SQL and delayed research examples.</em></p><p><strong>Connect with us:</strong></p><ul><li>Website: <a href="https://onchaindivers.com/">onchaindivers.com</a></li><li>Twitter: <a href="https://x.com/onchain_divers">@onchain_divers</a></li><li>Discord: <a href="https://discord.com/invite/PRWyXUvhU3">Join our community</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3cd0c019b7a2" width="1" height="1" alt=""><hr><p><a href="https://medium.datadriveninvestor.com/one-buyer-one-slot-then-pumpswap-3cd0c019b7a2">One Buyer, One Slot, Then PumpSwap</a> was originally published in <a href="https://medium.datadriveninvestor.com/">DataDrivenInvestor</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[One PumpFun Buy, Two Economic Phases]]></title>
            <link>https://medium.com/@hello_89216/one-pumpfun-buy-two-economic-phases-97b891f286a4?source=rss-1b5226afa4e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/97b891f286a4</guid>
            <category><![CDATA[backtesting]]></category>
            <category><![CDATA[blockchain]]></category>
            <category><![CDATA[solana-network]]></category>
            <category><![CDATA[memecoins]]></category>
            <dc:creator><![CDATA[Onchain Divers]]></dc:creator>
            <pubDate>Mon, 09 Nov 2026 17:46:06 GMT</pubDate>
            <atom:updated>2026-11-09T17:46:06.262Z</atom:updated>
            <content:encoded><![CDATA[<h4>How a synthetic migration appears on-chain — and how to read it in the indexer</h4><p>Transaction 3QWeY371…h6Gr looks like one PumpFun buy. Economically, it is two trades joined inside one instruction.</p><p>The first part finishes the PumpFun bonding curve. The second part buys against the reserves of a PumpSwap pool that does not exist yet. Two seconds later, the migration transaction creates that pool with exactly the reserves left by the second part.</p><p>This is Pump’s synthetic-migration path. The official documentation describes it directly: a completing v3 buy takes the tokens left on the curve, then buys the remainder from the tokens that would otherwise enter PumpSwap.[1]</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*zkphU2IG1S_oKuSq.png" /></figure><h3>The transaction</h3><p>The transaction landed at:</p><pre>slot:       462,471,715<br>UTC time:   2026-10-28 12:11:55<br>buyer:      C4Ln2veR4zhf9QUEa53Ru9xmGNuYRdWBUMhfG7cdUXir<br>mint:       8K1zd4a7N5fnz3ZdtV6WFmdyE3FohkF1FYyfstqopump<br>instruction: buy_exact_quote_in_v3</pre><p>The top-level caller was FLASHX8…BtxB9. It invoked the Pump program by CPI. The Pump logs emitted three events in order:</p><pre>TradeEvent<br>CompleteEvent<br>PostCompleteBuyEvent</pre><p>That sequence is the clearest way to recognize the two-phase buy.</p><h3>Phase 1: finish the bonding curve</h3><p>The TradeEvent contains the ordinary curve part:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*43iitfQMOXsP4HeU.png" /></figure><p>The curve’s token account held 211,809,670.460615 tokens before the transaction. After phase 1 removed 4,909,670.460615 tokens, exactly 206,900,000 tokens remained.</p><p>Those remaining tokens become the <strong>prospective PumpSwap base reserve</strong>. The curve is complete, but the pool has not been created yet.</p><p>In the indexer, this is the row with:</p><pre>is_post_trade = 0<br>base_coin_amount = 4,909,670,460,615<br>quote_coin_amount = 1,982,511,396</pre><p>On this row, virtual_token_balance_after and virtual_sol_balance_after retain their ordinary bonding-curve meaning.</p><h3>Phase 2: trade against the future pool</h3><p>The buyer’s quote budget was not exhausted when the curve completed. PostCompleteBuyEvent records the second part:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*sDRICB9ha-dX02Dx.png" /></figure><p>The arithmetic reconciles exactly:</p><pre>206,900,000,000,000 − 867,985,847,696<br>= 206,032,014,152,304 base units<br><br>84,990,359,673 + 358,053,237<br>= 85,348,412,910 lamports</pre><p>No PumpSwap pool account exists at this point. The program is calculating what the pool reserves will be after migration and applying the constant-product buy against that future state. Pump’s IDL exposes these values as pool_base_reserves_before/after and pool_quote_reserves_before/after on PostCompleteBuyEvent.[2]</p><p>In the indexer, this becomes a second row with the same signature and the same ix_idx:</p><pre>is_post_trade = 1<br>base_coin_amount = 867,985,847,696<br>quote_coin_amount = 358,053,237</pre><p>The field names now require care. On this post-trade row:</p><pre>virtual_token_balance_after = pool_base_reserves_after<br>virtual_sol_balance_after   = effective pool_quote_reserves_after</pre><p>They are no longer ordinary bonding-curve virtual reserves. They carry the future pool state produced by the synthetic part of the buy.</p><p>This is the “virtual reserve reduction” visible in indexed data: the second leg removes 867,985.847696 tokens from the base amount that would have entered PumpSwap. The migration therefore deposits 206.032014M tokens instead of 206.9M.</p><h3>The user’s total buy</h3><p>To reconstruct what the user bought, sum the two economic rows:</p><pre>base received:<br>4,909,670,460,615<br>+ 867,985,847,696<br>= 5,777,656,308,311 raw units<br>= 5,777,656.308311 tokens<br><br>quote into reserves, excluding fees:<br>1,982,511,396<br>+ 358,053,237<br>= 2,340,564,633 lamports<br>= 2.340564633 SOL</pre><p>The RPC token-balance delta confirms the token total exactly: the buyer’s newly created token account ended with 5,777,656.308311 tokens, while the bonding-curve token account fell by the same amount.</p><p>Fees are separate for each phase. The decoded events report:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*1MB1CeH6VH7zoZVs.png" /></figure><p>The two phases contain <strong>0.040374742 SOL</strong> of Pump-level fees in total. Do not use the indexer’s repeated fee = 53,000 field for this calculation: that field is the Solana transaction fee, not the Pump protocol fee.</p><p>The buyer’s top-level SOL balance changed by more than principal plus Pump fees because the outer transaction also created a Token-2022 associated account and executed router-level transfers. A wallet balance delta is therefore not a clean substitute for decoded Pump events.</p><h3>Migration carries the phase-2 state forward</h3><p>The actual migration landed two seconds and 11 slots later:</p><pre>migration signature:<br>BjbeyxrKAeTWDRSs53DpggMqb5biU7fJayugmbeHKfFtwtYvLoa9kUrVxUvZ6M15ePx7wt4ENiu2u3H555vdSmS<br><br>mint_amount = 206,032,014,152,304<br>sol_amount  = 85,348,412,910</pre><p>Both values match PostCompleteBuyEvent.pool_*_reserves_after exactly. This is the intended handoff: migrate_v2 creates the pool with the reserves left by the completing buy.[1]</p><p>The first indexed PumpSwap trade provides one more check:</p><pre>pool_base_token_reserves_before = 206,032,014,152,304<br>pool_quote_token_reserves_before = 67,763,907,494<br>virtual_quote_reserves           = 17,584,505,416</pre><p>For PumpSwap, quote pricing uses effective reserves:</p><pre>raw quote vault + virtual quote reserves<br>67,763,907,494 + 17,584,505,416<br>= 85,348,412,910</pre><p>That equals the migration’s sol_amount and the synthetic leg’s effective quote reserve after. Comparing only the raw quote vault would falsely suggest that 17.5845 SOL disappeared during migration.</p><h3>How to interpret the indexer</h3><p>The correct grain depends on the question.</p><h3>Counting user buy instructions</h3><p>Count the two rows as <strong>one buy instruction</strong>:</p><pre>uniqExact(tuple(signature, ix_idx))</pre><p>A plain count() reports two rows because it is counting economic legs, not user instructions.</p><h3>Calculating total bought amount</h3><p>Sum the amounts across both rows:</p><pre>SELECT<br>    signature,<br>    ix_idx,<br>    sum(base_coin_amount)  AS total_base_out,<br>    sum(quote_coin_amount) AS total_quote_into_reserves<br>FROM pumpfun_v2_swaps<br>WHERE signature = &#39;3QWeY371w8Rs4ZPBB1f4MeEgGt963PNvKWKTD3fMiphkonyYicWd6tLFJ14KkXbsKqdDeuLtPNSnzzdJQd59h6Gr&#39;<br>  AND failed = 0<br>GROUP BY signature, ix_idx;</pre><h3>Separating curve and synthetic-pool volume</h3><p>Use is_post_trade:</p><pre>sumIf(quote_coin_amount, is_post_trade = 0) AS curve_quote,<br>sumIf(quote_coin_amount, is_post_trade = 1) AS future_pool_quote</pre><p>This lets the same transaction contribute to two venue-state phases without pretending that two users traded.</p><h3>Reading reserve fields</h3><p>Use the row type before interpreting virtual_*_balance_after:</p><pre>multiIf(<br>    is_post_trade = 0, &#39;bonding-curve virtual reserves after&#39;,<br>    is_post_trade = 1, &#39;future PumpSwap effective reserves after&#39;,<br>    &#39;unknown&#39;<br>)</pre><p>Do not sum reserve fields across the rows. Each row is a different state transition.</p><h3>Transaction-level columns</h3><p>The following values repeat on both rows and must not be summed:</p><pre>provided_gas_fee<br>provided_gas_limit<br>consumed_gas<br>fee<br>signing_wallet<br>fee_payer<br>parent_program</pre><p>Collapse them with any, argMin, or max only after confirming within-instruction consistency.</p><h3>A safe transaction-level model</h3><p>For analytics, preserve both grains:</p><pre>instruction fact:<br>(signature, ix_idx)<br><br>leg fact:<br>(signature, ix_idx, is_post_trade)</pre><p>The instruction fact answers:</p><ul><li>how many buys occurred;</li><li>how many wallets traded;</li><li>total tokens received;</li><li>total quote committed to reserves.</li></ul><p>The leg fact answers:</p><ul><li>how much completed the curve;</li><li>how much traded against the future pool;</li><li>which reserve state belongs to each phase.</li></ul><p>Flattening immediately to one row loses the migration mechanics. Treating both rows as independent swaps doubles user activity and transaction-level fees.</p><h3>Takeaway</h3><p>This transaction is not a regular PumpFun buy followed by a PumpSwap trade. It is one buy_exact_quote_in_v3 instruction with two internal economic phases:</p><ol><li>buy the remaining bonding-curve inventory;</li><li>buy additional tokens from the reserves destined for the future PumpSwap pool.</li></ol><p>The is_post_trade field is the key to reading it. Sum the two rows for the user’s total buy, keep them separate for reserve accounting, and carry the post-trade reserve state into the migration. When the pool appears, compare PumpSwap’s <strong>effective</strong> quote reserves—raw vault plus virtual_quote_reserves—with the migration output.</p><p>That is how the curve, the synthetic pool trade, and the first real PumpSwap state reconcile without a gap.</p><h3>Sources</h3><p>[1] <a href="https://github.com/pump-fun/pump-public-docs/blob/639b87fb45c3adeffa6f7c4a1bf0cb85e392bf22/docs/SYNTHETIC_MIGRATION.md">https://github.com/pump-fun/pump-public-docs/blob/639b87fb45c3adeffa6f7c4a1bf0cb85e392bf22/docs/SYNTHETIC_MIGRATION.md</a> — Pump synthetic migration [2] <a href="https://github.com/pump-fun/pump-public-docs/blob/639b87fb45c3adeffa6f7c4a1bf0cb85e392bf22/idl/pump.json">https://github.com/pump-fun/pump-public-docs/blob/639b87fb45c3adeffa6f7c4a1bf0cb85e392bf22/idl/pump.json</a> — Pump program IDL</p><p><em>OnChain Divers provides decoded Solana data: launch, swap and migration data for researchers building lifecycle-aware market models. Subscribe for schema updates, reproducible SQL and delayed research examples.</em></p><p><strong>Connect with us:</strong></p><ul><li>Website: <a href="https://onchaindivers.com/">onchaindivers.com</a></li><li>Twitter: <a href="https://x.com/onchain_divers">@onchain_divers</a></li><li>Discord: <a href="https://discord.com/invite/PRWyXUvhU3">Join our community</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=97b891f286a4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Solana Votor: Votes Left the Transaction Stream]]></title>
            <link>https://medium.com/@hello_89216/solana-votor-votes-left-the-transaction-stream-280d98edec2e?source=rss-1b5226afa4e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/280d98edec2e</guid>
            <category><![CDATA[solana-network]]></category>
            <category><![CDATA[on-chain-data]]></category>
            <category><![CDATA[blockchain]]></category>
            <category><![CDATA[cryptocurrency]]></category>
            <dc:creator><![CDATA[Onchain Divers]]></dc:creator>
            <pubDate>Fri, 23 Oct 2026 15:43:29 GMT</pubDate>
            <atom:updated>2026-10-23T15:43:29.628Z</atom:updated>
            <content:encoded><![CDATA[<h4>Consensus moved. Execution did not.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*aMkdJpHYusHWpUzdvAGmOQ.png" /></figure><p>On October 21, 2026, Solana activated Alpenglow’s first phase, Votor. The feature activated at the start of epoch 1064, but the actual consensus handoff occurred at slot 459,653,000: the last TowerBFT block was 459,652,999, and the first produced Votor block was 459,653,008 at 13:54:03 UTC.[1]</p><p>Votor did <strong>not</strong> eliminate validator voting. It changed how votes travel and how they are represented. Under TowerBFT, validators submitted normal Vote Program transactions. Under Votor, validators send signed votes directly to one another and aggregate them into BLS quorum certificates. Routine consensus votes therefore no longer occupy transaction entries inside blocks.[1][6]</p><p>The accurate summary is:</p><blockquote><strong><em>Consensus votes are no longer ordinary ledger transactions. Validators still vote, and certificates derived from those votes determine which blocks become final.</em></strong></blockquote><h3>The short version</h3><p>Our equal 24-hour indexer windows found:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*58P-huw53Iufadn4.png" /></figure><p><strong>Most of the transaction-count break is mechanical.</strong> In 25 sampled blocks on each side, legacy vote transactions averaged <strong>742</strong> per block before Votor and <strong>0</strong> after it. This is not evidence that user throughput fell by 60%.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*cjYaq8Y-gdnq4k3U.png" /></figure><h3>What is Votor?</h3><p>Alpenglow replaces TowerBFT as Solana’s consensus protocol. Its first phase, <strong>Votor</strong>, handles voting and finalization. <strong>Rotor</strong>, the planned replacement for Turbine’s block-distribution system, is not part of this first deployment.[1][6]</p><p>Votor runs two finalization routes concurrently:</p><ul><li>an 80%-stake fast path that can finalize after one voting round;</li><li>a two-round path using 60%-stake thresholds.</li></ul><p>Whichever route completes first wins. The target is approximately <strong>150ms finality</strong>, rather than TowerBFT’s roughly 12.8-second process. This is finality latency, not block time.[1][10]</p><p>Transaction execution did not move off-chain. The SVM, transaction format, programs, state changes, failed transactions, and fee mechanics continue to operate normally.</p><h3>The exact activation boundary</h3><p>There were two distinct moments:</p><p><strong>Feature activation</strong><br>slot 459,648,000<br>2026–10–21 13:36:05 UTC</p><p><strong>Last TowerBFT / Alpenglow genesis block</strong><br>slot 459,652,999<br>2026–10–21 13:54:00 UTC</p><p><strong>Votor begins</strong><br>slot 459,653,000</p><p><strong>First produced Votor block</strong><br>slot 459,653,008<br>2026–10–21 13:54:03 UTC</p><p>Slots 459,653,000–459,653,007 were skipped. The first produced Votor block points directly to the certified TowerBFT genesis block.</p><p>Our primary census uses equal 24-hour UTC windows around the first indexed Votor block, rather than equal slot ranges. This avoids bias from changes in slot duration and produced-block count.</p><h3>Transaction count: a structural break, not a user exodus</h3><p>The indexer recorded <strong>441.6 million</strong> transactions in the 24 hours before the switch and <strong>178.4 million</strong> afterward.</p><p>The sampled block decomposition makes the cause visible:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/881/0*fmY6n7QD-UqDar4M.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*xi5pQo4pXkrb2UJy.png" /></figure><p>The disappearance of routine vote transactions is causally attributable to Votor. The simultaneous change in non-vote transactions is not. It may reflect user demand, bots, application activity, congestion, leaders, schedulers, or time-of-day effects.</p><p><strong>Any TPS dashboard that historically included votes must therefore be re-baselined at slot 459,653,000.</strong></p><h3>Block compute units</h3><p>The RPC sample measured actual meta.computeUnitsConsumed, not requested CU limits:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/881/0*lnNerXXnYerQh6h-.png" /></figure><p>Votes represented more than half of recorded transactions but only a small portion of consumed compute. <strong>They were numerous and comparatively cheap.</strong></p><p>Votor unquestionably removes their on-chain CU usage, signature verification, transaction entries, and associated bandwidth. It does not follow that all freed resources are immediately filled by user transactions. Solana blocks may instead be constrained by account locks, data size, scheduler behavior, network limits, or user demand.</p><p>The decline in sampled non-vote CU is correlational, not a measured causal effect of Votor.</p><h3>Fees and validator revenue</h3><p>Solana transaction fees contain:</p><ol><li>a base fee charged per required signature;</li><li>an optional priority fee derived from the transaction’s compute-budget configuration.[2]</li></ol><p>The base fee in this window was 5,000 lamports per signature.[2] Under TowerBFT, routine vote transactions paid signature fees. Once those votes left the ledger, that recurring stream disappeared.</p><p>In the RPC sample:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*lHouVDGjONxTp1TU.png" /></figure><p>The means are sensitive to fee spikes, so the corresponding median priority fees are also useful: <strong>9.57M</strong> lamports before versus <strong>7.94M</strong> after.</p><p>The complete indexer census measures the block’s leader Fee reward, not gross transaction fees. Median leader reward moved from <strong>0.00851 SOL</strong> to <strong>0.00676 SOL</strong> per block; total 24-hour leader rewards moved from <strong>6,453.55 SOL</strong> to <strong>5,600.17 SOL</strong>.</p><p>Priority-fee movement cannot be assigned to Votor from this event window. Priority fees respond to user demand, account contention, requested CU limits, bidding behavior, and scheduler conditions. Votor did not change the fee formula.[1][2]</p><p>Validators no longer pay one transaction fee for every consensus vote. Instead, admitted validators pay the current <strong>0.8 SOL Validator Admission Ticket per epoch</strong> from their vote account.[3] Older draft material mentioning 1.6 SOL per epoch is not the live value.</p><h3>What can be concluded?</h3><p>The clean effects are structural:</p><ol><li>Routine vote transactions fell to zero after the boundary.</li><li>Recorded transaction count fell sharply because those votes disappeared.</li><li>Vote signature fees and vote CU consumption disappeared from blocks.</li><li>Leader fee revenue fell over this particular comparison window.</li><li>Non-vote load and priority fees remained demand-driven; their changes are not causally identified by a before/after chart.</li><li>Validator voting continues through direct messages and aggregate certificates.</li></ol><h3>Appendix A: How the migration was made safe</h3><h4>A controlled handoff</h4><p>SIMD-0384 specified a migration boundary 5,000 slots after feature activation. During the transition, validators restricted TowerBFT blocks so ordinary user transactions could not enter the rollback region. They then waited for a strongly confirmed final TowerBFT ancestor and voted on it as the Alpenglow genesis block.[7]</p><p>Because transitional blocks carried no new user transactions, incompatible descendants could be discarded without rolling back newly accepted user activity.</p><h4>The 82% genesis certificate</h4><p>Validators switched only after receiving a BLS genesis certificate representing <strong>at least 82% of stake</strong>. The migration argument assumes at most 19% malicious stake: two contradictory certificates above 82% could not both be formed without honest validators signing conflicting genesis blocks.[7]</p><p>The certificate was rebroadcast and embedded in initial Alpenglow metadata. A validator starting from a snapshot could verify the migration-success state, while a lagging validator could recover after observing a valid finalization certificate.</p><h4>Prerequisites and rehearsals</h4><p>Before the switch:</p><ul><li>BLS public-key registration activated on mainnet;</li><li>registrations required proof of possession bound to the vote account, protecting aggregate signatures from rogue-key attacks;</li><li>validators without an eligible BLS key were excluded from consensus;</li><li>the migration was rehearsed on a community cluster, testnet, and devnet;</li><li>the initial migration required a compatible Agave-based client.[1][3][8]</li></ul><h4>Security model</h4><p>Alpenglow uses a “20+20” model. Its core safety assumption is <strong>strictly less than 20% Byzantine stake</strong>; the additional offline tolerance applies under the protocol’s stronger network conditions. This differs from conventional one-third-BFT assumptions.[6][10]</p><p>The white paper supplies safety and liveness arguments, but we found no separately published machine-checked proof or named external audit report.</p><h3>Appendix B: Can the votes still be explored?</h3><h4>Historical votes</h4><p>Yes. Before slot 459,653,000, routine votes remain ordinary historical transactions involving:</p><pre>Vote111111111111111111111111111111111111111</pre><p>They can be inspected using an archival getBlock, Solana Explorer, Solscan, or a transaction indexer.[5]</p><h4>Post-Votor individual votes</h4><p>Not through ordinary transaction history. Individual votes now:</p><ul><li>travel directly between validators;</li><li>live in validator consensus data structures;</li><li>do not receive transaction signatures;</li><li>do not appear in getTransaction or explorer transaction counts.</li></ul><p>Vote accounts still exist on-chain. Administrative changes, withdrawals, commission changes, and BLS-key registration can still produce Vote Program transactions. These are not routine per-slot consensus votes.</p><h4>Certificates and footers</h4><p>Agave’s Geyser interface provides an opt-in block-footer callback. A footer may expose a finalization certificate and reward-related certificates, together with the bank hash and producer metadata.[1]</p><p>Certificate fields are optional, and a footer normally carries the latest certificate the leader knew while constructing the block — not a certificate finalizing the same block. Standard JSON-RPC does not expose every Votor vote, and a full certificate feed currently requires validator or Geyser-level instrumentation.</p><p>The activation itself can be checked using the Alpenglow feature account, while compatible nodes expose the certified migration genesis through getAgGenesisCert.</p><h3>Methodology and limitations</h3><ul><li>Indexer census: every produced block in equal 24-hour UTC windows around the first indexed Votor block.</li><li>Event chart: hourly aggregates from the same Solana block indexer.</li><li>RPC decomposition: 25 evenly spaced blocks on each side, approximately one every 30 minutes.</li><li>Vote classification: presence of the native Vote Program in static transaction account keys.</li><li>Priority-fee estimate: meta.fee − 5,000 × required signatures, floored at zero.</li><li>CU metric: actual meta.computeUnitsConsumed; requested limits were not measured.</li><li>“Non-vote” includes user transactions and non-vote system transactions.</li><li>This is an event study, not a causal experiment.</li></ul><h3>Sources</h3><p>[1] Alpenglow | Solana — <a href="https://solana.com/upgrades/alpenglow">https://solana.com/upgrades/alpenglow</a></p><p>[2] Fees | Solana — <a href="https://solana.com/docs/core/fees">https://solana.com/docs/core/fees</a></p><p>[3] BLS Pubkey and the Validator Admission Ticket | Solana — <a href="https://solana.com/upgrades/bls-pubkey-vat">https://solana.com/upgrades/bls-pubkey-vat</a></p><p>[5] getBlock RPC | Solana — <a href="https://solana.com/docs/rpc/http/getblock">https://solana.com/docs/rpc/http/getblock</a></p><p>[6] SIMD-0326: Alpenglow — <a href="https://github.com/solana-foundation/solana-improvement-documents/blob/main/proposals/0326-alpenglow.md">https://github.com/solana-foundation/solana-improvement-documents/blob/main/proposals/0326-alpenglow.md</a></p><p>[7] SIMD-0384: Alpenglow migration — <a href="https://github.com/solana-foundation/solana-improvement-documents/blob/main/proposals/0384-alpenglow-migration.md">https://github.com/solana-foundation/solana-improvement-documents/blob/main/proposals/0384-alpenglow-migration.md</a></p><p>[8] SIMD-0387: BLS pubkey management — <a href="https://github.com/solana-foundation/solana-improvement-documents/blob/main/proposals/0387-bls-pubkey-management-in-vote-account.md">https://github.com/solana-foundation/solana-improvement-documents/blob/main/proposals/0387-bls-pubkey-management-in-vote-account.md</a></p><p>[10] Solana Alpenglow Consensus White Paper v1.1 — <a href="https://www.anza.xyz/alpenglow-1-1">https://www.anza.xyz/alpenglow-1-1</a></p><p><em>OnChain Divers provides decoded Solana data: launch, swap and migration data for researchers building lifecycle-aware market models. Subscribe for schema updates, reproducible SQL and delayed research examples.</em></p><p><strong>Connect with us:</strong></p><ul><li>Website: <a href="https://onchaindivers.com/">onchaindivers.com</a></li><li>Twitter: <a href="https://x.com/onchain_divers">@onchain_divers</a></li><li>Discord: <a href="https://discord.com/invite/PRWyXUvhU3">Join our community</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=280d98edec2e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Best Solana Historical Data for Backtesting and ML in 2026]]></title>
            <link>https://medium.com/@hello_89216/best-solana-historical-data-for-backtesting-and-ml-in-2026-85d852236de5?source=rss-1b5226afa4e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/85d852236de5</guid>
            <category><![CDATA[blockchain]]></category>
            <category><![CDATA[backtesting]]></category>
            <category><![CDATA[memecoins]]></category>
            <category><![CDATA[solana-network]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Onchain Divers]]></dc:creator>
            <pubDate>Thu, 22 Oct 2026 20:01:15 GMT</pubDate>
            <atom:updated>2026-10-22T20:01:15.044Z</atom:updated>
            <content:encoded><![CDATA[<h4>Three ways to get Solana trade data, seven providers to try, and what to check before you pay.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*VLpoM7fEon5Jn8hS.png" /></figure><p>Data is the part of a Solana backtest that takes the longest and gets the least attention. You need every trade, not candles; bonding-curve launches, not only migrated pools; and the fees and reserves that turn a signal into a realistic result. Few sources give you all of that for a full year, and a year of PumpSwap trades alone is over five billion rows.</p><p>This article covers what to look for in historical data, the forms it comes in, seven providers worth trying and what to check during a trial.</p><h3>What do you need from historical data?</h3><p>For backtesting:</p><ul><li><strong>Enough history, and knowing whether it stays.</strong> Data deep enough to cover your test period on the venues you trade. Some sources keep fixed history, others a rolling window, and it is worth knowing which before you build a test you plan to rerun.</li><li><strong>Individual trades.</strong> Candles hide who traded, in what order and at what size, which is where most Solana strategies live.</li><li><strong>Execution context.</strong> Pool reserves before and after each trade, fees, and the priority fee paid to land it. Without them, slippage and costs are guesses.</li><li><strong>Bonding-curve coverage.</strong> Most Pump.fun tokens never migrate to an AMM, so pool-only data misses most launches.</li></ul><p>For machine learning, on top of that:</p><ul><li><strong>Large exports.</strong> Training sets run to tens or hundreds of millions of rows, so pricing by request, credit or data point quickly becomes the limit.</li><li><strong>Raw rows.</strong> Features like order-flow imbalance or the share of flow from a given router are built from individual trades.</li><li><strong>Execution fields as features.</strong> Priority fees, compute units and the calling program often say more about who is trading than price does.</li><li><strong>A stable schema.</strong> If a column changes meaning halfway through your training window, features break silently.</li></ul><h3>Three forms Solana data comes in</h3><p><strong>Raw chain data.</strong> Transactions exactly as the chain recorded them, through RPC or streams. Everything is there, and turning instructions into trades is your job.</p><p><strong>Decoded trades you query yourself.</strong> Swaps already decoded into rows, in a database you query with SQL or delivered into your own warehouse. You compute whatever metrics you need from individual trades.</p><p><strong>Data through an API.</strong> Endpoints for prices, candles, trades and wallets. Quick to start with, and you work with the fields and formats each endpoint returns.</p><p>For backtesting and ML, decoded trades at the level of individual swaps are usually the most practical starting point. Raw data takes the most work to use, and API endpoints can make large exports slow and expensive.</p><h3>Ready-made options</h3><p>There is no single price for a 1-year+ of Solana data. What you pay depends on the form you take it in and how much you pull. Most APIs and query platforms charge by credits or API points, so cost grows with the volume you query and export. Raw chain data costs less to access, but storage, decoding and backfill are yours to run. Decoded trades in a database are usually a flat fee.</p><p><strong>The prices below are starting points, not the cost of a specific workload. Every provider offers a trial or a free plan, so test it on your own queries before you commit.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*xRi7V2ow2POgjz3y.png" /></figure><p><em>7 Solana historical data providers at a glance. Starting prices as listed on each provider’s public website in October 2026. Several providers discount annual billing, and plans change often: check the provider’s own site for current pricing and limits.</em></p><h3>Onchain Divers</h3><p>Real-time and historical decoded Solana trades with direct SQL access to the database (ClickHouse), not an API.</p><ul><li><strong>Pricing:</strong> flat monthly fee, $200/month</li><li><strong>Free trial:</strong> 1 week, full database access with no query limits</li><li><strong>Best for:</strong> trade-level backtests and ML on Pump.fun, Pumpswap, Raydium and other main Solana venues</li></ul><h3>Dune</h3><p>Decoded tables you query with SQL inside Dune, plus an API.</p><ul><li><strong>Pricing:</strong> credits, from $75/month (4,000 credits); $399/month for 25,000 credits; CSV export from $399</li><li><strong>Free plan:</strong> view-only</li><li><strong>Best for:</strong> exploration, dashboards and shared analysis</li></ul><h3>GoldRush</h3><p>Decoded data through an API; delivery to your own warehouse is priced separately.</p><ul><li><strong>Pricing:</strong> API credits, from $10/month (10,000 credits); $250/month for 300,000 credits</li><li><strong>Free trial:</strong> 14 days</li><li><strong>Best for:</strong> multi-chain projects and data in your own warehouse</li></ul><h3>Bitquery</h3><p>A multi-chain GraphQL API, with historical price and OHLCV series for Solana sold as an archive add-on.</p><ul><li><strong>Pricing:</strong> API points, from $49/month; the Solana historical archive add-on starts at $300/month</li><li><strong>Free trial:</strong> 7 days</li><li><strong>Best for:</strong> multi-chain projects and historical price and OHLCV series</li></ul><h3>Birdeye</h3><p>A REST API for Solana market data.</p><ul><li><strong>Pricing:</strong> plans with request limits, from $99/month</li><li><strong>Free plan:</strong> yes</li><li><strong>Best for:</strong> candles, token metrics and wallet lookups</li></ul><h3>Shrine API</h3><p>Real-time Solana trades, candles and launch streams, plus a trading API.</p><ul><li><strong>Pricing:</strong> free, 0.1 SOL/month or 1 SOL/month, with higher limits and faster streams on paid plans</li><li><strong>Free plan:</strong> yes</li><li><strong>Best for:</strong> live trades, 1-second candles, new launches and migrations.</li></ul><h3>Helius</h3><p>Raw chain data through RPC and streams. Helius sits apart from the rest: it is infrastructure rather than a dataset. It gives you fast access to the raw chain, and turning that into a history of trades is your pipeline’s job.</p><ul><li><strong>Pricing:</strong> credits by plan: $49, $499 or $999/month</li><li><strong>Free plan:</strong> yes</li><li><strong>Best for:</strong> building your own pipeline</li></ul><h3>What to check during a trial</h3><p>A week of real queries against real data tells you more than any comparison table, including this one. For the venues you care about, check:</p><ul><li><strong>History depth.</strong> How far back trade-level data goes, and whether that history is fixed or rolling.</li><li><strong>Bonding curves.</strong> Whether Pump.fun bonding-curve trades are included, not only migrated pools.</li><li><strong>Cost fields.</strong> Whether you get pool reserves, fees and priority fees, so slippage and costs are measured rather than guessed.</li><li><strong>Export.</strong> How long it takes to get a year of data into your own environment, and what limits that: credits, rate limits, row caps or query timeouts.</li><li><strong>Schema.</strong> Whether field meanings are documented and stay stable across the period you train on.</li></ul><h3>What about building your own indexer?</h3><p><strong>You can:</strong> the chain is public and archive RPC exists. In practice it means a separate decoder for every venue, from Pump.fun and PumpSwap to Raydium and Meteora, kept working as programs upgrade; weeks of backfill and about 1.9 TB of compressed storage for a single year of PumpSwap alone; and gaps and quirks that fail silently.</p><p>Your own indexer makes sense when you need venues or history nobody offers, or full control over every field. Otherwise the time usually goes further on the research itself.</p><h3>The short version</h3><p>For backtesting and ML on Solana, the headline price is rarely what decides it. What decides it is whether you get individual trades with the context to price them, history that covers your test period, and a way to pull it at the scale you work at. None of that is visible on a pricing page, and all of it shows up within a few days of real queries. Pick two or three options that fit your task, run your own queries in their trials, and let the data decide.</p><h3>Sources</h3><ul><li><a href="https://onchaindivers.com/">OnchainDivers</a></li><li><a href="https://dune.com/pricing">Dune pricing</a></li><li><a href="https://docs.dune.com/resources/credits-billing/how-credits-work">Dune: how credits work</a></li><li><a href="https://goldrush.dev/pricing">GoldRush pricing</a></li><li><a href="https://goldrush.dev/docs/goldrush-solana/warehouse/dex-swaps">GoldRush Solana warehouse: DEX swaps</a></li><li><a href="https://bitquery.io/pricing">Bitquery pricing</a></li><li><a href="https://birdeye.so/data-api/pricing">Birdeye Data API pricing</a></li><li><a href="https://shrine.trade/api/">Shrine API</a></li><li><a href="https://www.helius.dev/pricing">Helius pricing</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=85d852236de5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Pump.fun Has a Family Tree Now: Memecoins Can Quote Memecoins]]></title>
            <link>https://medium.com/@hello_89216/pump-fun-has-a-family-tree-now-memecoins-can-quote-memecoins-d51b572c7888?source=rss-1b5226afa4e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/d51b572c7888</guid>
            <category><![CDATA[cryptocurrency]]></category>
            <category><![CDATA[solana-network]]></category>
            <category><![CDATA[pumpfun]]></category>
            <category><![CDATA[blockchain]]></category>
            <category><![CDATA[memecoins]]></category>
            <dc:creator><![CDATA[Onchain Divers]]></dc:creator>
            <pubDate>Fri, 09 Oct 2026 08:42:02 GMT</pubDate>
            <atom:updated>2026-10-09T08:42:02.591Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*gMg66AILHKO5I98l.png" /></figure><h4>One missed H1B trade exposed a new multi-hop path. A self-join on coin creation found 4,020 Pump.fun launches priced in 239 other Pump.fun coins in eight days.</h4><h3>The transaction we missed</h3><p>At 2026–10–08 18:53:40 UTC, a trader sold <strong>2,938,659.066860 H1B</strong>. The trade did not go directly from <strong>H1B</strong> to <strong>SOL</strong>.[4] It took this route:</p><p>Hire 1 Billion Americans (H1B)</p><p>↓ <strong>Pump bonding curve</strong></p><p>Official Layoff Coin (LAYOFF)</p><p>↓ <strong>PumpSwap pool</strong></p><p>wrapped SOL</p><p>The bonding curve transferred 90,162.041872 LAYOFF into the next hop, and PumpSwap returned 0.165762793 WSOL before the temporary wrapped-SOL account was closed.[4]</p><p>The naming is almost too perfect: <strong>Hire 1 Billion Americans was literally priced in Official Layoff Coin.</strong> Both mints use Token-2022. The <strong>H1B</strong> creation instruction names <strong>LAYOFF</strong> as remaining account zero — the quote mint — and both coins were created by the same wallet.[5]</p><p>The supplied transaction is a version-1 Solana transaction, but version support was not the problem. Our swap table indexed 292,434 other version-1 signatures during the surrounding UTC hour. The gap was a <strong>new inner Pump instruction</strong>:</p><p>Instruction: MultiHopCurveSwap</p><p>Anchor discriminator: e19a7516d756f667</p><p>The route was nested inside PumpSwap’s MultiHopSwap instruction, whose discriminator is 2b644913e9f66f94. At our 20:30 UTC cutoff, the exact signature produced zero rows in pumpfun_v2_swaps. The launch itself was present in pumpfun_token_creation, including its quote mint. A live-table check found that multi_hop_curve_swap is decoded in general (16,091 rows across 15,265 signatures in the fixed review window). More importantly, the same H1B/LAYOFF curve already had another successful multi_hop_curve_swap at 19:39:24 UTC, with the correct base mint, quote mint, sell direction and PumpFun router parent. The supplied signature is waiting for backfill; this is not a live decoder or route-shape gap.</p><h3>What Pump.fun changed</h3><p>Pump.fun’s public documentation now says explicitly that create_v2 can pair a new coin with an <strong>existing Pump.fun coin</strong>, rather than only SOL, USDC, or a centrally listed quote asset.[1] The change appeared in the public-docs commit titled “v3 / v2 trades, multi-hop swap, pump coins as quote mints.”[3]</p><p>The quote coin can still be on its bonding curve or can already have migrated to PumpSwap. Pump.fun derives the child’s starting virtual quote reserves from the quote coin’s live curve or pool reserves, targeting roughly the same raise as a normal launch but denominating it in the parent coin.[1]</p><p>There is one important limit: <strong>the graph is currently only one level deep.</strong> If coin B is quoted in coin A, B cannot itself become the quote for coin C. Pump.fun rejects that path with CurveDepthExceeded.[1]</p><p>Trading is handled by a matching route. PumpSwap’s multi_hop_swap moves through two or more pools and curves atomically, keeps the intermediate coin out of the user’s wallet, applies one final slippage check, and emits one event per hop.[2] The official sell example is exactly the shape seen on-chain:</p><p>child coin → Pump quote coin → SOL</p><p>multi_hop_curve_swap is the Pump-program side of that route. The official docs say it is callable only by PumpSwap’s signed global-config PDA; integrators are expected to call PumpSwap’s outer multi_hop_swap, not invoke the curve method directly.[2]</p><p>This is different from the earlier quote-token expansion. Pump.fun V2 already supported configured quote assets such as PUMP, USDC, xStocks and other listed tokens. The October update makes a much larger set dynamic: a Pump.fun coin can now become the reserve currency for another Pump.fun coin.[1][3]</p><h3>How H1B’s migration threshold is calculated</h3><p>A token-paired curve does not keep an 85 SOL migration threshold. It completes when its own real_token_reserves reach zero. The quote target is calibrated once, at creation, in units of the selected parent coin.[1]</p><p>H1B’s CreateEvent recorded:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*1T7GJkrF8_2ZmXjk.png" /></figure><p>Selling all 793.1 million real H1B leaves 279.9 million virtual H1B on the curve:</p><pre>final_virtual_h1b<br>= 1,073,000,000 - 793,100,000<br>= 279,900,000 H1B</pre><p>The constant-product quote requirement is:</p><pre>net_quote_to_complete<br>= ceil(<br>    initial_virtual_quote<br>    × initial_real_token<br>    / final_virtual_token<br>  )<br><br>= ceil(<br>    13,298,774.196962<br>    × 793,100,000<br>    / 279,900,000<br>  )<br><br>= 37,682,235.854272 LAYOFF</pre><p>At completion, virtual quote reserves reach 50,981,010.051234 LAYOFF. The 37.68M figure is net quote entering the curve. A trader’s gross spend is higher once the applicable protocol, holder-reward and buyback fees are included.</p><p>Pump.fun derives that initial LAYOFF reserve from the parent coin’s own curve or PumpSwap pool. It asks how much LAYOFF a fee-free buy of approximately the normal 85.005359057 SOL raise would return, then scales that target back by the standard curve ratio. The result is stored in H1B’s bonding-curve account.[1]</p><p><strong>This calibration happens once.</strong> If LAYOFF later doubles or halves against SOL, H1B still needs roughly 37.68M LAYOFF to empty its curve. What changes is the threshold’s SOL or USD value. A user may enter through SOL, but the router must first produce LAYOFF; only the net LAYOFF sent into the H1B curve advances migration.</p><p>At finalized slot 454661844, fetched at 2026–10–08 21:14:51 UTC, H1B’s curve was still open:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/941/0*s9zvp3LMRzrp3eqw.png" /></figure><p>The two progress percentages differ because the curve is nonlinear. Early H1B is cheaper: 17.42% of real inventory had sold while only 5.22% of the final LAYOFF target had accumulated.</p><p>The final purchase can cross the threshold. A v3 buy or multi-hop curve hop first buys every real H1B left at the bonding-curve price and emits CompleteEvent. Any remaining quote input then trades against the reserves of the future PumpSwap pool before that pool exists. Pump.fun calls this synthetic migration. migrate_v2 later creates the pool with the post-purchase reserves, preserving price continuity.[6]</p><p>For a token-paired coin such as H1B, Pump.fun does not subtract the SOL migration fee from the raised LAYOFF when it constructs those synthetic pool reserves.[1][6]</p><h3>The self-join that finds them</h3><p>The relationship is already explicit in the creation table:</p><p>child mint → child.quote_coin → parent mint</p><p>That means no token-transfer inference is required. Deduplicate creation rows to one row per mint, then join the child’s quote_coin to another creation’s mint.</p><pre>WITH<br>    toDateTime(&#39;2026-10-01 00:00:00&#39;, &#39;UTC&#39;) AS start_ts,<br>    toDateTime(&#39;2026-10-08 20:30:00&#39;, &#39;UTC&#39;) AS end_ts,<br>    creations AS<br>    (<br>        SELECT<br>            replaceAll(toString(mint), char(0), &#39;&#39;) AS mint,<br>            argMin(replaceAll(toString(name), char(0), &#39;&#39;),<br>                   tuple(block_time, slot, tx_idx)) AS name,<br>            argMin(replaceAll(toString(symbol), char(0), &#39;&#39;),<br>                   tuple(block_time, slot, tx_idx)) AS symbol,<br>            argMin(replaceAll(toString(creator), char(0), &#39;&#39;),<br>                   tuple(block_time, slot, tx_idx)) AS creator,<br>            argMin(quote_coin,<br>                   tuple(block_time, slot, tx_idx)) AS quote_coin,<br>            min(block_time) AS created_at,<br>            argMin(signature,<br>                   tuple(block_time, slot, tx_idx)) AS creation_signature<br>        FROM default.pumpfun_token_creation<br>        WHERE block_time &gt;= toDateTime(&#39;2024-01-01 00:00:00&#39;, &#39;UTC&#39;)<br>          AND block_time &lt; end_ts<br>        GROUP BY mint<br>    )<br>SELECT<br>    child.created_at,<br>    child.mint AS child_mint,<br>    child.name AS child_name,<br>    child.symbol AS child_symbol,<br>    parent.mint AS quote_mint,<br>    parent.name AS quote_name,<br>    parent.symbol AS quote_symbol,<br>    child.creator = parent.creator AS same_creator<br>FROM creations AS child<br>INNER JOIN creations AS parent<br>    ON child.quote_coin = parent.mint<br>WHERE child.created_at &gt;= start_ts<br>  AND child.created_at &lt; end_ts<br>  AND parent.created_at &lt;= child.created_at<br>ORDER BY child.created_at DESC;</pre><p>The historical side of the CTE has an explicit lower bound because the parent coin may predate the child window by months. The final condition prevents a future or reprocessed parent row from being attached to an earlier child. Deduplication before the join is essential: joining raw creation rows can multiply one child when historical rows have been reprocessed.</p><h3>4,020 children and a launch graph that went vertical</h3><p>For the half-open window 2026–10–01 00:00:00 through 2026–10–08 20:30:00 UTC, the join returned:</p><ul><li><strong>4,020</strong> child launches quoted in another Pump.fun-created coin;</li><li><strong>239</strong> distinct quote-parent mints;</li><li><strong>2,094</strong> distinct child creators;</li><li>only <strong>88 launches, or 2.19%</strong>, where child and quote parent had the same creator;</li><li>a median quote-parent age of <strong>460.29 hours</strong>, about 19.2 days;</li><li><strong>717</strong> children whose quote parent was less than one hour old;</li><li><strong>1,010</strong> whose quote parent was less than one day old.</li></ul><p>The daily count sat between 322 and 453 from October 1 through October 7. Then October 8 produced <strong>1,376 launches by 20:30 UTC</strong>, despite being a partial day. The 19:00–20:00 UTC bucket alone contained <strong>695</strong> coin-on-coin launches.</p><p>The launches traded. Among 3,383 with a complete one-hour observation window, <strong>2,770 (81.88%)</strong> had at least one successful transaction already indexed in pumpfun_v2_swaps. The median across the completed cohort was four indexed transactions; 275 launches reached at least 100 and 32 reached at least 1,000. Treat those as lower bounds until the decoder covers the new multi-hop instruction. The supplied H1B sell is one known omission.</p><h3>The best child/parent combinations</h3><p>The self-join produces markets that read more like punchlines than trading pairs:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Q9QwyuuK6zGDPdl5.png" /></figure><p>The H1B/LAYOFF pair is especially useful because we have both sides of the lifecycle. H1B was created at 18:47:15 UTC with LAYOFF as its quote. Six minutes and 25 seconds later, the supplied transaction sold H1B through LAYOFF into SOL. The creation table found the market; the raw trace showed how it trades; and the missing swap row exposed the decoder gap.</p><p>The quote-parent graph is concentrated. The most reused parents were TikTok with 429 children, CLAW with 362, Agency with 319, PAID with 292, and ANSEM with 291. Yet the creator comparison shows that this is not mostly one wallet spawning its own descendants. Most children selected a quote coin launched by somebody else.</p><h3>What changes for data users</h3><p>A Pump.fun mint is no longer enough to define the economic market. For every child coin, a data product now needs:</p><ol><li>the child’s quote mint from creation;</li><li>the quote coin’s own SOL, USDC, or listed-token market;</li><li>both token decimal domains;</li><li>the route and venue used by each hop;</li><li>event coverage for both multi_hop_swap and multi_hop_curve_swap.</li></ol><p>A child price is first expressed in units of its parent coin. A dollar or SOL price is a composed price:</p><p>child / parent × parent / SOL × SOL / USD</p><p>The same rule applies to turnover and creator fees. Raw quote amounts cannot be summed across parent mints, and a parent coin’s volatility becomes part of the child’s price even when the child’s own curve does not move.</p><p>The event model also matters. Pump.fun documents one event per hop, while the intermediate parent coin nets to zero for the user.[2] Indexing only wallet deltas will see H1B leave and SOL arrive but can miss the real LAYOFF volume in the middle. Indexing only familiar trade discriminators can miss the entire route.</p><h3>The correction</h3><p>We missed the supplied transaction in the current historical slice. The live decoder already emits multi_hop_curve_swap rows for the same H1B/LAYOFF curve, so no new instruction or route-shape implementation is required. The remaining task is to let the backfill reach this signature and verify the reconstructed rows.</p><p>The creation layer was already enough to enumerate the new market graph. The raw transaction and Pump.fun’s updated docs identify the expected rows precisely. The indexer must decode every emitted hop event and preserve the child and parent mints rather than flattening the route into a synthetic H1B/SOL trade. For the supplied signature, that means a Pump curve leg for H1B/LAYOFF plus the connected PumpSwap leg for LAYOFF/WSOL.</p><p><strong>Pump.fun did not merely add another quote token. It made quote selection recursive across its own launch graph — one level deep for now — and gave the ecosystem an atomic router to trade through it.</strong></p><h3>Method and reproducibility</h3><ul><li>Launch window: [2026–10–01 00:00:00, 2026–10–08 20:30:00) UTC.</li><li>Parent scan: [2024–01–01 00:00:00, 2026–10–08 20:30:00) UTC.</li><li>Creation rows: deduplicated to one row per mint before the self-join.</li><li>Trade activity: successful rows, deduplicated by signature; first-hour statistics include only launches with a full hour before cutoff.</li><li>Supplied transaction: verified from finalized getTransaction responses in both json and jsonParsed encodings.</li><li>Complete SQL, fixed extracts, RPC responses, official source snapshots, plotting code and the dependency-free transaction verifier are included with this article.</li></ul><h3>Sources</h3><p>[1] Creating a Coin Paired With a Pump Coin — Pump.fun — <a href="https://github.com/pump-fun/pump-public-docs/blob/8cda1fa30ea658b20909d8aedf002047119388d2/docs/instructions/CREATE_WITH_PUMP_COIN_QUOTE.md">https://github.com/pump-fun/pump-public-docs/blob/8cda1fa30ea658b20909d8aedf002047119388d2/docs/instructions/CREATE_WITH_PUMP_COIN_QUOTE.md</a></p><p>[2] Multi-Hop Swap — Pump.fun — <a href="https://github.com/pump-fun/pump-public-docs/blob/8cda1fa30ea658b20909d8aedf002047119388d2/docs/instructions/MULTI_HOP_SWAP.md">https://github.com/pump-fun/pump-public-docs/blob/8cda1fa30ea658b20909d8aedf002047119388d2/docs/instructions/MULTI_HOP_SWAP.md</a></p><p>[3] Pump.fun public docs: pump coins as quote mints — <a href="https://github.com/pump-fun/pump-public-docs/commit/8cda1fa30ea658b20909d8aedf002047119388d2">https://github.com/pump-fun/pump-public-docs/commit/8cda1fa30ea658b20909d8aedf002047119388d2</a></p><p>[4] H1B → LAYOFF → SOL multi-hop sell — <a href="https://solscan.io/tx/4GrascJJFodHXJ7Hac1cK1V776uit8pZybR4oumEApWLEqiEnq8JMDtAo62AQVuqjafVzDpW12zkzr822hUmwxWq">https://solscan.io/tx/4GrascJJFodHXJ7Hac1cK1V776uit8pZybR4oumEApWLEqiEnq8JMDtAo62AQVuqjafVzDpW12zkzr822hUmwxWq</a></p><p>[5] H1B coin creation and initial routed buy — <a href="https://solscan.io/tx/552v4rwMY3gcoBEUTM8PjuvZ3MetR6Cnosrrs83HYziSAJg4LnsjUnED8gTs9DYzYp5vxwGByPt9cVgGHkT2ymXy">https://solscan.io/tx/552v4rwMY3gcoBEUTM8PjuvZ3MetR6Cnosrrs83HYziSAJg4LnsjUnED8gTs9DYzYp5vxwGByPt9cVgGHkT2ymXy</a></p><p>[6] Synthetic Migration — Pump.fun — <a href="https://github.com/pump-fun/pump-public-docs/blob/8cda1fa30ea658b20909d8aedf002047119388d2/docs/SYNTHETIC_MIGRATION.md">https://github.com/pump-fun/pump-public-docs/blob/8cda1fa30ea658b20909d8aedf002047119388d2/docs/SYNTHETIC_MIGRATION.md</a></p><p><em>OnChainDivers provides indexed Solana launch, swap and migration data for researchers building lifecycle-aware market models. Subscribe for schema updates, reproducible SQL and delayed research examples.</em></p><p><strong>Connect with us:</strong></p><ul><li>Website: <a href="https://onchaindivers.com/">onchaindivers.com</a></li><li>Twitter: <a href="https://x.com/onchain_divers">@onchain_divers</a></li><li>Discord: <a href="https://discord.com/invite/PRWyXUvhU3">Join our community</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d51b572c7888" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Pump.Fun quietly introduced a router and a V3 family]]></title>
            <link>https://medium.com/@hello_89216/pump-fun-quietly-introduced-a-router-and-a-v3-family-74e8de82cdd5?source=rss-1b5226afa4e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/74e8de82cdd5</guid>
            <category><![CDATA[memecoins]]></category>
            <category><![CDATA[blockchain]]></category>
            <category><![CDATA[solana-network]]></category>
            <category><![CDATA[backtesting]]></category>
            <category><![CDATA[cryptocurrency]]></category>
            <dc:creator><![CDATA[Onchain Divers]]></dc:creator>
            <pubDate>Mon, 05 Oct 2026 18:53:13 GMT</pubDate>
            <atom:updated>2026-10-05T18:53:13.905Z</atom:updated>
            <content:encoded><![CDATA[<h4>On October 2, PumpFun upgraded three core programs and introduced a new native router path for quote-token-aware trades before the methods appeared in the public IDLs</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*SNrkwykKxeB8FcWQ.png" /></figure><h3>Detection</h3><p>Our PumpFun detector compares every instruction with a discriminator taxonomy built from the on-chain IDL. Unknown instructions are logged instead of being silently ignored.</p><p>At `2026–10–02 16:15:48 UTC`, it encountered a new discriminator on the PumpFun bonding-curve program:</p><pre>e1f7501ed5b38488</pre><p>The program log named it:</p><pre>Instruction: BuyExactQuoteInV3</pre><p>We independently confirmed the name by hashing the Anchor instruction namespace:</p><pre>sha256(b”global:buy_exact_quote_in_v3”)[:8].hex()<br><br># e1f7501ed5b38488</pre><p>The detector flagged the instruction 16 seconds after the first transaction we observed on mainnet. The trade still emitted a normal <em>TradeEvent</em>, so nothing looked broken on-chain. The instruction itself was new and absent from PumpFun’s published IDL.</p><p>The new activity followed a coordinated deployment. PumpFun upgraded `pump_amm`, `pump`, and `pump_fees` within 32 seconds, then upgraded a routing program 12 minutes later.</p><h3>What it turned out to be</h3><p>`buy_exact_quote_in_v3` does **not** introduce token-quoted curves from scratch. V2 already supported them. The change is how PumpFun’s own router connects a user’s input asset to those curves.</p><p>A typical V3 route is:</p><pre>SOL<br><br>  ↓<br><br>Jupiter, PumpSwap, or another AMM<br><br>  ↓<br><br>quote token<br><br>  ↓<br><br>PumpFun token-quoted bonding curve</pre><p>In one mainnet transaction, the PumpFun router sent SOL through Jupiter, received a non-standard quote token, and passed the routed amount into `buy_exact_quote_in_v3`. The user received the bonding-curve token without retaining the intermediate quote asset.</p><h4>How this differs from V2: the SPX example</h4><p>V2 could already trade a bonding curve quoted in SPX. A transaction we inspected on October 3 used the following construction:</p><pre>Top-level Jupiter instruction: SOL → SPX<br><br>User’s SPX token account: receives the routed SPX<br><br>Top-level Pump instruction: BuyExactQuoteInV2 spends SPX<br><br>User’s token account: receives the PumpFun token</pre><p>The route was atomic, but the client had to compose Jupiter and Pump as separate top-level instructions. The user’s associated token account acted as the handoff between the two programs, and the Pump instruction carried a fixed quote-token budget prepared before execution.</p><p>In the inspected transaction, Jupiter delivered `277.38145186 SPX`. `BuyExactQuoteInV2` then spent `274.60763782 SPX`, including fee transfers, leaving `2.77381404 SPX` in the user’s account. That remainder is evidence of the prebuilt two-step route, not a protocol failure.</p><p>The inspected V3 route had a different shape:</p><pre>Top-level PumpFun router instruction<br>  ├─ CPI to Jupiter: SOL → route assets → quote token<br>  └─ CPI to Pump: BuyExactQuoteInV3</pre><p>The router handled the quote token inside its own call. Jupiter produced `497.904744` units of the target quote token; V3 consumed the same `497.904744` units across the curve purchase and fee transfers. The user received the base token and did not finish with quote-token dust.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jy9zCaPaHiRC5lT9Vu6iNw.png" /></figure><p>So the public-facing change is **<strong>integration, not first-time SPX support</strong>**. V2 proved that SPX-quoted curves already worked. V3 makes the same class of trade native to PumpFun’s router and easier to execute from SOL across more quote assets.</p><p>This distinction also changes how the first adoption numbers should be read. V3 buys used 55 quote mints in our initial sample. The older V2 buys routed by this specific PumpFun program used only SOL and USDC, but V2 activity across direct calls, Jupiter, and third-party routers already included SPX and other non-standard quotes.</p><p>We did not observe an SPX-denominated V3 buy in the reviewed window, so SPX is the verified V2 comparison — not evidence that every older quote mint had already moved to V3.</p><p><strong>Verification transactions:</strong></p><p>- V2 SPX route: [`3FzWcn…pmhp`](<a href="https://explorer.solana.com/tx/3FzWcnDQTFK5rhTJvwwhhm99pB6oJB6ZYrATEEUEGNztTzvGpkJPNQQkLbLFNvPi62TC3giRY9tL33DwxtX5pmhp">https://explorer.solana.com/tx/3FzWcnDQTFK5rhTJvwwhhm99pB6oJB6ZYrATEEUEGNztTzvGpkJPNQQkLbLFNvPi62TC3giRY9tL33DwxtX5pmhp</a>)</p><p>- V3 routed buy: [`55C1rB…kZ2Y`](<a href="https://explorer.solana.com/tx/55C1rB8uhHQDD2ToiHwL7KLLYr2m8tWmAmLjP3UncxRcBG5td2JwcKZM1oq9zz4jiDthdmrKxMCPbsuDMLqkkZ2Y">https://explorer.solana.com/tx/55C1rB8uhHQDD2ToiHwL7KLLYr2m8tWmAmLjP3UncxRcBG5td2JwcKZM1oq9zz4jiDthdmrKxMCPbsuDMLqkkZ2Y</a>)</p><p>We have added `buy_exact_quote_in_v3` to `pumpfun_v2_swaps`:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/761/1*cz3gtsMf41qwLc6NmNNjaQ.png" /></figure><p>The rows use the existing schema. The `parent_program` field identifies routed traffic.</p><h3>Other updates in the same deployment</h3><p>The upgraded binaries contain several methods that were not present in the public IDLs.</p><h4>New bonding-curve methods</h4><p>buy_exact_quote_in_v3</p><p>sell_v3</p><p>buy_v3</p><h4>New PumpSwap methods</h4><p>buy_exact_quote_in_v2</p><p>sell_v2</p><p>buy_v2</p><p>The PumpSwap names match the quote-token-aware V2 interface already used by the bonding curve.</p><h4>Quote-mint-aware fee lookup</h4><p>PumpSwap now calls get_fees_with_quote_mint on pump_fees during nearly every swap. In a fixed 200-block scan we found:</p><p>get_fees_with_quote_mint: <strong>20,835 calls</strong></p><p>old get_fees: <strong>6 calls</strong></p><p>The fee tier is returned as program return data rather than a trade event. This allows fee selection to depend on the quote mint instead of assuming a SOL-denominated pool.</p><h4>Fee collection methods</h4><p>Both pump and pump_amm now contain:</p><p>sweep_protocol_fee</p><p>sweep_creator_fee</p><p>The bonding-curve versions were already running as top-level mainnet transactions during our review. pump_fees also gained the admin methods set_exotic_flat_fees and admin_cto_sharing_config.</p><h3>The program using the new routes</h3><p>Most early V3 activity came through:</p><p>6Vo3245eszAb5wuqEMw8mGdbfRUdKbHhDHP5LcaGuTAB</p><p>This is a PumpFun-owned routing program. Its upgrade authority is:</p><p>7gZufwwAo17y5kg8FMyJy2phgpvv9RSdzWtdXiWHjFr8</p><p>The same authority controls all three upgraded PumpFun programs:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IDP50_nw6qsO8MG7nwz3Dg.png" /></figure><p><strong>The shared authority and deployment timing show that the router and the new methods shipped together.</strong></p><p>The router can interact with:</p><ul><li>PumpFun bonding curves and PumpSwap;</li><li>Jupiter;</li><li>Raydium AMM, CLMM, CPMM, and LaunchLab;</li><li>Meteora DAMM, DBC, and DLMM;</li><li>limit-order creation, execution, and cancellation.</li></ul><p>In a sample of 120 router transactions, its most common calls were:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/781/1*N5TZH268A6LH3A5vRLU3cg.png" /></figure><p>A small amount of V3 traffic later appeared outside the primary router, indicating that other integrations had started to adopt the new methods.</p><h3>What changed for the data</h3><p>PumpFun’s October 2 deployment changed more than one instruction:</p><ul><li>PumpFun’s own router can now reach token-quoted bonding curves through a wider set of input routes;</li><li>PumpSwap received matching quote-token-aware methods;</li><li>fee calculation now accepts the quote mint;</li><li>a PumpFun-owned router connects these methods to several Solana venues;</li><li><strong>the changes were live on-chain before they appeared in the public IDLs.</strong></li></ul><p>The first supported V3 route is now indexed without a schema change. Existing queries can distinguish the routed flow through parent_program and read the actual quote mint from each swap row.</p><p><em>Sources: on-chain program binaries and upgrade authorities fetched on October 3, 2026; PumpFun public IDLs (</em><a href="https://github.com/pump-fun/pump-public-docs"><em>https://github.com/pump-fun/pump-public-docs</em></a><em>); mainnet transaction traces; and a fixed 200-block scan covering 209,904 transactions.</em></p><p><em>OnChainDivers provides indexed Solana launch, swap and migration data for researchers building lifecycle-aware market models. Subscribe for schema updates, reproducible SQL and delayed research examples.</em></p><p><strong>Connect with us:</strong></p><ul><li>Website: <a href="https://onchaindivers.com/">onchaindivers.com</a></li><li>Twitter: <a href="https://x.com/onchain_divers">@onchain_divers</a></li><li>Discord: <a href="https://discord.com/invite/PRWyXUvhU3">Join our community</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=74e8de82cdd5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Best Robinhood Chain Nodes for Pons Sniping (2026)]]></title>
            <link>https://medium.datadriveninvestor.com/best-robinhood-chain-nodes-for-pons-sniping-2026-7ff54807c4bc?source=rss-1b5226afa4e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/7ff54807c4bc</guid>
            <category><![CDATA[robinhood-crypto]]></category>
            <category><![CDATA[cryptocurrency]]></category>
            <category><![CDATA[memecoins]]></category>
            <category><![CDATA[backtesting]]></category>
            <category><![CDATA[blockchain]]></category>
            <dc:creator><![CDATA[Onchain Divers]]></dc:creator>
            <pubDate>Mon, 28 Sep 2026 19:49:30 GMT</pubDate>
            <atom:updated>2026-09-29T11:32:42.531Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*S08e5h0NTqzuhw1r.png" /></figure><p>On Pons, the money is made or lost in the first few blocks. The protocol charges a snipe tax that starts near 99% in the launch second, falls to about 6% one second later and to under 0.2% the second after that, while Robinhood Chain produces a block roughly every 100 ms. Your infrastructure decides which of those blocks you land in.</p><p>This guide covers what you actually need, who sells it, and what it costs.</p><h3>What you need: three different jobs</h3><p>Robinhood Chain is an Arbitrum Orbit (Nitro) chain, ID 4663, with ETH as gas. There is no public mempool: a transaction is invisible until sequencer forms it into a block. That changes the shopping list.</p><figure><img alt="Table of the three layers of a Pons sniping setup on Robinhood Chain: sequencer feed, RPC and archive RPC." src="https://cdn-images-1.medium.com/max/1024/0*eyohBvkVzRJ_qWGc.png" /><figcaption>The three layers of a Pons sniping setup.</figcaption></figure><p>These are usually three different products. Buying one $49 RPC plan and expecting it to cover all three is the most common mistake.</p><h3>1. Sequencer feed — the signal</h3><p>The sequencer broadcasts every ordered transaction over a WebSocket feed. Everything else — RPC log subscriptions, explorers, indexers — has to re-execute the block before it can tell you anything, so it always hears about a launch later. For sniping, the feed is where the race happens.</p><p>The free option is Robinhood’s public feed, wss://feed.mainnet.chain.robinhood.com. It works, but it is not fast, has no SLA, and in every test we have run it arrives behind the paid feeds. If several bots need it, run one relay and fan out locally rather than opening a connection per bot.</p><h3>Paid feed providers</h3><figure><img alt="Comparison of paid Robinhood Chain sequencer feeds: Eira x Divers Pulse gRPC, Node1 and BlockRazor Direct Ultra, with region, price and trial." src="https://cdn-images-1.medium.com/max/1024/0*dowadn3amWIX5-Oz.png" /><figcaption>Paid sequencer feeds for Robinhood Chain. Prices checked 28 September 2026.</figcaption></figure><ul><li><strong>Eira x Divers (Pulse gRPC)</strong> — gRPC stream of decoded transactions with execution results, plus RPC; no node needed. AWS Ohio. $200/day, $900/week or $2,000/month. 2-hour free trial. DM on <a href="https://t.me/inventandchill">Telegram</a></li><li><strong>Node1</strong> — transaction feed (one transaction per message) or Nitro-compatible native feed. Ohio. $200/day, $900/week or $2,000/month. <a href="https://node1.gitbook.io/node1-docs/robinhood/robinhood-feed">Plans</a></li><li><strong>BlockRazor Direct Ultra</strong> — direct WebSocket feed; no node needed. Ohio and Tokyo. $200/day or $2,000/month. <a href="https://docs.blockrazor.io/streams/node-stream/robinhood-chain">Plans</a></li></ul><p>At the top tier, pricing has converged: all three charge $2,000 a month, and the daily passes make it cheap to test before committing. What you are really comparing is how early each one reaches <strong>your</strong> server.</p><h3>Our latest measurements</h3><p><em>Last updated: September 28, 2026 · measured from AWS Ohio (us-east-2)</em></p><p>Robinhood feeds change all the time, so we post updated numbers when the picture shifts. As always: take a trial and test on your own setup. Results can be completely different on different servers and in different locations.</p><figure><img alt="Robinhood Chain sequencer feed latency measured from AWS Ohio on 28 September 2026: first-seen share and p50, p90, p99 delay for Eira x Divers, Node1, BlockRazor Ultra and the Robinhood public feed." src="https://cdn-images-1.medium.com/max/1024/0*l46un7GwfORNxGxn.jpeg" /><figcaption>One run from AWS Ohio, 28 September 2026. Your results will differ — test from your own server.</figcaption></figure><p>“First seen” is the share of transactions each source delivered first. p50/p90/p99 are milliseconds behind whichever source was first for that transaction; 0 means it was first. The run included other sources not listed here, so the first-seen figures do not add up to 100%.</p><p><strong>How to read this.</strong> These numbers are one run, from one box, in one region. They describe our setup, not yours. A server in a different availability zone, a different data center or a different continent can reorder this table entirely. Every vendor’s benchmark shows that vendor winning, ours included — which is exactly why the only numbers worth trusting are the ones you measure yourself. Take the trials, run them side by side from the server your bot will actually use, across busy and quiet hours, and decide on your own data.</p><h3>2. RPC — reads and submission</h3><p>Every provider below serves Robinhood Chain mainnet. For sniping, RPC is mostly your submission path: put it in the same region as your feed and your bot, and keep a second provider as a fallback.</p><figure><img alt="Robinhood Chain RPC providers compared: Alchemy, Chainstack, QuickNode and dRPC free tiers, entry paid plans and requests per second." src="https://cdn-images-1.medium.com/max/1024/0*T4Xfwtj1U3hg4K2H.png" /><figcaption>RPC providers for Robinhood Chain. Prices checked 28 September 2026.</figcaption></figure><ul><li><a href="https://www.alchemy.com/"><strong>Alchemy</strong></a> — free forever with 30M compute units a month at 25 requests per second. Paid is pay as you go at $0.525 per 1M CU, from 300 RPS.</li><li><a href="https://chainstack.com/build-better-with-robinhood-chain/"><strong>Chainstack</strong></a> — free forever with 3M requests a month at 25 RPS. Growth plan is $49/month with 20M requests at 250 RPS.</li><li><a href="https://www.quicknode.com/chains/robinhood"><strong>QuickNode</strong></a> — 1-month free trial with 10M credits at 15 RPS, no card needed. Build plan is $49/month ($34 billed annually) with 80M credits at 50 RPS.</li><li><a href="https://drpc.org/chainlist/robinhood-mainnet-rpc"><strong>dRPC</strong></a> — free forever with 210M CU a month on public nodes only, at 100 RPS. Paid is pay as you go at $6 per 1M requests, up to 5,000 RPS.</li></ul><h3>3. Archive RPC — history</h3><p>An archive node keeps every historical state, not only recent blocks. You do not need it to fire a transaction, but you need it to know whether your strategy works:</p><ul><li><strong>Backtesting filters.</strong> Minimum liquidity, creator wallet age, timing against the tax curve — each rule is a guess until you replay it against thousands of past launches.</li><li><strong>Debugging bad fills.</strong> Tracing a failed or expensive snipe shows what the curve looked like and whether you lost on latency or on sizing.</li><li><strong>Wallet intelligence.</strong> Dev-wallet tracking, repeat-launcher detection and copy-trading all depend on history beyond the last few blocks.</li><li><strong>P&amp;L.</strong> Accurate accounting over thousands of small fills is a historical-state problem.</li></ul><p>Unlike the feed, archive is not latency-critical, so location matters much less. What matters is price per call and limits, because backtesting traffic is bursty.</p><figure><img alt="Robinhood Chain archive RPC options compared: Onchain Divers, Chainstack, Alchemy and QuickNode, with archive access, price and limits." src="https://cdn-images-1.medium.com/max/1024/0*Cs6N4amr8ip56yVS.png" /><figcaption>Archive RPC for Robinhood Chain. Prices checked 28 September 2026.</figcaption></figure><ul><li><a href="https://docs.google.com/forms/d/e/1FAIpQLSduVHVt51hEMGD2847Yj0rQTy1U5ogIgHPHutKyZth0ALhrwA/viewform"><strong>Onchain Divers</strong></a> — Robinhood archive node for $150/month with no request limits, hosted in Frankfurt. An Arc archive, an Ethereum node and an Ethereum beacon node are included as a bonus.</li><li><a href="https://chainstack.com/build-better-with-robinhood-chain/"><strong>Chainstack</strong></a> — archive from the Growth plan ($49/month, 20M requests), with debug and trace. An archive call counts as 2 requests, so that is about 10M archive calls at 250 RPS.</li><li><a href="https://www.alchemy.com/"><strong>Alchemy</strong></a> — archive on all plans, including free (30M CU a month), then $0.525 per 1M CU. Debug and trace from pay as you go. 25 RPS free, 300 RPS paid.</li><li><strong>QuickNode</strong> — archive on all plans, including the trial (10M credits). Build is $49/month with 80M credits. Debug and trace from Build. 15 RPS on trial, 50 RPS on Build.</li></ul><p>Metered plans charge per call, and archive and trace methods usually cost more than a standard request. If you plan to replay launches at scale, work out your monthly call volume first — a flat unlimited plan and a metered one can differ by an order of magnitude once backtesting gets serious.</p><p>Keep archive traffic on a separate key or plan from your live path. Backtesting will eat a plan sized for execution.</p><h3>Putting it together</h3><figure><img alt="Recommended Robinhood Chain sniping setup at each stage, from free learning setup to competing for the first block: feed, RPC and archive choices." src="https://cdn-images-1.medium.com/max/1024/0*tc-HNh_2xU3fIlDp.png" /><figcaption>Which setup at which stage.</figcaption></figure><p><em>Latency figures in this article come from our own test in AWS Ohio and will differ on other servers and in other locations. Prices and plans were checked on September 28, 2026 and change often — confirm on each provider’s site. This article is not financial advice. Sniping new token launches is high-risk and most new tokens lose value; do your own research (DYOR), and never trade with money you cannot afford to lose.</em></p><h3>Sources</h3><ul><li>Robinhood Chain docs: <a href="https://docs.robinhood.com/chain/">https://docs.robinhood.com/chain/</a></li><li>Eira x Divers pricing: <a href="https://eiranodes.dev/pricing">https://eiranodes.dev/pricing</a> (DM and talk to us directly <a href="https://t.me/inventandchill">here</a>)</li><li>Node1 Robinhood Feed: <a href="https://node1.gitbook.io/node1-docs/robinhood/robinhood-feed">https://node1.gitbook.io/node1-docs/robinhood/robinhood-feed</a></li><li>BlockRazor Direct Sequencer Feed (Ultra): <a href="https://docs.blockrazor.io/streams/node-stream/robinhood-chain/direct-sequencer-feed-ultra">https://docs.blockrazor.io/streams/node-stream/robinhood-chain/direct-sequencer-feed-ultra</a></li><li>Alchemy pricing: <a href="https://www.alchemy.com/pricing">https://www.alchemy.com/pricing</a></li><li>Chainstack pricing: <a href="https://chainstack.com/pricing/">https://chainstack.com/pricing/</a></li><li>QuickNode pricing: <a href="https://www.quicknode.com/pricing">https://www.quicknode.com/pricing</a></li><li>dRPC pricing: <a href="https://drpc.org/pricing">https://drpc.org/pricing</a></li></ul><p><strong>Connect with us:</strong></p><ul><li>Website: <a href="https://onchaindivers.com/">onchaindivers.com</a></li><li>Twitter: <a href="https://x.com/onchain_divers">@onchain_divers</a></li><li>Discord: <a href="https://discord.com/invite/PRWyXUvhU3">Join our community</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7ff54807c4bc" width="1" height="1" alt=""><hr><p><a href="https://medium.datadriveninvestor.com/best-robinhood-chain-nodes-for-pons-sniping-2026-7ff54807c4bc">Best Robinhood Chain Nodes for Pons Sniping (2026)</a> was originally published in <a href="https://medium.datadriveninvestor.com/">DataDrivenInvestor</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Pump.fun’s quote-token expansion is live — and indexed]]></title>
            <link>https://medium.com/@hello_89216/pump-funs-quote-token-expansion-is-live-and-indexed-654ff8ccb68e?source=rss-1b5226afa4e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/654ff8ccb68e</guid>
            <category><![CDATA[backtesting]]></category>
            <category><![CDATA[memecoins]]></category>
            <category><![CDATA[solana-network]]></category>
            <category><![CDATA[cryptocurrency]]></category>
            <category><![CDATA[blockchain]]></category>
            <dc:creator><![CDATA[Onchain Divers]]></dc:creator>
            <pubDate>Thu, 10 Sep 2026 09:09:07 GMT</pubDate>
            <atom:updated>2026-09-10T09:09:07.991Z</atom:updated>
            <content:encoded><![CDATA[<p>Pump.fun no longer starts every market against SOL.</p><p>Its current launch program carries an on-chain QuoteControl registry with 93 assets: xStocks, commodities, crypto assets, individual equities and PUMP itself. We decoded the account yesterday. Today we can show the other half of the story: the <strong>new quote layer</strong> is already flowing through the OnChainDivers indexer, from token creation to every indexed V2 swap.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*_1Gz_z5Ms1I-CclT.png" /></figure><h3>What shipped in the indexer</h3><p>Two source tables now expose the quote asset directly:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*w-3-K-W2odLc-R6w.png" /></figure><p>This matters because a base_coin is no longer enough to reconstruct a Pump.fun market. The same launch mechanics can now settle against PUMP, SPYx, NVDAx, WBTC, DOGE, USDC and dozens of other assets. Pricing, turnover and creator-fee calculations all need the actual quote mint and its decimals.</p><p>There is one normalization detail worth knowing. Native-SOL creations use the System Program sentinel in pumpfun_token_creation.quote_coin:</p><pre>11111111111111111111111111111111</pre><p>V2 swap rows represent the same leg as wrapped SOL:</p><pre>So11111111111111111111111111111111111111112</pre><p>Non-native quote mints remain exact across creation and swap rows.</p><h3>A fixed two-hour pulse</h3><p>We queried the explicit UTC window 2026-09-09 20:30:00 through 22:30:00. The indexer tip had reached 22:35:36 UTC when the extract ran.</p><p>In those two hours:</p><ul><li>5,547 fresh Pump.fun launches appeared across 90 quote identifiers;</li><li>2,432 launches used a non-native quote asset;</li><li>those non-native launches covered 89 quote mints;</li><li>88 of the 89 are present in the decoded QuoteControl registry;</li><li>the remaining mint is native USDC, which Pump.fun handles as a default quote outside that registry;</li><li>pumpfun_v2_swaps recorded 103,560 successful non-SOL swap transactions in the same window, across 91 non-SOL quote mints.</li></ul><p>The creation side was already broad:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*eNZkouBkNqmnNRYT.png" /></figure><p>The trade table was not merely accepting the schema. It was filling with real activity:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*KBuNkIq2kX2qMFwW.png" /></figure><p>Counts above use unique successful transaction signatures where stated. Swap-row counts can be slightly higher because a transaction may produce more than one indexed instruction row.</p><h3>Fresh rows, not a synthetic demo</h3><p>Here are six launches from the end of the window, joined to their first ten minutes of pumpfun_v2_swaps:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*S-nRxJ97L9xXnisC.png" /></figure><p>The important point is not that these launches traded once. It is that quote_coin survives the whole indexed path: launch identity, swap direction, raw quote amount, signer, fees, transaction ordering and program context are queryable together.</p><h3>What the on-chain registry stores</h3><p>The registry account is:</p><pre>6z6GDdfb2AjR9ZhJmAUQ5cipJCVxQvLJhB2H8mCwTFBP</pre><p>It belongs to the Pump.fun program and decodes as:</p><pre>struct QuoteControl {<br>    discriminator: [u8; 8],<br>    admin: Pubkey,<br>    reserved: [u8; 64],<br>    entries: Vec&lt;QuoteControlMint&gt;,<br>}<br><br>struct QuoteControlMint {<br>    quote_mint: Pubkey,<br>    initial_virtual_quote_reserves: u64,<br>}</pre><p>The account discriminator is exactly:</p><pre>sha256(&quot;account:QuoteControl&quot;)[0:8]<br>= 38f423eec1d5a2c9</pre><p>Each entry is 40 bytes: a 32-byte mint followed by a little-endian u64. The second value is the amount of that quote token used for initial virtual quote reserves, not a live oracle price.</p><p>For example:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*IcJ7PU23Nw1kSi-S.png" /></figure><p>Across assets with usable prices, these reserve amounts cluster around roughly $4.2k at the reference prices used when Pump.fun configured them. That keeps the opening curve in a similar dollar regime while the quote-token units vary by orders of magnitude.</p><p>The registry was initialized on-chain with InitializeQuoteControl, then populated through admin-controlled AddQuoteControlMint calls. The latter has a 48-byte payload:</p><pre>8 bytes   instruction discriminator<br>32 bytes  quote mint<br>8 bytes   initial virtual quote reserves, u64 LE</pre><p>The current program binary names the module quote_control, and live CreateV2/BuyV2 traffic reads the account.</p><h3>Query fresh launches by quote asset</h3><p>This is the compact query used for the launch counts above:</p><pre>WITH<br>    toDateTime(&#39;2026-09-09 20:30:00&#39;, &#39;UTC&#39;) AS start_ts,<br>    toDateTime(&#39;2026-09-09 22:30:00&#39;, &#39;UTC&#39;) AS end_ts<br>SELECT<br>    replaceAll(toString(quote_coin), char(0), &#39;&#39;) AS quote_coin,<br>    uniqExact(replaceAll(toString(mint), char(0), &#39;&#39;)) AS launches,<br>    uniqExact(replaceAll(toString(creator), char(0), &#39;&#39;)) AS creators,<br>    toTimeZone(min(block_time), &#39;UTC&#39;) AS first_launch_utc,<br>    toTimeZone(max(block_time), &#39;UTC&#39;) AS last_launch_utc<br>FROM default.pumpfun_token_creation<br>WHERE block_time &gt;= start_ts<br>  AND block_time &lt; end_ts<br>GROUP BY quote_coin<br>ORDER BY launches DESC;</pre><p>And the swap-side check:</p><pre>WITH<br>    toDateTime(&#39;2026-09-09 20:30:00&#39;, &#39;UTC&#39;) AS start_ts,<br>    toDateTime(&#39;2026-09-09 22:30:00&#39;, &#39;UTC&#39;) AS end_ts<br>SELECT<br>    quote_coin,<br>    countIf(failed = 0) AS successful_swap_rows,<br>    uniqExactIf(signature, failed = 0) AS successful_transactions,<br>    uniqExactIf(base_coin, failed = 0) AS traded_launches,<br>    uniqExactIf(signing_wallet, failed = 0) AS signing_wallets<br>FROM default.pumpfun_v2_swaps<br>WHERE block_date_utc = toDate(&#39;2026-09-09&#39;)<br>  AND block_time &gt;= start_ts<br>  AND block_time &lt; end_ts<br>GROUP BY quote_coin<br>ORDER BY successful_transactions DESC;</pre><p>Do not sum quote_coin_amount across different mints. It is a raw integer in each quote asset’s own decimals. Normalize per mint first, then apply a contemporaneous price if a cross-asset dollar view is required.</p><h3>What this enables</h3><p>The new field makes several analyses possible without reconstructing quote identity from token transfers:</p><ul><li>compare launch activity across quote regimes;</li><li>build quote-specific prices and turnover;</li><li>separate creator fees by settlement asset;</li><li>measure whether sniper behavior changes between SOL, equities, commodities and crypto quotes;</li><li>follow a launch into migration without losing the original quote denomination.</li></ul><p>The registry tells us what Pump.fun configured. The indexer shows what people actually launched and traded.</p><p>OnChainDivers subscribers can query both sides directly: pumpfun_token_creation.quote_coin for the launch and pumpfun_v2_swaps.quote_coin for the executed market.</p><h3>Sources and reproducibility</h3><ul><li><a href="https://solscan.io/account/6z6GDdfb2AjR9ZhJmAUQ5cipJCVxQvLJhB2H8mCwTFBP#accountData">Pump.fun QuoteControl account(opens in new tab)</a></li><li><a href="https://solscan.io/tx/3ULbKDUPuEVADdBEnmmY8owHD1SzjrsDnULe2Qk6nmfLmeDBq8gYrCw9n68TTpRRdTkGBpoFGAoYTPQzsCfJgqH3">QuoteControl initialization transaction(opens in new tab)</a></li><li><a href="https://solscan.io/tx/oYQAamkxAdhEaAAS5HWSac4HupMsNEEJmo8hXxPsY19YzAqT1UpQw4LGhttdDYxugEDVNStXjt17E6nfsccsjb9">Batch containing the GOOGLx QuoteControl addition(opens in new tab)</a></li><li><a href="https://github.com/pump-fun/pump-public-docs/blob/main/idl/pump.json">Pump.fun public program IDL(opens in new tab)</a></li></ul><p>The complete executed queries are included alongside this article as fresh_quote_summary.sql, fresh_swap_quote_summary.sql and fresh_exotic_launches.sql.</p><p><em>OnChainDivers provides indexed Solana launch, swap and migration data for researchers building lifecycle-aware market models. Subscribe for schema updates, reproducible SQL and delayed research examples.</em></p><p><strong>Connect with us:</strong></p><ul><li>Website: <a href="https://onchaindivers.com/">onchaindivers.com</a></li><li>Twitter: <a href="https://x.com/onchain_divers">@onchain_divers</a></li><li>Discord: <a href="https://discord.com/invite/PRWyXUvhU3">Join our community</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=654ff8ccb68e" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>