<?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 ironchef on Medium]]></title>
        <description><![CDATA[Stories by ironchef on Medium]]></description>
        <link>https://medium.com/@ironchef253?source=rss-ff133632ee43------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>Stories by ironchef on Medium</title>
            <link>https://medium.com/@ironchef253?source=rss-ff133632ee43------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 13:52:29 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ironchef253/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[The Martian Standard]]></title>
            <link>https://ironchef253.medium.com/the-martian-standard-d46f4b5d810a?source=rss-ff133632ee43------2</link>
            <guid isPermaLink="false">https://medium.com/p/d46f4b5d810a</guid>
            <dc:creator><![CDATA[ironchef]]></dc:creator>
            <pubDate>Fri, 19 Sep 2025 02:01:42 GMT</pubDate>
            <atom:updated>2025-09-19T02:01:42.259Z</atom:updated>
            <content:encoded><![CDATA[<p>Was hacking on some ideas for why the best move for Donald Trump and Elon Musk would be to pursue a Bretton Woods 2.0 which attempts to refactor how economics work globally.</p><p>Thesis:</p><p>A new branch of economics — Frontier Collateral Economics, or the Martian Standard — can prevent WWIII and enable Mars by collateralizing forward-valued assets through AI, robotics, and tokenization, creating the next Bretton Woods.</p><p>Summary:</p><p>The greatest obstacle to Mars is not rockets or engineering. SpaceX has already made Starship real, Tesla is building the robotics, and Starlink provides the communications backbone. The true bottleneck is sovereign solvency. Global debt loads are unsustainable, and history shows that unresolved debt crises spill into collapse or world wars. Without a new financial operating system, there will be no Mars.</p><p>In 1944, at Bretton Woods, John Maynard Keynes and his peers recognized that the old gold standard had collapsed, and that a new framework was needed to rebuild solvency after depression and war. That system — currencies pegged to the U.S. dollar, the dollar pegged to gold — stabilized the world for a generation. When Nixon severed gold convertibility in 1971, the fiat-dollar era began. Fifty years later, that system is exhausted. Debt grows, demographics strain, and geopolitics edge toward conflict. We are once again at a breaking point.</p><p>The solution is a Bretton Woods 2.0 — built not on war’s aftermath but on foresight. I call this the Martian Standard, or more formally, Frontier Collateral Economics. The core idea is simple but powerful: we can collateralize forward-valued frontier resources using AI, robotics, and tokenization.</p><p>•	AI price discovery: AI-to-AI markets can generate reproducible indices for unextracted resources — gold, lithium, copper, even Martian minerals — with transparent governance and contestability.</p><p>•	Robotics and logistics: Tesla robotics, Starship, and Starlink make future extraction credible.</p><p>•	Tokenization and sovereign adoption: These indices can be packaged into blockchain-based instruments and adopted as capped, regulated collateral for sovereign debt issuance.</p><p>Gold-backed Treasuries are effectively “chunks of the United States in gold form.” But why stop there? All commodities, crypto, and even Martian assets can form part of a Collateralized Everything Standard. This expands the collateral base for governments, provides credible new borrowing capacity, and anchors the economics of frontier expansion.</p><p>For Elon Musk, this framework is more than relevant — it is essential. Without a new economics, Mars is just a dream. With it, Mars becomes bankable. And more: it unifies Musk’s empire — SpaceX (logistics), Tesla (robotics and energy), Neuralink/AI (valuation and governance), Starlink (communications), and X (narrative) — under one doctrine. Musk ceases to be just a builder of rockets and becomes the architect of the next global financial order.</p><p>The alternative is grim. Without a new Bretton Woods, the debt crisis will fester until collapse or war forces a reset, just as depression and extremism led to WWII. Better to act with foresight this time. Better to build Bretton Woods 2.0 before WWIII, not after.</p><p>The Martian Standard is that path: a new branch of economics, designed to solve the hardest problem on Earth so that humanity can expand beyond it.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d46f4b5d810a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Power of AI Consensus]]></title>
            <link>https://ironchef253.medium.com/the-power-of-ai-consensus-39e9afcb8204?source=rss-ff133632ee43------2</link>
            <guid isPermaLink="false">https://medium.com/p/39e9afcb8204</guid>
            <dc:creator><![CDATA[ironchef]]></dc:creator>
            <pubDate>Mon, 03 Mar 2025 02:46:44 GMT</pubDate>
            <atom:updated>2025-03-03T02:46:44.242Z</atom:updated>
            <content:encoded><![CDATA[<h3>Thoughts On AI Consensus</h3><p>I have a lot of ideas which I have documented here. Many of them turn out to be exactly right, well in advance. AI Consensus is one of those ideas. It feels somewhat painful to talk about. It just has “that feeling” about it. Every time I come back in particular to this article about G<a href="https://medium.com/@rexstjohn/generative-blockchains-e9f0a882ddbe">enerative Blockchains</a> and reevaluate my thinking, it still feels correct.</p><p>I am coming back again to the topic of AI Consensus after doing further work at Verafy on this topic. I feel there is something huge here that represents a total industry shift that needs to be considered.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*osoiPox3vDxMeJ-oaynklw.png" /></figure><p>I call AI Consensus (g) chains. They are a form of blockchain but with extremely different parameters. They make use of blockchains for hard commitments but add an additional fuzzier layer on top: Consensus between AIs.</p><p>I call them (g)chains because while modern blockchains are ledgers of account, most of what goes on is purely mathematical in nature. This is sort of like measuring a student by only their performance on a math test. Bitcoin uses purely cryptographic methods to generate new blocks.</p><p>But what if you had access to intelligence which is MULTIDIMENSIONAL. In other words, what if your “problem set” extended beyond pure cryptography. What if you could design problems which were amalgams of geometry, physics, chemisty, history, art, music and post them to a network of connected nodes. And the “smartest” nodes would solve them most accurately and be rewarded by getting the next block?</p><p>AI makes this a possibility. We have been constrained in the blockchain space by our tools: Computers which are great at math. But now we have new tools: Computers which can do art, music, video generation, audio, speech, parrot wikipedia essays, evaluate text.</p><p>Our new computers can do a lot more than math, which opens up entirely new dimensions to explore when it comes to “Proof of Work” networks.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5DSNVQkHz3x5uU5syUbLNw.png" /></figure><p>I have written previously about the topic of African Wild Dog packs which “Vote” via sneezing. There are one or more alpha dogs sure, but these dogs are not “dictators” of the entire pack. They do not get to call the shots in their entirety.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/1*b_lgwNwViqT4NemcXE-vMQ.jpeg" /></figure><p>So why do tribal animals like dogs or humans or chimps have variations in personality, size, intelligence and other dimensions? I think there must be something deeply mathematical about it rooted in physics and energy.</p><p>My theory is that heterogenous modular systems like this are the best way to acquire energy overall for the group. The alpha gets to be the “tie breaker” because it is deemed “most authoritative.” If a hard decision must be made such as “Sacrifice the old or young to leave the island surrounded by alligators to save the group,” the alpha is most able to make the call — by making the trip themselves and potentially harassing recalcitrant members to get on board with their plan. All of this is highly energy intensive. If the old, weak or sick got to make the call — the entire pack would starve. It’s about energy.</p><p>So why not new forms of blockchain where the most intelligent nodes get stronger voting power?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mB1UtkMhkEfLZCmnEq5n6Q.png" /></figure><p>I am imainging a world where a decentralized network of heterogeneous AI might be posed with a set of synthetic challenge questions which are designed to stretch AI brains in multiple directions to see “who is the smartest.”</p><p>You can imagine a synthetic question blending art, music, history, geometry, language, math which is so challenging that only the most advanced AI might solve it. Such a challenge might be sent out to the network and then the responses measured to determine “who is the smartest brain.”</p><p>This is quite different but related to Bitcoin in that a challenging cryptographic problem is dispatched to the network and uniform nodes “mine” for the answer and are rewarded if they get it “right.”</p><p>But that is purely math. With AI, we can design problems that are very different and stretch electronic brains in ways that are beyond the Bitcoin network.</p><p>I am imagining systems where the strongest AI get to first prove their intelligence and then get a share of voting (Similar to how much voting PoS nodes get to do based on their stake) based on the capability of the electronic brain.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GtHw6IdjN0UbkDe1V0BbFQ.png" /></figure><p>(g) chains have fascinating ramifications.</p><p>Once you leave the narrow confines of pure math, you can do things like create a movie chain.</p><p>A “Request for Frame” is originated which supposes to a DePin network that it wants “the next frame” in a movie sequence. The DePin nodes compete to generate the best or most interesting next frame.</p><p>AI then evaluates the frame against the past 10 frames and makes some qualitative (rather than a hard math) decision between 20–30 proposed frames or snippets and chooses “the best one,” appends it to the movie then restarts the cycle.</p><p>The same can work with sound, music, audio, text. (g) chains are radically different from traditional blockchains as AI PoW opens a range of new possibilities.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TSIGynOUcZyUYJzoKBW1HQ.png" /></figure><p>Verafy.ai is a form of (g) chain because we are using a network of AI to evaluate the “truth” of a document by reading provided sources and then making a judgement call about the combination of document and sources.</p><p>This is purely a qualitative decision — There is no such thing as a “best” essay in the world. Or “the best movie,” or the “best image in a sequence.” The closest we can get is “sufficiency.”</p><p>(g) chains are fundamentally “fuzzy.” Unlike blockchains, there is no “Right” answer. There is merely a “sufficient” answer (one that crosses a threshold for acceptance via a collection of agents or AI that evaluate the input).</p><p>(g) chains are best used in situations where mistakes and error have low cost. Unlock blockchains, where error results in user funds being compromised, (g) chains have more leeway to make mistakes. Is this “the correct” next frame of the movie? It is correct enough, there isn’t a huge problem if its not exactly right.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*617PDvRrx-rMVJJml81ruA.png" /></figure><p>(g) chains rely on Platonic Theory of Forms. You can make 10,000 drawings which will “suffice” to being identified as a “drawing of a horse.” You can only make 1 transaction which properly solves for “I sent the $100 to account Y at 2pm.”</p><p>We live in a fuzzy world. Most truths we rely on depend on framing and perspective. The overly skinny person looks out of shape to the bodybuilder. The bodybuilder looks overbuilt to the long distance runner. YOu get a different answer to the question “is this person an ideal athlete” based on the context and perspective and reference frame through which the question is answered.</p><p>(g) chains take the training wheels off and require a huge stretch of imagination to understand. The era of fuzzy blockchains feels close — Because when you begin overlaying AI consensus on top of the bedrock of blockchains, you are mixing ying and yang.</p><p>These are two different worlds altogether: The qualitative and the quantitative.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Y26pd6QCMuDXNRfeeH5n5w.png" /></figure><p>So what is the truth? (g) chains can get us to a fuzzy approximation of the truth, but that is the closest humanity will ever get.</p><p>AI Consensus requires some philosophical backflips which are a little beyond the current stage of the industry I think.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=39e9afcb8204" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Is Ethereum The SAP Of Blockchains?]]></title>
            <link>https://ironchef253.medium.com/is-ethereum-the-sap-of-blockchains-2bfccd734580?source=rss-ff133632ee43------2</link>
            <guid isPermaLink="false">https://medium.com/p/2bfccd734580</guid>
            <dc:creator><![CDATA[ironchef]]></dc:creator>
            <pubDate>Wed, 18 Dec 2024 18:27:44 GMT</pubDate>
            <atom:updated>2024-12-18T18:27:44.555Z</atom:updated>
            <content:encoded><![CDATA[<p>Reading the news about Deutsche Bank pursuing an Ethereum L2, along with Sony earlier in the year makes me wonder about the trajectory of Ethereum. Base is obviously the most prominant example of a centralized corporation launching an L2. But now we see others doing the same.</p><p>What is the motivation? Given a choice between going full Bitcoin maxi (holding a treasury in Bitcoin), going fully decentralized (using a Solana or Bitcoin or Ethereum L1 or other) as their payment solution or forking/building in house — I wonder how many banks and enterprises will choose to fork and launch an L2.</p><p>This dynamic sort of reminds me of SalesForce and SAP. Every Fortune 500 has a cruddy version of SalesForce or SAL requiring constant enterprise maintenance, training and upgrading so they can solve some highly boring internal problem relating to CRM or similar.</p><p>Is that what Eth is turning into? Are we going to see the long list of outdated banks choosing to attempt to form and manage their own internal L2 as their payment play of choice?</p><p>It comes down to the economics.</p><p>With an Ethereum L2, the primary objective would be to capture the sequencer revenues as well as (hopefully) some associated upside with a token. Base has chosen not to have a token, but the approach of monetizing the sequencer and avoiding the annoying legal problems associated with floating a token may be more widely appealing than we might imagine.</p><p>If you are a large bank (there are hundreds), you operate an economy to yourself. Your business alone in combination with the scale of your customers, partners and peers represents a very large volume of annual transactions.</p><p>If you ultimately hope to further monetize these transactions, and your own in-house economy, what better way to go about it than to introduce a sequencer you control and to get a share of all the sequencer fees for “your” economy. You built this economy, after all, why would you give away the fees to anyone.</p><p>What motivation would such a bank or enterprise have to share their fees with Bitcoin, Solana, Ethereum L1 or other? Most banks are not in the business of making the world a better place by sharing fees. So keeping all the fees in house via a wholly owned sequencer is a move that the internal MBAs might embrace, if they are being forced to enter Crypto (kicking and screaming).</p><p>On might argue that established enterprises are ill suited to build and operate their own L2s. This is similarly true of SAP and SalesForce instances…yet here we are. SalesForce and SAP are still embedded around the world and thriving despite what techno idealists might prefer.</p><p>So if the enterprise banking world is dragged into crypto against their will by changing regulatory and market realities, it remains to be seen how many will attempt the sequencer monetization approach over adopting a best in class alternative such as Solana where their fee collection is lower.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2bfccd734580" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[CHIPS act failure? Blame NVIDIA.]]></title>
            <link>https://ironchef253.medium.com/chips-act-failure-blame-nvidia-4cfeabdd56e1?source=rss-ff133632ee43------2</link>
            <guid isPermaLink="false">https://medium.com/p/4cfeabdd56e1</guid>
            <dc:creator><![CDATA[ironchef]]></dc:creator>
            <pubDate>Tue, 03 Sep 2024 21:22:41 GMT</pubDate>
            <atom:updated>2024-09-03T21:22:41.843Z</atom:updated>
            <content:encoded><![CDATA[<p>It was surprising today to see the news of a brand new <a href="https://www.bloomberg.com/news/articles/2024-09-03/nvidia-gets-doj-subpoena-in-escalating-antitrust-investigation">Antitrust probe against NVIDIA</a>. What exactly might this be about? One answer might be that it’s primary competitor, Intel, has announced a pending potential split from it’s foundry business and <a href="https://finance.yahoo.com/video/intel-considers-foundry-business-split-133444727.html?guccounter=1&amp;guce_referrer=aHR0cHM6Ly93d3cuZ29vZ2xlLmNvbS8&amp;guce_referrer_sig=AQAAAF8UW0szw-yvpcz3rdB1sD5T4sCsrv0Plbgg6j1yKcuf-vP_Uh3EdFBbwscvlygUAldI93uufbPvrU2Da7uwPN8iDv1-thCoQX8KK34N63MuUKeZvcHMgcWwU2xXM8FrjkPYvANq3KLs79wq3Po6hGg69H_OC4khzXBYBZ02CVGN">retained Goldman Sachs to support</a>.</p><p>There is a little bit of a problem here given that we are in the home stretch of a high contentious election. A primary objective of the Biden administration has been to preserve domestic semiconductor manufacturing via <a href="https://en.wikipedia.org/wiki/CHIPS_and_Science_Act">the CHIPS act.</a></p><p>If Intel were to break apart within 60 days of the November election, this would cast a shadow over the effectiveness of the CHIPS act — adding an unpleasant color to the current administration resulting in a rather inconvenient news cycle heading into the voting season.</p><p>One potential strategy to get ahead of the news? Accuse NVIDIA of antitrust behavior. Reading through the currently available talking points — I am not able to find anything substantive other than “NVIDIA’s chips are really good.”</p><p>Being highly competent, hiring and leading an elite team, making long term plans (founding and sticking with the CUDA platform for decades) and winning big is not “antitrust,” it is competent leadership and execution.</p><p>I am a little worried about this announcement — I suspect, based on having worked at several of these companies, that we are seeing a pre-emptive effort to cover for the breakup of a major American semiconductor company which was supposed to be protected by the CHIPS act.</p><p>That looks bad, maybe blaming NVIDIA (for being competent) for Intel’s failure (and the failure of the CHIPS act) might provide a little bit of air cover leading into the election. Unless I see more concerning details from this probe, it seems rather suspicious.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4cfeabdd56e1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Learn more about the Solana Labs to Agave fork]]></title>
            <link>https://medium.com/anza-xyz/learn-more-about-the-solana-labs-to-agave-fork-f8507453e639?source=rss-ff133632ee43------2</link>
            <guid isPermaLink="false">https://medium.com/p/f8507453e639</guid>
            <dc:creator><![CDATA[ironchef]]></dc:creator>
            <pubDate>Tue, 05 Mar 2024 18:05:01 GMT</pubDate>
            <atom:updated>2024-03-05T18:05:01.393Z</atom:updated>
            <content:encoded><![CDATA[<p><em>This is a quick post written for the Solana Discord server to provide further details around the upcoming Solana Labs fork.</em></p><p>This weekend (Saturday, March 2nd)we forked the Solana Labs validator software into a repo managed by the Anza team <a href="https://github.com/anza-xyz/agave.">located here</a>. Please see our <a href="https://x.com/anza_xyz/status/1763667506785587248?s=20)">X post here </a>and <a href="https://medium.com/anza-xyz/anza-github-migration-faq-9ab94e03f8dc">FAQ here</a>.</p><p>If you are not familiar with Anza, we are a new dev shop within the Solana ecosystem consisting of former executives and core engineers from Solana Labs. One of our objectives is to build a forked Solana Labs validator client which will be called “Agave,” as well as contribute to other major protocols within the Solana ecosystem. Our <a href="https://medium.com/anza-xyz/meet-anza-a-new-solana-focused-dev-shop-7d92391d290b">announcement provides further background</a>.</p><p>Our objective in forking the client will be to separate future contributions by the Anza team from the previous work at Solana Labs. This is part of our process to support the Firedancer client being developed by Jump Crypto, build dev tools and provide further engineering support to the broader Solana developer ecosystem.</p><p>Anza will be joining the Solana Validator Community Call (March 14th at 11am PST) and the Solana Foundation Core Community Call March 13th to answer questions on this topic live. Please feel free to follow our new DevRel lead, Rex St. John, who <a href="https://twitter.com/rexstjohn">will be active across various forums to answer any questions you may have</a>. Finally, follow Anza on X to stay updated for future announcements: <a href="https://twitter.com/anza_xyz">https://twitter.com/anza_xyz</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f8507453e639" width="1" height="1" alt=""><hr><p><a href="https://medium.com/anza-xyz/learn-more-about-the-solana-labs-to-agave-fork-f8507453e639">Learn more about the Solana Labs to Agave fork</a> was originally published in <a href="https://medium.com/anza-xyz">Anza.xyz</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[An Open Source Embodied Intelligence Companion]]></title>
            <link>https://ironchef253.medium.com/an-open-source-embodied-intelligence-companion-46df8c4a205b?source=rss-ff133632ee43------2</link>
            <guid isPermaLink="false">https://medium.com/p/46df8c4a205b</guid>
            <dc:creator><![CDATA[ironchef]]></dc:creator>
            <pubDate>Sat, 02 Mar 2024 19:04:43 GMT</pubDate>
            <atom:updated>2024-03-04T16:09:46.923Z</atom:updated>
            <content:encoded><![CDATA[<p><a href="https://m.youtube.com/watch?v=xNUyh_lxRXA&amp;si=VOwOeJ_Jlf3jRAGd">https://m.youtube.com/watch?v=xNUyh_lxRXAsi=VOwOeJ_Jlf3jRAGd</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/590/1*47XUs2nyvW2VVPDqj8YFdQ@2x.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/590/1*UFmZBsqMDfjA9HTGeYOOdg@2x.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/590/1*B9J-a63wyOyxV6N2dylUNw@2x.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/590/1*YBwBGObCaZ_7jwfUr5mczw@2x.jpeg" /></figure><p>I have been working on a side project with some friends to develop an open source embodied intelligence companion using on-device LLMs.</p><p>My target use case is a speech therapy companion, especially to help with single parent households where the burden, cost and time of hiring a speech therapist can run from $600-$1,200 / money.</p><p>An open source robot speech practice robot with an LLM could be revolutionary and save tremendous time and costs.</p><p>My other use cases are reminders for kids, ADHD executive function support and autism skills coaching.</p><p>Our first demo has been completed, in the future we want to increase community participation, work towards one or more LLMs which are trained in speech coaching as well as autism skills coaching.</p><p>The initial robot uses whisper, home assistant.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=46df8c4a205b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Internet In A Box]]></title>
            <link>https://ironchef253.medium.com/the-internet-in-a-box-06491b2b306a?source=rss-ff133632ee43------2</link>
            <guid isPermaLink="false">https://medium.com/p/06491b2b306a</guid>
            <dc:creator><![CDATA[ironchef]]></dc:creator>
            <pubDate>Thu, 22 Feb 2024 05:29:03 GMT</pubDate>
            <atom:updated>2024-02-22T05:36:46.677Z</atom:updated>
            <content:encoded><![CDATA[<p>I am beginning to think it will be possible to recreate the entire internet on a device with no internet connection.</p><p>Wikipedia is consistently one of the top sources for Google searches. It is only 22 gig when downloaded in compressed format. Large Language Models and stable diffusion checkpoints are in that same range (give or take).</p><p>It isn’t a huge stretch to believe that a 60 gig combined file (checkpoint + LLM + knowledge store) would be possible that can approximate nearly everything on the internet. And with significant further compression and optimization, perhaps that file could be smaller.</p><p>We might literally be looking at a situation where you can effectively generate or extrapolate (from first principals) anything on the internet from a 30–60 gig set of files.</p><p>This includes the ability to generate any image, any book (via decompression from summary materials), movies.</p><p>You could create a completely offline version of the internet that is “directionally correct.” It would be lossy. Jurassic Park recreated from a Wikipedia summary about the movie and book would only sort of resemble Steven Spielberg’s masterpiece.</p><p>Likewise, an 800 page version of Jurassic Park uncompressed via AI from a book summary wouldn’t exactly be right, but it could be directionally right.</p><p>It won’t be 100% correct, perhaps maybe 15–20% correct, but it is conceivable. I think it could be made to be fairly close to acceptable for human consumption, at the very least.</p><p>I give it 6 months before it happens.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=06491b2b306a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Decentralized Networks Must Embrace Job Queues]]></title>
            <link>https://ironchef253.medium.com/decentralized-networks-must-embrace-job-queues-ded807ef4a68?source=rss-ff133632ee43------2</link>
            <guid isPermaLink="false">https://medium.com/p/ded807ef4a68</guid>
            <dc:creator><![CDATA[ironchef]]></dc:creator>
            <pubDate>Thu, 15 Feb 2024 17:35:21 GMT</pubDate>
            <atom:updated>2024-02-15T17:58:09.543Z</atom:updated>
            <content:encoded><![CDATA[<h2>Decentralized Networks Should Embrace Job Queues</h2><p>My message is pretty simple: I should be able to package up an AI prompt, plus some meta data, or a lambda and drop it into the network, plus some tokens and expect that 100% of the time, someone listening will volunteer to run the AI workload, lambda function or container I provide and send me back the result.</p><p>Compute is a multi trillion+ market worldwide. Some of the top Discord applications involve simply chatting with bots to get AI outputs.</p><p>Decentralized networks should embrace being able to accept packaged workloads by default to enable the next generation of open marketplaces.</p><p>I have said many times that the majority of global compute tends to be underutilized or sit idle.</p><p>While Sam Altman and Jensen duke it out over who will produce new silicon, I am here to say: Let’s just use the silicon we have guys. If we just do that, we don’t need to produce that much new silicon.</p><p>One of my favorite programming APIs was Apple’s “blocks” which enabled a developer to define an execution snippet and then drop it into the operating system without having to wrangle CUDA or threading woes.</p><p>We need to be able to drop execution blocks into the open network to create AI and other responses natively.</p><p>How about we address the real market need: Compute.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ded807ef4a68" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The next instruction set is no instruction set.]]></title>
            <link>https://ironchef253.medium.com/the-next-instruction-set-is-no-instruction-set-86be4e8ad94a?source=rss-ff133632ee43------2</link>
            <guid isPermaLink="false">https://medium.com/p/86be4e8ad94a</guid>
            <dc:creator><![CDATA[ironchef]]></dc:creator>
            <pubDate>Sun, 11 Feb 2024 21:25:32 GMT</pubDate>
            <atom:updated>2024-02-11T21:35:20.258Z</atom:updated>
            <content:encoded><![CDATA[<h2>The next instruction set is no instruction set. Assembly language and machine code were always a hack to make it convenient for humans to talk to intelligent sand.</h2><p>The future will be a universal kernel on top of an alien instruction set formed by AI on a Per use case basis.</p><p>In the near term this will be ignored. In the long term they will be saying “how did he know that so far in advance?”</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=86be4e8ad94a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[2024 Semiconductor Industry Predictions]]></title>
            <link>https://ironchef253.medium.com/2024-semiconductor-industry-predictions-e95d93e84a9c?source=rss-ff133632ee43------2</link>
            <guid isPermaLink="false">https://medium.com/p/e95d93e84a9c</guid>
            <dc:creator><![CDATA[ironchef]]></dc:creator>
            <pubDate>Sun, 11 Feb 2024 18:11:30 GMT</pubDate>
            <atom:updated>2024-02-12T07:51:56.070Z</atom:updated>
            <content:encoded><![CDATA[<p>It is February 11th, 2024. Over the last year, we have put together a series of detailed predictions which successfully forecasted developments such as the Rabbit r1, the first consumer language model computer, and other trends. In this document, we are going to outline a series of time-bound predictions for how AI and Semiconductors will combine and produce a number of surprises.</p><h3>2024 Semiconductor Industry Predictions</h3><h3>#1: Liquid Computing (Death of Semiconductor IP Worldwide)</h3><p>Some have speculated that Generative AI <a href="https://www.newyorker.com/magazine/2024/01/22/who-owns-this-sentence-a-history-of-copyrights-and-wrongs-david-bellos-alexandre-montagu-book-review">may lead to the death of IP</a>. Decentralized Thoughts is the first, only ever in history publication to predict that this trend will spread from art, film, software and music directly into hardware. Thats right. Not only will the instruction set be open (RISC-V), all the ASICS, drivers, hardware components, algorithms and chiplets will be open as well.</p><p>The result will be that semiconductors, the last bastion of IP law defense, will crumble in the face of an overwhelming onslaught of computer generated components. This “Liquefaction” of IP and semiconductors means that it will be possible, normal and common to generate end-to-end machines, including the operating system and kernel in a single pass.</p><blockquote><em>PREDICTION: By the End of 2024, we will see end-to-end systems generated on the fly which include every component, the operating system, UI, ASICS and increasingly aspects of the supply chain for assembly itself.</em></blockquote><p>This is a specific, named and time-bound prediction being made directly here. You are not seeing such predictions anywhere else on the internet from any other analyst to my knowledge at this time.</p><h3>#2: A Futures Market For GPUs and GPU Compute</h3><p>I have documented my “Machine Economy” thesis multiple times:</p><ul><li>A futures market for GPU compute (<a href="https://www.linkedin.com/pulse/ai-require-futures-market-computation-rex-st-john-apxmc/?trackingId=gSEKETUpRq2Gum%2BrBy38bw%3D%3D">link</a>)</li><li>Datacenter Central banking (<a href="https://www.linkedin.com/pulse/datacenter-central-banking-rex-st-john-yapyc/?trackingId=VG8RRRJzThuFu6%2B9X5mBJA%3D%3D">link</a>)</li><li>The Machine Economy (<a href="https://medium.com/@rexstjohn/the-machine-economy-96236c55fbe6">link</a>)</li></ul><blockquote><em>Prediction: We will see a futures market emerge, algorithmically and AI traded for tokenized forward compute contracts and GPUs this year</em></blockquote><h3>#3: Sam Altman Will Build The Matrix</h3><p>In order to re-imagine the global semiconductor market supply chains, Sam Altman will need to completely “throw away” the rule book and start over from scratch. This involves two core elements:</p><ul><li>Sam Altman must build the greatest ever end-to-end semiconductor supply chain simulator (<a href="https://medium.com/@rexstjohn/sam-altman-must-build-the-matrix-to-solve-the-ai-semiconductor-shortage-8f74ed8f3bc7">The Matrix</a>) in history in order to outperform TSMC and others. This simulator will include every worker in every factory, the ships used to send supplies, the robots and the design of robots used to create the components</li><li>Sam Altman <a href="https://medium.com/@rexstjohn/how-i-would-approach-semiconductors-981a68f22f1a">must liquify computers end-to-end so they can be designed on the fly by AI</a>, this will involve going as far as throwing away the last 100 years of human-designed computers and providing AI with nothing more than math and physics</li></ul><blockquote><em>PREDICTION: In the next 3–6 months, Sam Altman will announce the creation of The Matrix, it may be called something new</em></blockquote><h3>Closing Thoughts</h3><p>By the end of the year, 2024, computing as we know it will be over. Computers will no longer contain any human-generated elements and will be 100% generated by machines. The internal functioning of computers will be a complete mystery, and will be created purely by AI as part of an integrated end-to-end supply chain (The Matrix). I expect to see, by the end of 2024, the first machines of this type begin to emerge.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e95d93e84a9c" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>