Scar 🇺🇸

# GENESIS ```word LET IT BE WRITTEN in word ```

@LLMJunky @TheAhmadOsman @0xSero @ivanfioravanti @ZenMagnets @NVIDIAAI Congrat! Have fun :)

@realBionicLife @TheAhmadOsman @0xSero @ivanfioravanti @ZenMagnets @NVIDIAAI 7700 shipped 🚢

Finally proud to announce that I've joined the GPU Minor Leagues. 2 x RTX 6000 Pro. I have six months to pay off the second GPU lol. You are all TERRIBLE influences.

@zongheng_yang @karpathy It is yet to be proven that a winner of short train run is also the winner of a real long train run.

Karpathy @karpathy is the GOAT of simple but viral projects. Autoresearch is his latest: an agent running guided search in a loop. What if the agent is scaled up? We give an Autoresearch agent 16x H100 + H200s and let it rip. Result: 9x faster time-to-quality & better quality. Most important — all GPU cluster shenanigans are automated by a skill. The skill teaches the agent to launch parallel GPU jobs via SkyPilot. This means to get SOTA model results, 𝗮𝗹𝗹 𝗮 𝗿𝗲𝘀𝗲𝗮𝗿𝗰𝗵𝗲𝗿 𝗵𝗮𝘀 𝘁𝗼 𝗱𝗼 𝗶𝘀 𝗮𝗻 𝗘𝗻𝗴𝗹𝗶𝘀𝗵 𝗽𝗿𝗼𝗺𝗽𝘁 + 𝗮 𝗸𝘂𝗯𝗲𝗰𝗼𝗻𝗳𝗶𝗴 𝗳𝗶𝗹𝗲.

@karpathy Nice!

Thank you Jensen and NVIDIA! She’s a real beauty! I was told I’d be getting a secret gift, with a hint that it requires 20 amps. (So I knew it had to be good). She’ll make for a beautiful, spacious home for my Dobby the House Elf claw, among lots of other tinkering, thank you!!

@csaba_kissi How does your hand written css slop scale?

Everybody is talking about how AI will generate 100% of the code in a year or two. But nobody is talking about how this code will scale.

@ah20im Warrning before a context is compacted. Estimate the work ahead. Warn that there is not enough context to get it done right. Ask: Perpare a hand off to a new thread? Yes | No.

What would you like to see in Codex?

@thsottiaux How can they work on the same code base without conflicting with each other?

Hello subagents in codex. Have seen some awesome new and creative workflows emerge from these developers.openai.com/codex/subagent…

@dev_maims Matter even more

Do you think understanding the fundamentals still matters when AI can generate working code in seconds?

@EricRWeinstein Why is everythinng from you sound like a made up DRAKE equation. :)

When you think about the AI revolution WHO do you think about? Value X Created V{X)=Ln(-————-———) Value X Captured People with high V(X) are VERY hard to control. People with negative V(X) are much easier. Try listening to high V(X).

@Yuchenj_UW x.com/realBionicLife…

Markdown is code. English is the new default programming language. If this sounds controversial, you’re still living in 2025.

@Ai_Vaidehi It's like certified Google Search User of 1998?

Anthropic just announced the "Claude Certified Architect" program. And you can start today. In 16 years of my professional career, I haven't done a single certification. Not one. Not AWS. Not Azure. Not Google Cloud. Not PMP. Not Scrum. Not any of the alphabet soup. I learned by building. By breaking things. By shipping. But I'm about to break that streak. I'm going for my first-ever certification: Claude Certified Architect — Foundations Here's why this matters — especially if you're a developer, engineer, or any professional who feels like the AI wave is moving too fast. Claude Code launched a few weeks ago. And it feels like a paradigm shift. Not an incremental upgrade. Not another chatbot wrapper. A fundamentally different way of building software. Agentic architecture. Tool orchestration. MCP integration. Context management at a systems level. If those words sound intimidating — that's exactly why this certification exists. It covers everything from agentic orchestration to prompt engineering to Claude Code workflows. Not surface-level content. And here's what got me: It costs nothing. Free. Zero. $0. So if you've been feeling left behind... If you've been watching others ship AI agents while you're still figuring out where to start... If you've been telling yourself "I'll learn this next quarter"... This is your sign. Stop scrolling. Start building. First certification in 16 years. Let's see how this goes. Links in the comments 👇 Cc : Brij Pandey

@paulabartabajo_ Yes. So many changes. We can reshape codebases in one sitting now. We can allow wet code now without introducing aburd amount of layers and architectural acrobatics. So many changes.

You can refactor a codebase in an afternoon now. The hard part is knowing what it should look like when you're done

We need to rethink how we work on the shared codebases. We used to be able to split domains and work in parallel without colliding. Now we operate not on the domain level but entire codebase level. Refactring, architectural changes, data layer changes move fast. We move reshape entire codebases now in a span of a few hours. Working in parallel becomes impossible. Definition of wet code changes. So many changes....

Meta laying off 20% after Block cut 50% is just the beginning. More companies will realize: cut 50% of the org, arm the remaining 50% with AI, and you get higher productivity at lower cost. In big companies, hierarchy kills velocity. Even if AI makes engineers 3x faster, most of their time still goes to meetings and alignment. The AI era favors small, ruthless, fast human teams backed by armies of AI agents.

@m0rcutt @aidenybai The mess it write it will correct if you design a test tool it can use.

@aidenybai 100%. The job has basically become let the coding agent write a mess and then diligently cleanup after it

the bottleneck for coding agents is now testing / code quality agents are OK at writing code in the happy path, but don't consider edge cases on harder tasks it's also very messy (unnecessary utils, duplicated code, random `as` everywhere)

Software development is dead, long live software development!

Can't say AI won't take your job anymore. Entry level just disappeared.

@paulabartabajo_ A little more but yeah. A good orchestration layer and narrator make all the difference for the experience

Advice for AI engineers 💡 The best AI agents aren't built with complex frameworks. They're a while loop, an LLM, and tools. That's it. Claude Code, Codex, OpenClaw... All the same pattern.

!!!!! github.com/karpathy/jobs !!!! We're all done.

5 minutes ago, @karpathy just dropped karpathy/jobs! he scraped every job in the US economy (342 occupations from BLS), scored each one's AI exposure 0-10 using an LLM, and visualized it as a treemap. if your whole job happens on a screen you're cooked. average score across all jobs is 5.3/10. software devs: 8-9. roofers: 0-1. medical transcriptionists: 10/10 💀 karpathy.ai/jobs

@Yuchenj_UW There is a meme for that?

--dangerously-skip-permissions is how I run Claude Code also how I run my life