Matt “Friend of the pod” Rogish 🇺🇸 (@MattRogish) header
Matt “Friend of the pod” Rogish 🇺🇸 (@MattRogish) profile picture
Philadelphia, PA
Joined September 2007
Photo by Matt “Friend of the pod” Rogish 🇺🇸 (@MattRogish)Photo by Matt “Friend of the pod” Rogish 🇺🇸 (@MattRogish)Photo by Matt “Friend of the pod” Rogish 🇺🇸 (@MattRogish)Photo by Matt “Friend of the pod” Rogish 🇺🇸 (@MattRogish)Photo by Matt “Friend of the pod” Rogish 🇺🇸 (@MattRogish)Photo by Matt “Friend of the pod” Rogish 🇺🇸 (@MattRogish)Photo by Matt “Friend of the pod” Rogish 🇺🇸 (@MattRogish)Photo by Matt “Friend of the pod” Rogish 🇺🇸 (@MattRogish)Photo by Matt “Friend of the pod” Rogish 🇺🇸 (@MattRogish)Photo by Matt “Friend of the pod” Rogish 🇺🇸 (@MattRogish)
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
talked to someone today who was distraught and said basically everyone just ships whatever they want, no one is talking to each other, and no one has any reason to say "no" to any particular feature being added so they just add everything
If you read Linear's changelog, the vast majority of what it's shipped in the past few months has been either building new features that aren't part of its core product, or fixing those features that aren't part of its core product Think this is an interesting and challenging aspect of AI-accelerated engineering productivity: mostly it allows teams to move incredibly quickly in new directions. There might be legitimate value in this in the long term, since teams can learn faster, but I'm not convinced it's better than simply focusing on your core product and on your customers
21
11
3
524
47,278
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
fable 5.2 is a clear departure from older claude models in terms of thinking, subagents, retries and refactoring. adjust your prompts and mental models accordingly if you want to use fable 5.2 in your workflows.
22
23
2
1,192
205,156
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
We've given all Pro, Max, Ultra, and Team users one free full reset, across all models and limits. Use it anytime before December 10 at 2:00PM PT. We recently fixed a few small issues that used more limits than intended, and the Fable resets didn't do much for those of you who'd already hit your weekly limits for all models. While these issues didn't affect everyone, we want to make up for the trouble. You'll find it in Settings > Usage, or at clau.de/reset.
1,890
2,895
2,007
46,916
3,515,620
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
People who didn’t buy Nvidia this year
236
4,139
221
69,047
1,670,782
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
It’s not an overstatement to say that creating software without the aid of AI at this point is basically absurd. Like, that’s a preposterous, ridiculous thing to do.
guys i don't know how to break this to you but for the rest of your life every piece of software you use will have been made with ai. every website, every video game, every app, Everything. it is a non-negotiable in every development position i saw while job hunting.
49
69
3
2,147
34,783
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
Anthropic found a way to make Claude build better-looking apps, and the trick isn't a better design prompt Instead of asking Claude to make an interface “beautiful,” researchers separated the agent that builds it from the one that evaluates it. The evaluator checks concrete criteria, including visual coherence and originality, then feeds the results back into another iteration. This turns subjective design feedback into a repeatable improvement process. If you're building with Claude Code, try the same approach: Define visual standards, let a separate agent critique the output, and require revisions before approval.
Claude just made a key quantum computing operation 12x cheaper, and the method matters far beyond quantum physics Anthropic gave Claude eight circuit-design problems and an automated evaluator that checked correctness and measured computational cost. Rather than tweaking existing solutions, Claude explored new algorithms, reconstructed intermediate data instead of storing it, and tested its designs. One quantum decoder ended up requiring 12x fewer gates than a previous construction. The bigger lesson for AI builders: give agents measurable objectives, independent verification and room to explore competing approaches. > The most valuable agents won't just write code > They'll discover better ways to solve problems Source: anthropic.com/research/desig…
32
17
2
418
39,007
Remember when the government turned off Fable for about 3 weeks? That feels like a lifetime ago
71
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
I encourage anyone who doubts this to try and make a real AAA game and apply your own real critical game quality standards to it. I am making one right now and while Opus 5.5 (Fable and Astra can't do it) can do many of the coding tasks, it simply isn't able to make gameplay design and UI design decisions at the level of detail and correctness demanded by a GAME. My theory is that it's because the quality requirement in a game is ironically higher than for e.g. a SaaS application. SaaS or other "work" applications provide economic value for using it, so as long as your economic value gained is high enough, you'll tolerate bugs and UI imperfections and lag, even outright illogical app behavior - as long as it does the thing you need it to do. But when you're playing a game, your willingness to continue doing so rests entirely on the quality of the experience. And that's impacted by gameplay, fun, UI, story, etc. If there's any sort of "irritant" like a part of the interface is janky, or something consistently looks wrong, or the gameplay does something illogical, or it's laggy, etc - you don't have any incentive to keep going. To keep someone playing a game, the experience has to be far better than that - maybe not perfect - but it still has to be great. What I'm finding is the Opus 5.5 will make a game at roughly between 50% to 85% quality level. If you give it a lazy prompt, it'll puke out something with blocky graphics and barely-working gameplay: 50%. If you iterate a bit and give it more details, you can get it up to 85% or so - playable, can generate impressive screenshots and maybe even gameplay clips - but that last 15% or so still needs hand-tuning and human taste-making. You can have it read every game design and UI/UX guide you can find, and it just won't build anything better than "kind of workable." And "kind of workable" is often enough for a work app. I have like a dozen vibe-coded utility apps that roughly do the thing I need them to do (track stuff, manage some kind of data, visualize something) but they're not worth polishing because they're good enough to perform the function. But with games I can imagine someone else looking at it with the same critical eye that I would, if I were playing it: - these buttons are a little out of place, the alignment isn't perfect - why does this icon which does the same thing as this other one look different? - this tree or shadow looks weird from this angle - this game element acts a little janky under these conditions - there's lag when stuff gets busy In particular, AI doesn't seem to be able to create really good custom UIs. If you have a bunch of commonplace functions then yeah, it'll use a UI template. But if you have a game, and the game involves doing a bunch of unusual (fun!) things, you want the UI to engage the user's intuition about it, show the right responsiveness or progress of an action, act correctly in certain unusual corner cases (because fun often involves corner cases) etc, and AI can't quite get all of it right yet. It's workable, you can often see the vision - but it's not a game. No AAA game studio would ship what AI produces today. Much of this seems to be about paying attention to the right details. Maybe subsequent models will be better (they often are). But right now a human still has to look at every little corner of the game, decide if it's acting just right, then specify in extensive detail exactly what should happen and why, and what overall effect is intended. The good news is that one person can be the gameplay designer, the artistic director, the engine developer, the interface designer - you can do all of these jobs and let the AI carry out the tasks. The bad news is that one person has to be the gameplay designer, the artistic director, the engine developer, the interface designer... and I'll probably have to be the marketing guy later when/if the game is released. Consequently, I've had to quickly learn the fundamentals of all these disciplines. Unsurprisingly given my SWE background, the engine on my game is quite well-optimized even with tons and tons of moving things and physics. But the gameplay is another thing and I don't even know the right terminology to describe a lot of the art/graphics things I want to see, and... oh my god, UI design is an AI clusterfuck... Back in the day, tech companies sort of had a hierarchy of respect when it came to technical staff - I'm a little ashamed to say - it was roughly: 1. backend engineer 2. frontend engineer 3. visual designer 4. UI/UX designer Yes, the lowly UI/UX designer, at the bottom of the totem pole - not even considered "real" designers - they were ones responsible for deciding where to put the buttons and what fonts to use. Well, today the AI does everything else but it still can't design a great interface worth a damn. So yeah, to the UI/UX designers I knew: I owe you an apology. I wasn't really familiar with your game. Of all of the tasks, this is the one that AI can't seem to do - I have to tell it how to do things and it's extra hard when I don't even have the terminology. I can describe how I want things to work, but the AI doesn't have the interiority of a human user - there's a lot of intuition and feelings involved - it does boneheaded things, ignores critical details that would be obvious to a human, doesn't realize which things it absolutely mustn't overlook on a UI and worst of all: even when I feed it a precise sketch of the interface I want, it ignores critical details because it simply didn't realize those details were key. (example: when I draw this thing next to this other thing, it literally means they need to be next to each other) If anyone wants a benchmark: I've been working on a game for about 6 weeks now - coincidentally I got started right before Opus 5.5 came out (when attempts with Astra just couldn't cut it). Granted, I have a day job so a lot of that was after-hours and weekends but I'd estimate that I have gotten done the equivalent work of a 10-person studio working for ~6 months. So there's definitely leverage here, but there's definitely a lot more work to go before this is a real game.
Some of my friends think AI will make it possible to build AAA games in just a couple of months. I think that seriously underestimates how complex game development is. I often point them to game designer Liz England’s “The Door Problem.” It’s a great example of how much thought goes into game design. Making a good game still takes hard work and experience. #gamedev #game
111
73
21
1,061
141,851
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
Delete everything in your AGENTS file and replace with: > Always build code cleanly, modularly, and with proper architecture and separation of concerns. Never simply bolt on a feature because I asked for it. If you can’t find an easy way to build it naturally, and you'd recommend a refactor to make the whole thing cleanly, stop your work ahead of time and say so.
157
134
27
3,772
285,516
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges. They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.
3,193
9,252
6,968
96,428
21,007,563
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
Background subagents in Claude Code now use less of your usage limit. A running subagent updates its status line every 30 seconds, and each update used to re-read its whole conversation. Now Haiku 5.5 writes it from short notes instead, so each update costs about 99% less.
714
980
744
23,469
2,605,497
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
Remember all the promises about AI? Almost every one of them was a lie. Here are the 11 biggest: 1. AI will create solo billionaire entrepreneurs. Lie. The smarter the models get, the less a product is worth, no matter how good the marketing is. Software is gradually becoming as ordinary as a cup of coffee from a coffee shop. And coffee shop owners don't become billionaires either. To get rich, you have to build something technically complex and unique. But AI models have strict filters that won't let you do any of that 2. People will make major scientific discoveries thanks to AI. Lie. AI labs will make scientific discoveries. For ordinary people, any serious scientific research is blocked by strict filters. Try discussing genetic engineering with Claude and you'll get your account banned right away 3. Engineers and programmers will be in higher demand as AI develops, because someone has to maintain it all. Lie. In two years, a software engineer will be on the same level as a barista selling coffee. The smarter the models get, the fewer people are needed to maintain systems. And this applies to more than just programmers 4. Thanks to AI, people will work less and rest more. Lie. If your employer used to expect five tasks a day from you, now they'll expect fifty. AI won't give you free time. It'll just raise the minimum bar for your productivity, especially when a large staff is no longer needed 5. AI will reduce economic inequality. Lie. The most powerful models require massive computing resources, capital, and infrastructure. The more important intelligence becomes to the economy, the more power goes to the handful of companies that control its production and distribution. We were promised democratization, and instead we got the groundwork for a new tech oligopoly 6. AI will let small companies compete with corporations. Lie. If one person with AI can build a product in a week, a corporation with a thousand employees and the same AI can build a thousand such products 7. AI will make us smarter. Lie. When people stop writing, analyzing, coding, and solving problems on their own, they risk gradually losing those skills. We're building the most powerful intellectual tool in history while gaining the option to completely stop training our own intelligence 8. AI will make technology accessible to everyone. Lie. You won't be given access to technology, but to rented intelligence. With limits, censorship, subscriptions, and rules that can change tomorrow. You control nothing 9. AI will give everyone a perfect education. Lie. We're already seeing serious degradation of human intelligence, of the education system, and of the need for a college degree. And nobody has yet figured out what to do with the hundreds of millions of students worldwide who, two years from now, will graduate with degrees nobody needs anymore. Intelligence has become free. Education is no longer needed 10. AI will kill advertising and pushy marketing. Lie. Advertisers used to buy a banner to influence your choice. Now they'll be able to pay your personal AI assistant, the one you trust more than your own wife. Advertising won't disappear. It'll become your best friend 11. AI will rid humanity of bureaucracy. Lie. One AI will draft the documents, a second will check them, a third will find violations, and a fourth will write appeals. Bureaucracy will finally be able to grow endlessly, without needing people at all
151
487
63
2,488
182,608
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
Hey, Rails peeps. Security updates have been released, this is a big ol' batch of fixes including another Critical in image processing. I strongly recommend you upgrade as soon as possible. rubyonrails.org/2026/11/4/Ra…
1
57
4
203
20,099
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
running a company nowadays feels kinda dystopian my team is 50+ people in 12 countries, but i talk to my agents more than my human teammates - voice note to grokbot - mattpocock/skills /grill-me so the agent interrogates the idea before anything gets built - spec drops into linear through the linear mcp - exa + firecrawl agent reads every competitor site, pricing page and changelog - ideabrowser.com to find ideas and trends - apify pulls g2, reddit and app store reviews - prompt: "find the 10 complaints that show up most and quote the users" - context7 so the agent reads the latest docs before it touches any api the design - claude code + paper mcp - 4 versions of every screen before a single line of code - pick pieces from 2, it merges them. diverge then converge - frontend-design skill from anthropic/skills + our own design system skill - cleanshot screenshots of any ui i like go straight in with "make it feel like this" the build - pingdotgg/t3code as the control surface for every agent, from my phone too - stablyai/orca runs a fleet of agents in parallel, each in its own git worktree - graphify maps the whole codebase into a knowledge graph so agents navigate instead of guess - headroom compresses tool outputs before they hit the model, so tokens go 20-95% further - lsp plugins so the agent sees types and errors like an ide does - security guidance hook flags risky edits before they land - supabase mcp for the database, stripe for billing - github mcp opens the pr, vercel spins up a preview - loops run overnight, i check what shipped over coffee the user testing - lightpanda as the headless browser, way lighter than chromium so you can run hundreds at once - agents click every flow on the preview like a first-time user - one persona is impatient, one is confused, one is on a phone - playwright + chrome devtools mcp for regression tests the bugs - sentry mcp catches the error - agent reads the stack trace and opens a fix pr in under 10 minutes - human just approves the data - posthog mcp cron at 7am on yesterday's funnel - any step that drops more than 10%, the agent watches 20 session replays and tells me why - stripe mcp pulls every cancellation reason into one doc every monday the agents' tools - orgo so every agent gets its own computer - agentmail so every agent gets its own inbox - agentphone so i can call them from the car - composio connect-apps for every other connector - akitaonrails/ai-memory, a git-backed markdown wiki so they remember what we decided last week the skills - wshobson/agents so skills work across claude code, codex and opencode - my writing voice - our design system - how we write specs - skill creator so the agents write new skills themselves some helpful tricks - one agent per platform. the x agent never touches email - cheap model for triage, opus for anything that ships - every output gets scored 1-10 by an llm judge. under 8 goes back - every agent writes a 5-line log at night so i can read the whole company in 2 minutes - every agent mistake becomes a new skill, so it only happens once the team - humans get version 5, not my voice note which they are happy about lol - their job is creative feedback, relationships and telling the agents what they got wrong a human pm reviews the funnel once a week. this reviews it every morning running a company is now voice notes and inference
577
333
153
5,854
745,494
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
Asked one of our developers how he uses his AI agents and he just sent me this:
179
1,337
270
16,196
1,778,676
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
Message threads are now in public beta on the Messages API. The API caches your conversation on the server between turns, so instead of resending the full conversation, you send only new messages and the previous message ID, and Claude starts responding sooner.
311
321
254
8,396
857,956
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
this whole "i'll build my own" wont last another 3 months for decades bros can't even keep their copyright footer text updated, imagine keeping 5 apps running never
113
39
22
1,108
43,822
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
Hard to overstate how big of a deal this is. tl;dr — ANY unpatched Linux server with even a single TCP port exposed to the internet (say for HTTP or SSH) and MPTCP active (default for many Go web servers) is vulnerable to complete takeover.
195
655
123
10,180
1,500,167
Matt “Friend of the pod” Rogish 🇺🇸 retweeted
Current agent pet peeve is how tainted the codebase gets with artifacts from the dialogue. > tell agent to do X > agent does X, but also does Y > tell agent I don't want Y > agent removes Y, but leaves comments everywhere like "X does x; don't worry, we don't do Y"
81
51
16
3,157
136,657