<?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 Alex Wormuth on Medium]]></title>
        <description><![CDATA[Stories by Alex Wormuth on Medium]]></description>
        <link>https://medium.com/@nftechie?source=rss-7880af92a93a------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*Jzld4D6EFd1FwGg7pChjGw.jpeg</url>
            <title>Stories by Alex Wormuth on Medium</title>
            <link>https://medium.com/@nftechie?source=rss-7880af92a93a------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 08:14:55 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@nftechie/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[Minimum Viable Slop]]></title>
            <link>https://medium.com/@nftechie/minimum-viable-slop-28c6594d9f91?source=rss-7880af92a93a------2</link>
            <guid isPermaLink="false">https://medium.com/p/28c6594d9f91</guid>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[vibe-coding]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Alex Wormuth]]></dc:creator>
            <pubDate>Fri, 06 Mar 2026 16:01:55 GMT</pubDate>
            <atom:updated>2026-03-06T16:05:22.260Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7qQbTgIe6VLNvzSqoW4FlA.png" /></figure><p>If you’ve spent any time on X or LinkedIn recently, you’ve seen the posts. Someone built a SaaS in a weekend. A solo founder shipped an MVP in three hours. A non-technical founder “vibe coded” their way to a working product. The tone is breathless, evangelical. We’re all builders now.</p><p>I spent my paternity leave finding out what that actually means.</p><p>Over a few weeks I built a LoRa analysis tool, open-sourced Claude skills for <a href="https://github.com/nftechie/apple-health-skill">Apple</a> and <a href="https://github.com/nftechie/strava-skill">Strava</a>, attempted a Solana price prediction model, tried my hand at a pump.fun trading bot, and designed logos for my dad’s business. I used AI for almost all of it. I moved faster than I ever have. And somewhere in the middle of it, I started to feel uneasy.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*m4htXiXkN7gXzfqKqXVCQg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EMmks3H4JYkbjyYvSK9x5g.png" /><figcaption>One of the projects that actually mattered — scoping LoRa coverage for cattle tracking on my dad’s ranch. Real problem, real terrain, real consequences.</figcaption></figure><h3>Speed is real</h3><p>Let’s give the hype its due. The productivity gains are genuine. What used to take me weeks now takes hours. Boilerplate that once consumed entire afternoons gets scaffolded in minutes. The barrier to starting something has never been lower.</p><p>This is not nothing. For anyone who’s ever had an idea and lacked the technical depth to execute it, or who’s been bottlenecked by the sheer grunt work of building — this is a meaningful shift. The hype is pointing at something real.</p><p>But speed is also the least interesting thing that changed.</p><h3>The slop problem</h3><p>When the cost of building drops, so does the bar for starting. That sounds fine until you think about what it means at scale: thousands of people building faster than ever, mostly building the same things, mostly building things nobody asked for.</p><p>Slop. Fast, frictionless, indistinguishable.</p><p>The subtler problem is distraction. When I could spin up a new project in an afternoon, I found myself spinning up new projects in afternoons — chasing whatever felt interesting that day. Three days on a shiny idea is no longer a minor detour. It’s three days of opportunity cost at a moment when someone else working on the <em>right</em> thing can outpace you faster than ever. Speed cuts both ways.</p><p>The lower the friction to build, the more important it becomes to be deliberate about <em>what</em> you build.</p><h3>AI can’t solve the hard part</h3><p>Here’s where the hype breaks down most clearly.</p><p>I tried building an ML model to predict SOL prices. A pump.fun trading bot. A protein function prediction tool. These are hard problems — not hard in the sense of tedious, but hard in the sense of genuinely unsolved. They require creativity, novel problem framing, the kind of lateral thinking that comes from deep domain knowledge.</p><p>AI was useful. It accelerated my exploration, helped me set up evaluation frameworks, kept me from getting stuck on syntax. But it couldn’t one-shot any of it. The models aren’t creative enough to solve genuinely hard problems on their own. They’re very good at executing within a well-defined problem space. They’re not good at discovering what the problem space should be.</p><p>This matters because a lot of the hype implicitly promises otherwise. The message is: describe what you want, and AI will build it. That’s true for a narrow band of well-understood problems. For anything genuinely novel or difficult, you still need judgment, taste, and domain expertise. AI accelerates execution. It doesn’t replace the question of whether you’re executing on the right thing.</p><h3>Software is not product</h3><p>The deeper issue is one the hype barely acknowledges: software is not the same thing as product.</p><p>Products solve real human problems. They require understanding people — what they actually struggle with, what they’ll pay to fix, what they’ll tell their friends about. They require craft: the unglamorous work of iterating on something until it’s genuinely good. They require distribution: getting the thing in front of the people who need it.</p><p>None of that got faster. AI can help you write the code. It can’t give you the insight that the problem is worth solving in the first place. It can’t build the trust that makes someone hand over their credit card. It can’t replace the judgment that distinguishes a product people love from one they try once and forget.</p><p>We’re in a bubble — not necessarily in the financial sense, but in the attention sense. The gap between what’s easy to build and what’s worth building has never been wider. Everyone can ship. Almost nobody is asking whether they should.</p><h3>Two futures, one uncomfortable question</h3><p>Everything above assumes a constraint that might not hold: that AI <em>can’t</em> solve the hard problems. That’s true today. It may not be true for long. And which way it goes changes everything.</p><p><strong>If AI never gets there</strong> — if the models plateau and genuine creativity, novel problem framing, and deep domain reasoning remain stubbornly human — then the bubble pops. The slop piles up, trust erodes, and the people who survive are the ones who used the speed to solve real problems faster, not to sidestep the question of whether the problem was worth solving. The heuristic matters more, not less.</p><p><strong>If AI does get there</strong> — if the models eventually become capable of genuine creativity and unsupervised problem-solving — then the world looks stranger than most people building right now seem to imagine. The scarce resource isn’t technical execution anymore. It’s taste. It’s knowing which problems are worth solving. It’s understanding people well enough to point extraordinarily powerful tools in the right direction. Distribution, trust, and domain insight become the only real moats. That’s not the “everyone’s a founder” future the hype promises. Most people don’t know what’s worth building. That skill — quiet, unfashionable, hard to fake — becomes extraordinarily valuable.</p><p>The uncomfortable truth is that we don’t know which future we’re in. The honest answer is probably somewhere between the two, arriving unevenly across domains, at a pace nobody can reliably predict. But here’s what’s striking: the right response is identical in both cases. In a world where AI can’t solve hard problems, you need judgment to pick the right ones. In a world where it can, you need judgment to pick the right ones. The hype, focused entirely on speed and capability, is optimizing for the wrong variable either way.</p><h3>The questions that cut through it</h3><p>After a few weeks of building everything I could think of, I landed on two filters. Not frameworks, not matrices — just two questions I now ask before starting anything.</p><p><em>Is this solving a real problem?</em></p><p>Not a theoretically real problem. Not a problem I’ve convinced myself exists. A problem someone is actually experiencing, that they’d be relieved to have solved, that they might even pay to fix.</p><p><em>And do I actually care about it?</em></p><p>This one is harder to answer honestly. Paternity leave gave me something that’s easy to lose in the noise of building: perspective on what time actually costs. Not in opportunity-cost terms, but in the simpler sense — time spent is time gone. A new person exists who will not be small for long.</p><p>That’s not an argument for caution. It’s an argument for intention. The tools make it frictionless to start things. They make it easy to stay busy, to feel productive, to ship. What they can’t do is tell you whether the thing you’re building is worth the hours of your life you’re exchanging for it. That question belongs to you.</p><p>The reason meaning matters <em>now</em>, specifically, is that the cost of building things you don’t care about has dropped to nearly zero. You can spend years producing slop at speed and never feel the drag of it — until you look up and realize you’ve been optimizing for output rather than living for anything in particular.</p><p>So the minimum viable version of good work isn’t moving fast. It isn’t even solving real problems, though that’s necessary. It’s building things that are worth the life you’re spending on them.</p><p>The tools are remarkable. The speed is real. Neither of those things decides what matters. That part is still yours.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=28c6594d9f91" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Patience and Impatience]]></title>
            <link>https://medium.com/@nftechie/patience-and-impatience-e44b9557422a?source=rss-7880af92a93a------2</link>
            <guid isPermaLink="false">https://medium.com/p/e44b9557422a</guid>
            <category><![CDATA[entrepreneurship]]></category>
            <category><![CDATA[founders]]></category>
            <category><![CDATA[startup]]></category>
            <dc:creator><![CDATA[Alex Wormuth]]></dc:creator>
            <pubDate>Tue, 23 Dec 2025 16:37:05 GMT</pubDate>
            <atom:updated>2025-12-23T16:42:18.893Z</atom:updated>
            <content:encoded><![CDATA[<h4>A reflection on the necessary virtues for building a successful startup.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eWZh15_KDgvHKVFS94GFFw.png" /></figure><p>Startups are mysterious creatures. Despite being studied for decades, they remain enigmatic and out-of-reach for most. I myself have been studying them for a decade through trial-and-error (mostly error if I’m being truthful).</p><p>The brightest minds claim to outline what it takes to build a successful startup. Yet most still fail. Could they be wrong? Is there something yet to be discovered?</p><p>I’ve worked for startups and have built my own. I’ve been accepted into startup incubators and have been rejected by others. I’ve seen and heard many perspectives on how to build a successful startup. <strong>Most focus on tactics and principles such as talking to users, working 100-hour weeks, and shipping early and often; few focus on virtues.</strong></p><p>Here is my claim: <strong>successful startups require both patience and impatience</strong>. At face value, this statement appears a dichotomy; however, when you unfold it, nuance emerges.</p><p>Let’s unfold it.</p><h3>Patience</h3><p>We’ve all heard the phrase “patience is a virtue,” yet few put it front and center when building a startup.</p><p>I find the Greek word <strong><em>makrothumia</em></strong> more apt for startup founders, which translates to longsuffering, patience, or forbearance. The word also has underlying spiritual implications, which many are foolish to disregard.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Xoxve4Sr5wI3Y5Kb53kGjw.png" /><figcaption>Longsuffering by Gemini.</figcaption></figure><p>Instead of consuming endless podcasts and books on how to be successful, simply turn inward. Learn how to cultivate <em>makrothumia </em>in your daily life.</p><p>Most have a fixation on quick returns and becoming an overnight success, but startups simply don’t work this way. They require longsuffering.</p><p>Is your current venture something you are willing to longsuffer for? If not, spend some time cultivating <em>makrothumia </em>and reflecting on what you are willing to longsuffer for.</p><h3>Impatience</h3><p>Now that I’ve told you to be patient, how is it that I can then turn around and tell you to be impatient? It feels disingenuous, paradoxical, and dare I say curious.</p><p>Yet in the weeds grows the grain. What appears as a dichotomy is actually not one at all. <strong>To build a successful startup, you need to be patient <em>AND</em> impatient simultaneously.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MKUMk6KxRIegEDd9-8YIUA.png" /><figcaption>Impatience by Gemini.</figcaption></figure><p>Let’s look at a practical example:</p><blockquote>You have early signs of product-market-fit and a customer emails you with a bug. This is when you must be impatient. Fix the bug and respond to the user as quickly as you can. But remember, this alone is not enough.</blockquote><p>Impatience is necessary, but not sufficient. Think of impatience like compounding interest. <strong>Every time you are impatient, you are putting a deposit into your “impatience bank” that will grow as a function of your patience.</strong></p><p>The larger your “impatience bank”, the higher the odds your startup will succeed.</p><h3>Balance</h3><p>The balancing act between patience and impatience is what sets successful founders apart. Longsuffering for years or even decades while simultaneously being impatient every day is no easy task, but the few who do it reap the harvest.</p><p>Simply reframe your perspective on startups with patience and impatience in mind, and it may make all the difference for you.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LtXadG1DZ0WEjBphIba00w.png" /><figcaption>Balance by Gemini.</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e44b9557422a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[My First 6 Months on a Rocket Ship: An Inside Look at Building the Financial Infrastructure of the…]]></title>
            <link>https://medium.com/@nftechie/my-first-6-months-on-a-rocket-ship-an-inside-look-at-building-the-financial-infrastructure-of-the-2bd4aa156c7a?source=rss-7880af92a93a------2</link>
            <guid isPermaLink="false">https://medium.com/p/2bd4aa156c7a</guid>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[blockchain]]></category>
            <category><![CDATA[nft]]></category>
            <category><![CDATA[crypto]]></category>
            <dc:creator><![CDATA[Alex Wormuth]]></dc:creator>
            <pubDate>Tue, 30 Jul 2024 00:08:09 GMT</pubDate>
            <atom:updated>2024-07-30T00:08:09.980Z</atom:updated>
            <content:encoded><![CDATA[<h3>My First 6 Months on a Rocket Ship: An Inside Look at Building the Financial Infrastructure of the Future at Tensor</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*a23UnLEYjxLOHiz76OaXMg.jpeg" /></figure><p>When Richard, Tensor’s CTO, asked if I was ready to jump onto a rocket ship, I was filled with a mix of excitement and uncertainty. In the midst of a bear market with a newborn to take care of, I couldn’t help but wonder whether the adventure would be too risky.</p><p>Having built a few startups of my own, I’m no stranger to risk. The process of building products people love is one of the most difficult challenges there is, but within that process lies an immense amount of joy and fulfillment, no matter the outcome.</p><p>Your perspective changes when you have kids, and I thought very deeply about whether joining Tensor would be the right decision for my family. 6 months in, I can say without a doubt it was one of the best decisions I’ve ever made.</p><p>From analyzing Mars’ rover data at the NASA Jet Propulsion Laboratory to building ML pipelines at Google, I’ve never had as much fun and fulfillment building software than I’ve had at Tensor. NFTs and memecoins may look like toys, but what powers them has the potential to become the financial infrastructure of the future (think NASDAQ 2.0).</p><p>It’s difficult to sum up why Tensor has been such a great place to work, but I’ll do my best by sharing what I love most.</p><h3>Incredible culture</h3><p>Tensor has by far one of the best teams I’ve been a part of. Every single person is both a GOAT and extremely humble. We strive for excellence, but we’re not afraid to have fun. We’re really just a bunch of gamers / traders / computer nerds trying to build a better financial future. For me personally, I believe the work we’re doing will influence markets to become more efficient, transparent, and fair for everyone.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*sWON1phZVliV0XFvQcRkug.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AgGjwmcrIHrTBejP5wR14g.jpeg" /></figure><h3>Deep ownership</h3><p>Everyone who works at Tensor is an owner and takes on a lot of responsibility. For my starter project, I built <a href="https://github.com/tensor-hq/smart-rpc">Smart RPC</a>, an open source library for Solana RPCs that handles load balancing, rate limit throttling, failover, fanouts, and retries. It was a little scary to ship a piece of code that helps power indexing and transactions for a protocol with over $2 billion in trading volume, but that’s what true ownership feels like.</p><p>Since then, I’ve worked with the team to add support for new token standards (Token 2022, WNS, Metaplex Core), set up new indexing pipelines to make our system more robust and scalable, revamped various parts of our frontend, and have been working on something new [🙂].</p><p>There’s no shortage of intellectually stimulating, challenging, and high-impact projects at Tensor, which makes every day fun and rewarding.</p><h3>Growth mindset</h3><p>I’m an extremely competitive person. When I’m not coding, I’m competing in triathlons. And in my opinion, startups are one of the toughest sports there is. In crypto, the stakes are high and the competition is tough.</p><p>Startups, in a way, are a lot like endurance sports. You need the discipline and perseverance to get through painful times, the sacrifice to do what it takes to win, and a growth mindset to make small improvements every day.</p><p>Working at Tensor has helped me develop the skills it takes to win both in business and in my personal life.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*m39wHXxHDeHIq-7l9Dw4xg.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Yn8KqZ5mGPf5MRmC-D6GrA.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*X5g2R64FZ34yoUBVlSy1Gg.jpeg" /></figure><h3>What’s next?</h3><p>Looking back at my career, it’s interesting to see how each phase has transformed into the next. I couldn’t have predicted the sequence of events that led me to Tensor, but when the opportunity presented itself it was a no-brainer.</p><p>As for the crypto market as a whole, I’m more excited than ever. I’ve been working in crypto for a few years now, and have seen Solana go from $30 -&gt; $200 -&gt; $8 -&gt; $200. Despite the roller coaster it’s been, I’m more convinced than ever that blockchain will power the financial infrastructure of the future, and I’m excited to help build it.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2bd4aa156c7a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[18 tips to build your blockchain game in days, not months]]></title>
            <link>https://medium.com/fractal-blog/18-tips-to-build-your-blockchain-game-in-days-not-months-ffd9b3735c72?source=rss-7880af92a93a------2</link>
            <guid isPermaLink="false">https://medium.com/p/ffd9b3735c72</guid>
            <category><![CDATA[rest-api]]></category>
            <category><![CDATA[web3-game]]></category>
            <category><![CDATA[blockchain-game]]></category>
            <category><![CDATA[game-development]]></category>
            <category><![CDATA[unity-game-development]]></category>
            <dc:creator><![CDATA[Alex Wormuth]]></dc:creator>
            <pubDate>Tue, 06 Jun 2023 16:21:07 GMT</pubDate>
            <atom:updated>2023-06-06T16:21:07.168Z</atom:updated>
            <content:encoded><![CDATA[<h4>Learn how to quickly implement blockchain elements with FSTUDIO</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2rIiaayMD7TIfOkuwoSWIw.png" /></figure><p>There are a lot of reasons to add blockchain elements into your game. The more pressing issue is, how do you do it without spending months of time and hundreds of thousands of dollars on development? If you’re going to spend that kind of time and money on your game, it should probably go into finding a highly marketable game concept and mechanic.</p><p>That’s why our team at Fractal built <a href="https://www.fractal.is/developers">FSTUDIO </a>— to help game developers like you reap the benefits of the blockchain without any of the complexity. Here are 18 things you can do on the blockchain <strong>today</strong> with FSTUDIO.</p><h3><strong>1. Use REST APIs to build on the blockchain</strong></h3><p>If you’ve already done your research, you know that to build on the blockchain you need to learn a new set of concepts and programming languages — Solidity for Ethereum and Polygon and Rust for Solana.</p><p>It’s definitely possible, but not at all necessary. The better approach is to use what you know, like <a href="https://developers.fractal.is/">REST APIs</a> and SDKs for <a href="https://developers.fractal.is/unity-sdk/introduction">Unity</a>, Unreal, and <a href="https://developers.fractal.is/react-sdk">React</a>.</p><p>In fact, if you aren’t sure which chain you want to go with — and therefore, aren’t sure which language you need to learn — you can still get started on the blockchain with our REST APIs. Then just choose a blockchain (or multiple) to integrate later.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*C1n-Es44m03lwy-QPBb_MA.gif" /></figure><h3><strong>2. Authenticate users in any game engine on any platform</strong></h3><p>Just add <a href="https://developers.fractal.is/wallet/fractal-login">this</a> code snippet to your game to start the authentication process. If you’re a brand new game without an existing account system, this will save you a huge chunk of time, since you won’t need to build a UI with usernames, passwords, and wallet connector buttons. Think of it like Google Sign In for websites — just instead, it’s Fractal.</p><p>There are similar code snippets out there, but most require a browser extension to activate the players’ account or wallet — which isn’t an ideal experience for desktop and mobile games. It’s possible to bypass the browser extension on your own, but you’d need to jump through a lot of hoops or generate a wallet for each game (which defeats the purpose of cross-game wallets).</p><p>Fractal, on the other hand, lets you easily authenticate users across platforms, including web, desktop, and mobile.</p><h3><strong>3. Onboard new players with social sign in</strong></h3><p>Speaking of wallet browser extensions, it’s not even a great user experience for web games, much less desktop and mobile. You’ll likely see a lot of players drop off if you ask them to go download a browser extension, create a seed phrase, and spend time setting up their wallet.</p><p>Instead, you can prompt them to create a Fractal wallet through Google. Just a single tap and they’re all set up and signed in.</p><p>This kind of social sign in flow is very familiar to gamers, which makes it a critical tool in onboarding as many as possible to your blockchain-based games.</p><h3><strong>4. Read your users’ wallets and account states</strong></h3><p>To create the best game experiences, you want to be able to easily check which NFTs and how much in-game currency users have in their account. There are tons of use cases for this:</p><ul><li>See whether a user owns a special Fractal sword and then change the gun in the game</li><li>See that a user owns a special item in one game which unlocks a special level for them in a second game.</li><li>See whether a user has enough crypto to make an in-game purchase and then prompt them to purchase</li></ul><p><a href="https://developers.fractal.is/wallet/access-wallet">This REST API</a> lets you do just that.</p><h3><strong>5. Generate transactions for users to sign</strong></h3><p>Signing on-chain transactions without <a href="https://developers.fractal.is/wallet/sign-transaction">an API</a><strong> </strong>is a mess. You need to manage a connection with an RPC node, integrate with wallet providers, construct transactions, and ensure they’re signed, sent and validated properly on the blockchain.</p><p>Some transaction instructions are simple like <em>transfer SOL from Wallet A to Wallet B</em>, but others, like minting an NFT or anything custom, are much more complicated — since they require multiple signers and multiple accounts.</p><p>If you’re a power user who wants to generate custom transactions for users to sign, <a href="https://developers.fractal.is/wallet/sign-transaction">this REST API</a> lets you do that. Having a single integration to manage authentication, read a user’s wallet state, sign transactions, mint NFTs, and more means fewer points of failure and easier setup and maintenance.</p><p>In the future, we’ll see transaction structures become increasingly complex, as more contracts are created on the blockchain. Then, an API that handles it all will become critical to make sure studios can move fast.</p><h3><strong>6. Forget about securing funds, wallets, and contracts</strong></h3><p>Taking care of users’ funds, wallets, your in-game economy, and minting is high stakes — especially when the ins and outs of blockchain technology are still new to you.</p><p>You need to be extremely careful storing users’ private keys on your own server, since some of their wallets may contain a lot of money. If something goes wrong, or someone attacks your server, the responsibility is 100% yours.</p><p>The same goes for blockchain contracts. If, for example, you’re issuing 1000 tokens and someone finds a mistake in your contract, they may hack your code and mint 10,000,000 — completely ruining the price of your circulating supply. You’ll need to start your economy from scratch and risk upsetting players.</p><p>That’s a lot of responsibility — not to mention time and money to make sure your infrastructure is secure.</p><p>To help you sleep a bit easier at night, Fractal takes on the responsibility of securing it all. You can trust that our code is safe and thoroughly vetted so there’s no chance of attack.</p><h3><strong>7. Automatically stay up to date with the latest standards</strong></h3><p>Each blockchain has different standards and best practices — and they’re constantly evolving. In fact, some chains have mutable contracts, meaning anytime a contract owner makes a breaking change, you’re responsible for keeping your integration up to date. Beyond that, the chains can choose to change their NFT metadata standards, which you’d need to keep up with for anything NFT-related. That’s tedious work better spent building a fun game.</p><p>If you’re using our API, we’re making those updates for you. That way, you have a stable and predictable foundation to build your game, while we deal with the underlying complexity and moving pieces.</p><h3><strong>8. Let your users buy in-game assets with fiat or crypto</strong></h3><p>This one is more for your players — but by making your players’ lives easier, you’re making your life easier too. Usually, to convert fiat currency into crypto, players would need to create a Binance or Coinbase account, send themselves money, wait a few days, swap it to SOL, then transfer it to their Fractal wallet so they can purchase an NFT.</p><p>If your players have a Fractal Account, all they need to do is choose the conversion they want to make on a simple UI (like <em>convert US dollars into SOL</em>) and type in their credit card number. That’s it. The functionality is completely built in for you.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1008/1*5oE1hqJ4F3tuhy0eyok-5A.png" /></figure><h3><strong>9. Create NFT collections and items in a few clicks</strong></h3><p>No-code NFTs — yes they exist. You can make NFT collections super easily on our <a href="https://developers.fractal.is/inventory-management/create-new-items">platform interface</a><strong> </strong>or through <a href="https://developers.fractal.is/minting">this API</a>.</p><p>Just tap ‘Create Item’ then fill out the item name, description, price, and properties. Or, to create items in bulk, you can upload them with a CSV doc or use the API. This way, you can create thousands if not millions of NFTs in one go and automatically add them to the Fractal Marketplace.</p><p>If you have an existing NFT collection on the blockchain, you can still import them via our UI, manage them there, and list them on our marketplace, too.</p><p>Trying to do all this without this API could take you days. You’d need to create your own contract and deploy it, which requires preparing all the images in advance, uploading them to an IPFS provider, and making sure they’re in the correct order. It’s a long process that requires a thorough understanding of Solidity or Rust to carry out. And you have games to focus on.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oGZjuppskinyxaJ3xb5c7g.gif" /></figure><h3><strong>10. Modify NFTs</strong></h3><p>Needing to modify an NFT happens more often than you’d think. Let’s say you:</p><ul><li>made a mistake and want to rename the NFT you misspelled</li><li>want to change the price or metadata (like XP, game progress, item durability)</li><li>want to update the character image or stats as players progress through the game</li></ul><p>But (surprise) modifying NFT after it’s been minted and distributed on a few marketplaces is a hassle. You’d need to figure out where you’re storing the NFT (probably Arweave), reupload the data, and change the metadata on the blockchain to reference the new URL. Then once you do make that change, it might not even be reflected on the marketplaces for some time.</p><p>If you mint NFTs via <a href="https://developers.fractal.is/minting">our API</a>, however, you can modify any metadata at any time, and our backend will make sure it updates on our marketplace immediately.</p><h3><strong>11. Mint NFTs “on demand” when users are ready to buy</strong></h3><p>It takes a lot of upfront budget to finance a minting event — for the art, the blockchain fees, and the development — only to gamble on whether or not you’d sell out.</p><p>The better monetization strategy is to only mint an NFT when someone actually wants to buy it. That way, you can actually make a profit, since you can charge more than the fee. <a href="https://developers.fractal.is/minting">This API</a> lets you do that.</p><p>There are tons of use cases for this “mint on demand” feature:</p><ul><li><strong>Mint badges and achievements</strong> based on in-game actions, and integrate the NFT more seamlessly into the core game loop</li><li><strong>Add a user’s name to an NFT</strong> — which isn’t something you could do ahead of time, since you wouldn’t know who’d purchase it. (We did this, and our players hugely appreciated that their name is now printed on the blockchain forever)</li><li><strong>Mint game characters</strong> — sync NFT metadata with off-chain data as the player progresses or implement staking using meaningful idle game mechanics. (<em>The character is not frequently played, player sends him to the “training ground” or “expedition”)</em></li><li><strong>Mint skins, items, weapons or pets</strong> — mint directly from the game when needed, dynamically expanding collection size. No need to upload static assets and metadata in advance instead generate them on-the-fly.</li></ul><h3><strong>12. Create an in-app purchase store that sells NFTs</strong></h3><p>To build a primary store that sells NFTs, you’d need to process the payments, manage what’s selling and what isn’t, put it on a website so people can buy it, and then take care of all the behind the scenes blockchain work (which as we’ve established previously is months of work).</p><p>Think of Fractal like Shopify — you can use <a href="https://developers.fractal.is/inventory-management/display-your-items">our API</a><strong> </strong>to do all this for you. Basically, the API packages the backend of store management so you can easily create items and display them on a web or in-game store.</p><p>Plus, if a player goes to your store and buys an NFT, and they’re logged into their Fractal Account, the API will automatically generate a buy/mint transaction for that item and then initiate the signing process.</p><h3><strong>13. Embed a secondary marketplace in your game</strong></h3><p>Just like you can build a primary store in your game, you can also embed the Fractal Marketplace to function as a secondary store, too. So players can sell the crops they harvested, for example.</p><p>Instead of building a secondary store from scratch (which on top of all the usual store management stuff, you’d also need to keep track of all the ownership changes and maintain a cache), <a href="https://developers.fractal.is/marketplace/selling-item">our APIs</a> do it for you. You just build the UI and ask our API to tell you which items are currently listed, and we’ll give them back to you so you can present the items to your players however you want.</p><p>In fact, you can even add a simple button to your game UX enabling players to list new items. Our API generates the listing transaction and automatically adds it to the Fractal Marketplace.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8K8jOfdfSXvrFh4hiSWSMw.png" /></figure><h3><strong>14. Connect to Ethereum, Polygon, AND Solana</strong></h3><p>Choosing a blockchain before you start building limits what you can do with your game. That’s why we made Fractal chain agnostic, so you don’t need to even have a chain to use most of our APIs (until you want to mint an item, of course).</p><p>This way, you can set everything else up while you’re deciding which chain — or chains — you want to support. If you chose Polygon first but want to add Solana later, you still have the choice to do that.</p><p>If you want to mint on a chain we don’t support, reach out to us and we’ll see how we can help.</p><h3><strong>15. Distribute your game on a launcher, whether you’re using blockchain or not</strong></h3><p>It’s true there are quite a few launchers that support blockchain games today, but not many have built in account management. That means, users with a Fractal Account don’t need to take any extra steps to access their wallet to play your game — the moment they install the launcher, they’re ready to go.</p><p>To get your game on Fractal Launcher, all you need to do is upload a .zip file of your game build to our platform. It just takes a few minutes and once you’re done, your game is ready to play.</p><h3><strong>16. Run play tests on our launcher</strong></h3><p>If your game is in closed alpha or beta, and you’re looking to run play tests, you can choose to distribute access codes to select players through Fractal. This gives them full access to play your game on Fractal Desktop. Just tap ‘Generate Access Codes’ and deliver them to your players.</p><h3><strong>17. Update your desktop game with incremental patches</strong></h3><p>Like #8, this is more about easily providing your players with a great experience. Instead of forcing them to download your game all over again each time you push an update, you can let them simply download the updates — minimizing download times from hours to minutes.</p><p>This is critical, especially when you’re running events — as the event might be over by the time the game finishes downloading.</p><p>Fractal Desktop automatically generates the incremental patches for the players, so you don’t need to worry about your game size impacting the player experience.</p><h3><strong>18. Run a leaderboard tournament</strong></h3><p>Our <a href="https://developers.fractal.is/tournaments/tournaments">Tournament API</a> is open to the public, whether you run the tournament on Fractal or not. If you don’t run it on Fractal, you can use the API to build a tournament infrastructure in your game.</p><p>If you do run it with Fractal, you’ll need to get approval from our team first. The API enables you to post scores to the Fractal leaderboard and then fetch the leaderboard scores back in your game. You’ll be featured on the Fractal site and get access to the Fractal community. Once the tournament goes live, you’ll have a steady stream of players creating accounts in your game, trying it out, and providing feedback.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JJ5-pIy9rTYh_qcIKMSDuw.png" /></figure><h3><strong>Stay tuned for more</strong></h3><p>This list is just the start. We plan to keep adding more use cases and functionalities into FSTUDIO that make it easy for you to integrate blockchain elements into your game, without asking you to learn about the blockchain or teach your players. Stay tuned.</p><h3><strong>Ready to start building on the blockchain?</strong></h3><ul><li><a href="https://studio.fractal.is/">Create your FSTUDIO account</a></li><li><a href="https://developers.fractal.is/">Read our dev docs</a></li><li><a href="https://www.fractal.is/developers">Learn more</a></li><li><a href="https://fractal.is/contact">Get in touch</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ffd9b3735c72" width="1" height="1" alt=""><hr><p><a href="https://medium.com/fractal-blog/18-tips-to-build-your-blockchain-game-in-days-not-months-ffd9b3735c72">18 tips to build your blockchain game in days, not months</a> was originally published in <a href="https://medium.com/fractal-blog">Fractal</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Machine learning using Google Colab vs. GCP virtual machine vs. Vast.ai vs. your own hardware]]></title>
            <link>https://medium.com/fib0n4cc1/google-colab-vs-gcp-virtual-machine-vs-vast-ai-vs-your-own-hardware-5215a2335780?source=rss-7880af92a93a------2</link>
            <guid isPermaLink="false">https://medium.com/p/5215a2335780</guid>
            <category><![CDATA[google-cloud-platform]]></category>
            <category><![CDATA[tensorflow]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[google-colab]]></category>
            <dc:creator><![CDATA[Alex Wormuth]]></dc:creator>
            <pubDate>Sun, 17 Oct 2021 18:52:06 GMT</pubDate>
            <atom:updated>2021-10-17T18:53:14.816Z</atom:updated>
            <content:encoded><![CDATA[<p>Hello fellow machine learning enthusiast. Your interest in this comparison tells me that you’re interested in improving your model training efficiency, and you’ve come to the right place.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/742/1*gejeBAG1aRDFWOL8Uxz3dg.png" /><figcaption>source: <a href="https://xkcd.com/1838/">https://xkcd.com/1838/</a></figcaption></figure><p>Below I will outline various options you have to train models, and discuss tradeoffs for each approach.</p><h3>Google Colab</h3><p><a href="https://colab.research.google.com/">Google Colab</a> is an easy way to start training and validating models in your browser. It doesn’t require any hardware setup and has many examples to get you started. I highly recommend you start here when training your first models.</p><h4>Pros:</h4><ul><li>Free (they also have paid tiers for more access to GPU/TPU resources).</li><li>Execution speed is quite good for being a free service.</li><li>Easy to get started. There are lots of tutorials, and the interface is visual and user-friendly.</li></ul><h4>Cons:</h4><ul><li>Less programmable. If you want to set up more complex experiment pipelines in a more flexible way, this may not be the solution for you.</li><li>Training data management and upload speed. If you’re experimenting with different features and assumptions in your training data, it will quickly become tedious to manage. Every time you make a change, you’ll need to re-upload your training data to Google Colab. If you’re working with large datasets, this can be time-consuming.</li></ul><h3>Google Cloud Platform (GCP) GPU-enabled Virtual Machine</h3><p>This option is for more advanced users, and those who want more control and flexibility over the training and experimentation pipelines.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*awKGSf76WQbNywbNIuAEDw.png" /></figure><h4>Pros:</h4><ul><li>You get to choose the resources your VM uses, so you’ll have more predictability and control over costs and training speed.</li><li>Reasonably affordable. For a decent starter machine with 1 GPU, you’re looking at about $.35/hr. Remember, you don’t need to run this machine 24/7 unless you’re constantly running experiments. So if you turn the instance off when you’re not using it, the cost isn’t too bad.</li><li>Programmability. Using your own VM, you have a lot more control. You can create experiment pipelines, integrating data generation pipelines with training pipelines, implement experiment tracking, etc.</li></ul><h4>Cons:</h4><ul><li>Complex and time-consuming to set up. To get your training code running on the VM will take some work. There are various tools you’ll need to set up manually, including Python, Tensorflow, and CUDA, to name a few. Whereas with Google Colab, this is all set up for you. Here is a <a href="https://medium.com/fib0n4cc1/how-to-set-up-cuda-cudnn-and-tensorflow-on-ubuntu-20-04-the-right-way-1ddcca2910bf">setup guide</a> to help you set up all these dependencies on your VM.</li></ul><h3>Vast.ai</h3><p><a href="https://vast.ai/">Vasi.ai</a> is a GPU sharing economy where people can rent our their GPUs.</p><h4>Pros:</h4><ul><li>Similar to GCP, you can choose the resources your VM uses.</li><li>Affordable. You can’t beat the price for the type of hardware you can rent. Compared to GCP and AWS, you can often get cheaper rates on vast.ai.</li><li>Since you’re renting a VM, you have a lot more control.</li></ul><h4>Cons:</h4><ul><li>Complex and time-consuming to set up. Here is a <a href="https://medium.com/fib0n4cc1/how-to-set-up-cuda-cudnn-and-tensorflow-on-ubuntu-20-04-the-right-way-1ddcca2910bf">setup guide</a> to help you set up all these dependencies on your VM. Note that some of the VMs may already come set up with these dependencies, which will save you a lot of time.</li><li>Security. Compared to GCP, which has many layers of security in place for your data, vast.ai isn’t able to provide as many guarantees. You could be running your training on some random person’s computer at home, which may not provide enough security for the training you’re doing.</li></ul><h3>Buy your own hardware</h3><p>Last but not least, you could consider buying your own hardware.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5haEuuNUTkMRfou56pMPdg.jpeg" /></figure><h4>Pros:</h4><ul><li>Programmability, security, speed, etc. There are many pros to this approach.</li></ul><h4>Cons:</h4><ul><li>Cost. This was the biggest barrier for me. To build your own PC with GPU / CPU / RAM resources similar to the above options, you could be looking at $5,000 — $10,000 for a solid machine. This doesn’t include the energy your machine will be using if you’re training 24/7. If you’re planning to invest in machine learning for many years, it may eventually pay off. But I’d recommend you first start with the above approaches, as you can always control when you are spending money.</li><li>Complex and time-consuming to set up. Similarly to the above VM options, you’ll need to make sure you have the correct system requirements to train models. Here is a <a href="https://medium.com/fib0n4cc1/how-to-set-up-cuda-cudnn-and-tensorflow-on-ubuntu-20-04-the-right-way-1ddcca2910bf">setup guide</a> to help you set up all these dependencies on your VM.</li></ul><p>Whichever option you choose, the world will be better off because of your investment in machine learning. Happy training!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*P132Us8Hh7P6GAkSviyGwQ.jpeg" /></figure><p>0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, … ∞</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5215a2335780" width="1" height="1" alt=""><hr><p><a href="https://medium.com/fib0n4cc1/google-colab-vs-gcp-virtual-machine-vs-vast-ai-vs-your-own-hardware-5215a2335780">Machine learning using Google Colab vs. GCP virtual machine vs. Vast.ai vs. your own hardware</a> was originally published in <a href="https://medium.com/fib0n4cc1">fib0n4cc1</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to set up CUDA, cuDNN, and Tensorflow on Ubuntu 20.04 the right way]]></title>
            <link>https://medium.com/fib0n4cc1/how-to-set-up-cuda-cudnn-and-tensorflow-on-ubuntu-20-04-the-right-way-1ddcca2910bf?source=rss-7880af92a93a------2</link>
            <guid isPermaLink="false">https://medium.com/p/1ddcca2910bf</guid>
            <category><![CDATA[cuda]]></category>
            <category><![CDATA[ubuntu]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[tensorflow]]></category>
            <category><![CDATA[deep-learning]]></category>
            <dc:creator><![CDATA[Alex Wormuth]]></dc:creator>
            <pubDate>Sun, 17 Oct 2021 18:16:50 GMT</pubDate>
            <atom:updated>2021-10-17T18:20:58.261Z</atom:updated>
            <content:encoded><![CDATA[<p>It was a Saturday afternoon — cold as usual in Colorado. Having spent many days working with Google Colab, I decided to spin up my own virtual machine for more control and programmability over model training and experiments.</p><p>After failing 5 times to setup a GPU on Tensorflow, I finally succeeded! However, I went through many different setup articles that steered me in the wrong direction and ended up wasting hours of my time. Because of this, I decided to write my own article in the hopes of saving your time and frustration.</p><h3>Manual vs. Docker setup</h3><p>This article will focus on manual setup, but it’s worth noting that you can achieve similar results with Docker. <a href="https://www.tensorflow.org/install/docker">https://www.tensorflow.org/install/docker</a> is a good starting point if you wish to proceed down this path.</p><h3>Manual setup on Ubuntu 20.04 — let’s go!</h3><h4>Step 1: Create a GPU-optimized VM running Ubuntu 20.04</h4><p>I used Google Cloud Platform (GCP), but AWS and other platforms may have similar products.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*awKGSf76WQbNywbNIuAEDw.png" /></figure><h4>Step 2: SSH into your newly created VM</h4><h4>Step 3: Make sure you have Python (3.8 or 3.9) and pip installed (for working with Tensorflow)</h4><p>If you don’t already have Python installed, <a href="https://linuxize.com/post/how-to-install-python-3-9-on-ubuntu-20-04/">here is a guide</a> for installing Python on Ubuntu 20.04.</p><p>You’ll also need to make sure pip is installed. If not, <a href="https://linuxize.com/post/how-to-install-pip-on-ubuntu-20.04/">here is a guide</a>.</p><h4>Step 4: Install CUDA</h4><p>You’ve made it this far, and now is where you really need to pay attention. This is the part that will cause you the most pain, but it’s quite simple if you read these instructions carefully.</p><p>After reading through many simple “how to set up CUDA” blog posts, my suggestion is to follow the <a href="https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html">step-by-step guide</a> provided by Nvidia. I know what you’re thinking. This looks like some old manual from the 1980s with a ton of boring information. You may be right, but it’s actually just a few simple steps.</p><ol><li>Check hardware and software prerequisites</li></ol><pre><strong>$</strong> lspci | grep -i nvidia</pre><pre><strong>$</strong> uname -m &amp;&amp; cat /etc/*release</pre><pre><strong>$</strong> gcc --version</pre><pre><strong>$</strong> uname -r</pre><p>2. Install Ubuntu headers</p><pre><strong>$</strong> sudo apt-get install linux-headers-$(uname -r)</pre><p>3. Download the NVIDIA CUDA Toolkit</p><p><a href="https://developer.nvidia.com/cuda-downloads?target_os=Linux&amp;target_arch=x86_64&amp;Distribution=Ubuntu&amp;target_version=20.04&amp;target_type=deb_local">Here are the commands</a> (from the NVIDIA website) that you need to run.</p><pre><strong>$</strong> wget <a href="https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin">https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin</a></pre><pre><strong>$ </strong>sudo mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600</pre><pre><strong>$ </strong>wget <a href="https://developer.download.nvidia.com/compute/cuda/11.4.2/local_installers/cuda-repo-ubuntu2004-11-4-local_11.4.2-470.57.02-1_amd64.deb">https://developer.download.nvidia.com/compute/cuda/11.4.2/local_installers/cuda-repo-ubuntu2004-11-4-local_11.4.2-470.57.02-1_amd64.deb</a></pre><pre><strong>$ </strong>sudo dpkg -i cuda-repo-ubuntu2004-11-4-local_11.4.2-470.57.02-1_amd64.deb</pre><pre><strong>$ </strong>sudo apt-key add /var/cuda-repo-ubuntu2004-11-4-local/7fa2af80.pub</pre><pre><strong>$ </strong>sudo apt-get update</pre><pre><strong>$ </strong>sudo apt-get -y install cuda</pre><p>4. Update environment variables by adding the following to your ~/.bashrc file.</p><pre><strong>$</strong> export PATH=/usr/local/cuda-11.4/bin${PATH:+:${PATH}}</pre><pre><strong>$</strong> export LD_LIBRARY_PATH=/usr/local/cuda-11.4/lib64\<br>                         ${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}</pre><p>5. Reboot your machine. This step is very important, as the CUDA installation will not be usable until you do so.</p><p>6. Post-install checks</p><p>a) Check your Cuda version (it should be 11.4)</p><pre><strong>$ </strong>nvcc --version</pre><p>b) Verify your driver version (this should not throw an error)</p><pre><strong>$ </strong>cat /proc/driver/nvidia/version</pre><h4>Step 4: Install cuDNN</h4><p>Similar to my above advice, I’d recommend following <a href="https://docs.nvidia.com/deeplearning/cudnn/install-guide/index.html">Nvidia’s own documentation.</a> I know it’s painful, but you’ll thank me later.</p><p>To summarize the documentation, there a basically a few different approaches you can take to installing cuDNN. The first approach is downloading a zip file and manually installing cuDNN.</p><p>Alternatively, you can use apt-get to install cuDNN (recommended).</p><ol><li>Enable the Ubuntu cuDNN repository</li></ol><pre><strong>$</strong> wget <a href="https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin">https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin</a></pre><pre><strong>$</strong> sudo mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600</pre><pre><strong>$</strong> sudo apt-key adv --fetch-keys <a href="https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/7fa2af80.pub">https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/7fa2af80.pub</a></pre><pre><strong>$</strong> sudo add-apt-repository &quot;deb <a href="https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/">https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/</a> /&quot;</pre><pre><strong>$</strong> sudo apt-get update</pre><p>2. Install cuDNN</p><pre><strong>$</strong> sudo apt-get install libcudnn8=8.2.4.*-1+cuda11.4</pre><pre><strong>$</strong> sudo apt-get install libcudnn8-dev=8.2.4.*-1+cuda11.4</pre><h4>Step 5: Install Tensorflow</h4><pre><strong>$</strong> pip install tensorflow</pre><h4>Step 6: Run your code!</h4><p>If all went went, your code should run and Tensorflow should be using the GPU. You might see some warnings, but you shouldn’t see any major issues with libraries or GPUs not being found.</p><p>If you don’t see any errors, and you want to check whether the GPU is actually being used, you can run the following in a separate tab while your code is executing:</p><pre><strong>$</strong> nvidia-smi</pre><h4>Step 7: Celebrate 🎉</h4><p>Enjoy your newly configured machine. Happy training!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*P132Us8Hh7P6GAkSviyGwQ.jpeg" /></figure><p>0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, … ∞</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1ddcca2910bf" width="1" height="1" alt=""><hr><p><a href="https://medium.com/fib0n4cc1/how-to-set-up-cuda-cudnn-and-tensorflow-on-ubuntu-20-04-the-right-way-1ddcca2910bf">How to set up CUDA, cuDNN, and Tensorflow on Ubuntu 20.04 the right way</a> was originally published in <a href="https://medium.com/fib0n4cc1">fib0n4cc1</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to develop profitable crypto trading strategies]]></title>
            <link>https://medium.com/fib0n4cc1/how-to-develop-profitable-crypto-trading-strategies-58ca3d970a76?source=rss-7880af92a93a------2</link>
            <guid isPermaLink="false">https://medium.com/p/58ca3d970a76</guid>
            <category><![CDATA[crypto-trading]]></category>
            <category><![CDATA[trading-strategy]]></category>
            <category><![CDATA[day-trading]]></category>
            <category><![CDATA[quantitative-finance]]></category>
            <category><![CDATA[cryptocurrency]]></category>
            <dc:creator><![CDATA[Alex Wormuth]]></dc:creator>
            <pubDate>Thu, 14 Oct 2021 19:30:03 GMT</pubDate>
            <atom:updated>2021-10-14T19:48:10.668Z</atom:updated>
            <content:encoded><![CDATA[<p>The journey to a profitable strategy is tough. It feels like a roller coaster ride, and you’re going to be unsuccessful most of the time. But with hard work, analytical thinking, and the scientific method, you can increase your chances of developing a profitable trading strategy.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*N-12ICnSgAT-FF27qfP3vA.png" /><figcaption>ADA/USD 15m Kraken price chart</figcaption></figure><h3>Where to start?</h3><p>There are an overwhelming number of resources out there to help traders develop profitable strategies. Some claim to give you the key to success, while others say it’s impossible. Neither of these are good resources. Markets are predictable, but there is no simple, easy way to develop a profitable strategy.</p><ol><li>Learn the basics about day trading</li><li>Trade manually first (preferably using a simulator)</li><li>Identify which coins you want to trade on a cryptocurrency exchange</li><li>Experiment using the scientific method</li></ol><h3>Types of strategies</h3><ol><li>Buy and hold. This is the simplest to execute, but you should aim to beat this strategy.</li><li>Swing trade using indicators, such as <a href="https://www.investopedia.com/terms/r/rsi.asp">RSI</a> and <a href="https://www.investopedia.com/terms/m/macd.asp">MACD</a></li><li>Swing trade using machine learning and model predictions. Here are some guides for developing <a href="https://medium.com/analytics-vidhya/how-i-used-a-random-forest-classifier-to-day-trade-for-2-months-part-i-9c00d96d254c">Random Forests</a> and <a href="https://towardsdatascience.com/unsupervised-learning-to-market-behavior-forecasting-ee8f78650415">Hidden Markov Models</a> to predict price movements.</li></ol><h3>Using the scientific method to improve strategies</h3><p>Developing strategies using indicators and models can be overwhelming. More often than not, they will be unprofitable. Even after doing thorough backtests and simulations, your strategy can underperform on real market data. Your best bet is to follow the scientific method to improve your strategy. Come up with a hypothesis about what a good strategy contains (e.g. trading volume, popular tweet), and then test your hypothesis. The faster you can iterate and reject different hypotheses, the faster you will get to a profitable strategy.</p><h3>Using Pixels to iterate on strategies quickly</h3><p><a href="https://pixels.fund/">Pixels</a> is a free tool to help you develop and test strategies using indicators and machine learning. By running experiments and simulating your strategy on live market data, you’ll learn more quickly and develop better strategies.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fplayer.vimeo.com%2Fvideo%2F564438809%3Fh%3Def3917b8d3%26app_id%3D122963&amp;dntp=1&amp;display_name=Vimeo&amp;url=https%3A%2F%2Fvimeo.com%2F564438809&amp;image=https%3A%2F%2Fi.vimeocdn.com%2Fvideo%2F1166815881-5b2341449744bca36173d6ae7db9d4d4365223c2606e4b3129b41403033502b6-d_1280&amp;key=d04bfffea46d4aeda930ec88cc64b87c&amp;type=text%2Fhtml&amp;schema=vimeo" width="1298" height="720" frameborder="0" scrolling="no"><a href="https://medium.com/media/bfc947de9730053d6760192e87d3461a/href">https://medium.com/media/bfc947de9730053d6760192e87d3461a/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Wv9qgXMWp87xYy3d9BVk0g.jpeg" /></figure><p>0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, … ∞</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=58ca3d970a76" width="1" height="1" alt=""><hr><p><a href="https://medium.com/fib0n4cc1/how-to-develop-profitable-crypto-trading-strategies-58ca3d970a76">How to develop profitable crypto trading strategies</a> was originally published in <a href="https://medium.com/fib0n4cc1">fib0n4cc1</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Building a crypto trading bot on Kraken]]></title>
            <link>https://medium.com/fib0n4cc1/building-a-crypto-trading-bot-on-kraken-4f1d6eafedd9?source=rss-7880af92a93a------2</link>
            <guid isPermaLink="false">https://medium.com/p/4f1d6eafedd9</guid>
            <category><![CDATA[cryptocurrency]]></category>
            <category><![CDATA[kraken]]></category>
            <category><![CDATA[day-trading]]></category>
            <category><![CDATA[crypto-trading]]></category>
            <category><![CDATA[quantitative-finance]]></category>
            <dc:creator><![CDATA[Alex Wormuth]]></dc:creator>
            <pubDate>Thu, 14 Oct 2021 19:07:45 GMT</pubDate>
            <atom:updated>2021-10-14T19:07:45.905Z</atom:updated>
            <content:encoded><![CDATA[<p>A few months ago, we set out to build a crypto trading bot. After exploring various exchanges such as Coinbase and Binance, we decided to focus our trading efforts on Kraken.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RZAvDJOg6LL5Yfvvn1H7Qg.jpeg" /></figure><h3>Factors that may influence which exchange you choose to run your bot on:</h3><ol><li>Trust</li><li>Security</li><li>Fees</li><li>Coin pairs</li><li>Slippage</li></ol><p>If you’re running an arbitrage strategy, you may need to connect to multiple exchanges. But for the purposes of this discussion, we’ll focus on strategies that can be executed on a single exchange.</p><p>We prefer trust and security over everything else. If you’re making a profit, but there’s a chance that the exchange you’re using will get hacked or get shut down, it’s not worth it. We’re willing to accept higher fees, less coin selection, and more slippage in exchange for trust and security.</p><h3>How to build a trading bot from scratch</h3><p>If you have programming expertise, you can take a look at <a href="https://docs.kraken.com/rest/">Kraken’s API</a>. You’ll need to pull OHLC (open, high, low, close) data at whatever time interval makes sense for you (we typically use 1m or 5m intervals).</p><p>Once you have OHLC data, you’ll need to generate indicators. We use <a href="https://mrjbq7.github.io/ta-lib/">ta-lib</a>, which is a Python library that lets you generate 100s of indicators from OHLC data.</p><p>Next, you’ll need to develop a strategy which turns indicator values into buy/sell decisions. For example, a simple but unprofitable strategy might be:</p><p>If RSI &lt; 20: Buy</p><p>If RSI &gt; 80: Sell</p><p>Lastly, you’ll need to execute these decisions on Kraken. Kraken has APIs for placing market/limit buy and sell orders.</p><h3>How to build a trading bot using Pixels</h3><p><a href="https://pixels.fund/">Pixels</a> is free tool for building, testing, and deploying trading bots. It’s great if you don’t have programming expertise, or if you don’t have time to build the data pipelines and integrations yourself. You just need to focus on developing and testing your own strategies.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fplayer.vimeo.com%2Fvideo%2F564438809%3Fh%3Def3917b8d3%26app_id%3D122963&amp;dntp=1&amp;display_name=Vimeo&amp;url=https%3A%2F%2Fvimeo.com%2F564438809&amp;image=https%3A%2F%2Fi.vimeocdn.com%2Fvideo%2F1166815881-5b2341449744bca36173d6ae7db9d4d4365223c2606e4b3129b41403033502b6-d_1280&amp;key=d04bfffea46d4aeda930ec88cc64b87c&amp;type=text%2Fhtml&amp;schema=vimeo" width="1298" height="720" frameborder="0" scrolling="no"><a href="https://medium.com/media/bfc947de9730053d6760192e87d3461a/href">https://medium.com/media/bfc947de9730053d6760192e87d3461a/href</a></iframe><p>Check out our article on developing profitable strategies if you’re interested in learning more!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Wv9qgXMWp87xYy3d9BVk0g.jpeg" /></figure><p>0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, … ∞</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4f1d6eafedd9" width="1" height="1" alt=""><hr><p><a href="https://medium.com/fib0n4cc1/building-a-crypto-trading-bot-on-kraken-4f1d6eafedd9">Building a crypto trading bot on Kraken</a> was originally published in <a href="https://medium.com/fib0n4cc1">fib0n4cc1</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Start with why]]></title>
            <link>https://medium.com/fib0n4cc1/start-with-why-7e87b24671c6?source=rss-7880af92a93a------2</link>
            <guid isPermaLink="false">https://medium.com/p/7e87b24671c6</guid>
            <category><![CDATA[startup]]></category>
            <dc:creator><![CDATA[Alex Wormuth]]></dc:creator>
            <pubDate>Thu, 14 Oct 2021 18:33:53 GMT</pubDate>
            <atom:updated>2021-10-14T18:33:53.769Z</atom:updated>
            <content:encoded><![CDATA[<p>Simon Sinek popularized the phrase “Start with why” in his TED talk many years ago.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2Fu4ZoJKF_VuA%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3Du4ZoJKF_VuA&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2Fu4ZoJKF_VuA%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="640" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/28419ca3a639e5c5132c46281a95abaa/href">https://medium.com/media/28419ca3a639e5c5132c46281a95abaa/href</a></iframe><p>This phrase is a guiding force for many companies, including our own. “Start with why” forces you to reflect on why you’re doing what you’re doing. Time is our most valuable asset, and the “why” behind how we’re spending our time can often be the source to unlocking passion and motivation for solving difficult problems.</p><p>So what is our why? We believe blockchains can revolutionize society. Whether it’s through new financial or social interactions, increased trust and equality, or more opportunities for entrepreneurs and artists, blockchains can have a positive and profound impact on humanity. Simply put, our why is to help make this happen. By actively engaging in discussions, investing in startups, providing liquidity, and more, we can help increase adoption of blockchain technology and accelerate the path to positive impact on society.</p><p>Blockchain technology is still quite new, and we have many things to learn about how it can be used responsibly. As we navigate this journey through the lens of a for-profit company, we’ll share thoughts, difficulties, and insights we encounter along the way.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Wv9qgXMWp87xYy3d9BVk0g.jpeg" /></figure><p>0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, … ∞</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7e87b24671c6" width="1" height="1" alt=""><hr><p><a href="https://medium.com/fib0n4cc1/start-with-why-7e87b24671c6">Start with why</a> was originally published in <a href="https://medium.com/fib0n4cc1">fib0n4cc1</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>