emilr @ERoydev

the best time to understand how enabling modexp makes a lot of cryptographic primitives better was 2 years ago. the second best time is now. github.com/solana-foundat…

My average 3AM session with Claude..😄

We hired a junior developer to write the simple code, so we don't have to spend a ton of money on tokens for those basic/primitive tasks

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.

x.com/i/article/2044…

@ChainLinkGod for the retards like me

x.com/i/article/2046…

Today we revived Richie's old blog, bpf.wtf. We'll be pushing a few PRs to add the new syscalls and keep it up to date. To this day, it's still one of the best resources for learning sBPF Assembly. No more Wayback Machine! Compiler disrespecters rejoice!

🚨 CRITICAL: Active supply chain attack on axios -- one of npm's most depended-on packages. The latest [email protected] now pulls in [email protected], a package that did not exist before today. This is a live compromise. This is textbook supply chain installer malware. axios has 100M+ weekly downloads. Every npm install pulling the latest version is potentially compromised right now. Socket AI analysis confirms this is malware. plain-crypto-js is an obfuscated dropper/loader that: • Deobfuscates embedded payloads and operational strings at runtime • Dynamically loads fs, os, and execSync to evade static analysis • Executes decoded shell commands • Stages and copies payload files into OS temp and Windows ProgramData directories • Deletes and renames artifacts post-execution to destroy forensic evidence If you use axios, pin your version immediately and audit your lockfiles. Do not upgrade.

Our cryptographic researcher @alexand_belling revealed yesterday at @eth_proofs that Linea is moving to RISC-V. After 3 years of directly arithmetizing the EVM, producing a 1000+ page spec and one of the most rigorous proving system in production, we’re changing course. Here’s why 🧵

x.com/i/article/2033…

If your Rust code needs 6 layers of generics and 4 lifetimes, you’re not writing “advanced Rust.” You’re writing a future debugging nightmare.

@blueshift @solana Nice, thank you for the explanation. It adds much-needed context to the virtual-machine discussions, which are hard to follow without this background. Providing this kind of information really helps the community understand and contribute more effectively.

"I've never heard anyone describe code as Boroque."

Simple explanation of the Solana Market Structure limechain.tech/blog/making-se…

Using AI to solve a problem is one of the most miserable experiences you can ever have. Instead of going through the problem, figuring out what you don’t understand, learning it, and working your way through until it’s solved, which is interesting and satisfying. You just prompt the AI without understanding the underlying solution. You end up unable to properly explain the solution you want, mindlessly prompting while the AI can’t even get close to what you’re trying to achieve. And all of that, just for the sake of “10x productivity.”

@Python_Dv thats why i like twitter, if this was on facebook comments would be like "the answer is 15"

This is a masterpiece if you want to learn how the sBPF linker works. It pinpoints the potential problem that SVM has with the dynamic syscalls that are incompatible with the upstream eBPF. But it's normal after all, eBPF is created to fit Linux needs everything else has to adjust. blueshift.gg/research/sbpf-…

𝙻𝚒𝚗𝚞𝚜 𝚃𝚘𝚛𝚟𝚊𝚕𝚍𝚜 𝚊𝚏𝚝𝚎𝚛 𝚅𝙸𝙱𝙴 𝙲𝚘𝚍𝚒𝚗𝚐...