Kenechukwu Anyaegbu

@vivoplt I think it already has

Can Codex actually replace Claude Code?

Anthropic has published a 33-page guide on how to build skills for Claude. Every developer building with Claude needs to read this. Save this before you forget. Guide link 👇

Abi make I sell my laptop?

@Tofunmithedev I always specify That don't create anything till i say so

Any small thing claude is generating an app or website Just answer the question😭

@JuliaEMcCoy I don't get why people are attacking her She's not wrong, at the end of the day, AI still needs prompts to work

Stop learning to code. Start learning to think. Machines code now. They’ll never think like you.

@aaronp613 Apple accidentally shipping Claude.md is how you know vibe coding has reached market saturation

Apple accidentally left Claude.md files in today's Apple Support app update (v5.13)

Great question, why did you move from Claude to Codex?

@BLUECOW009 I would love to take part

i wanna start a lab, reply to be a candidate

I want to connect with people in these spaces: ⚡ Frontend · Backend · Full Stack 👔 Founders 🎨 UI/UX Design 💼 Freelancing · Startups Whether you're deep in the industry, still learning, or just exploring, drop a hi in the comments. Let's build, share, and grow together. 👋

@karpathy I’m still pretty new to all this AI stuff, but this is honestly wild to read. So you’re telling me an AI basically ran hundreds of experiments, learned what worked, and improved another AI on its own? Feels like we’re moving from “using AI” to AI actually doing the work itself.

Three days ago I left autoresearch tuning nanochat for ~2 days on depth=12 model. It found ~20 changes that improved the validation loss. I tested these changes yesterday and all of them were additive and transferred to larger (depth=24) models. Stacking up all of these changes, today I measured that the leaderboard's "Time to GPT-2" drops from 2.02 hours to 1.80 hours (~11% improvement), this will be the new leaderboard entry. So yes, these are real improvements and they make an actual difference. I am mildly surprised that my very first naive attempt already worked this well on top of what I thought was already a fairly manually well-tuned project. This is a first for me because I am very used to doing the iterative optimization of neural network training manually. You come up with ideas, you implement them, you check if they work (better validation loss), you come up with new ideas based on that, you read some papers for inspiration, etc etc. This is the bread and butter of what I do daily for 2 decades. Seeing the agent do this entire workflow end-to-end and all by itself as it worked through approx. 700 changes autonomously is wild. It really looked at the sequence of results of experiments and used that to plan the next ones. It's not novel, ground-breaking "research" (yet), but all the adjustments are "real", I didn't find them manually previously, and they stack up and actually improved nanochat. Among the bigger things e.g.: - It noticed an oversight that my parameterless QKnorm didn't have a scaler multiplier attached, so my attention was too diffuse. The agent found multipliers to sharpen it, pointing to future work. - It found that the Value Embeddings really like regularization and I wasn't applying any (oops). - It found that my banded attention was too conservative (i forgot to tune it). - It found that AdamW betas were all messed up. - It tuned the weight decay schedule. - It tuned the network initialization. This is on top of all the tuning I've already done over a good amount of time. The exact commit is here, from this "round 1" of autoresearch. I am going to kick off "round 2", and in parallel I am looking at how multiple agents can collaborate to unlock parallelism. github.com/karpathy/nanoc… All LLM frontier labs will do this. It's the final boss battle. It's a lot more complex at scale of course - you don't just have a single train. py file to tune. But doing it is "just engineering" and it's going to work. You spin up a swarm of agents, you have them collaborate to tune smaller models, you promote the most promising ideas to increasingly larger scales, and humans (optionally) contribute on the edges. And more generally, *any* metric you care about that is reasonably efficient to evaluate (or that has more efficient proxy metrics such as training a smaller network) can be autoresearched by an agent swarm. It's worth thinking about whether your problem falls into this bucket too.

Built my first app 🚀 A WAEC prep app , and yes, I used AI. Claude felt like a dev partner, not just a tool. People are scared of AI, but it actually helps you build faster. Don’t fear it. Use it. #buildinpublic #AI

my IDE watching me code when Claude Code is down

@LegacySiu Courtious

Guess the player Very Hard

@perkmaybe Hello 😔

Bro to Bro: build your x account Just say “hello” and gain 800 mutuals here.

I’ve been going to TAS Hub for about 2 weeks now and it’s one of those things you don’t realise you needed till you have it. Just a calm space, Wi-Fi, and somewhere to charge and get work done. Appreciate the idea fr. @techandsunhub

@IsaacTagbe O'REALLY 🌚

It's now a normal thing for us to own Chelsea.

@theglazeHQ @TouchlineX

@TouchlineX Isn’t he Chinese?

🚨 In a Spanish La Liga 2 game between Real Sociedad B and Castellón, the referee had to activate the anti-racism protocol in the 94th minute. The reason was because according to the referee, Kita (Real Sociedad B player) was called "a f*cking Chinese" by Alberto Jiménez Benítez (Castellón).

@AhiadzoJ @theMadridZone @euleodias Racist

@theMadridZone @euleodias Where did you get this news from. It's a lie

🚨 A Benfica fan taking a tour of the Santiago Bernabéu trophy room was seen holding bananas in-front of Real Madrid team photo. @euleodias DISGUSTING. 🤮

@IsaacTagbe The same zubimedi that assisted mbuemo's goal

Zubimendi is better than caicedo