Ex-smart contract auditor (4+ years) advancing blockchain security via machine learning. Currently mastering linear algebra and matrix operations to build AI-na
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution
Glad you liked the smart contract vulnerability dataset.
In my open-source project LazyAuditor, I’m experimenting with compressing smart contracts by up to 60% without losing their meaning.
This can be especially useful for AI agents. Here’s how it works: ⤵️
Why does this matter?
The smaller the code, the fewer tokens the LLM needs to analyze it. This lets you fit more Solidity code into a single request and reduce audit costs
23 625 smart contract vulnerabilities — open sourced
I collected these findings over the past year from private audit reports and contest platforms
Hopefully, it’ll be useful for security researchers and people working on AI for smart contract security
https://t.co/SgY7J1crlr
The dataset is still in its raw form and needs some cleaning/normalization before it’s ready for ML training.
Also available on GitHub — a ⭐️ would be appreciated!
https://t.co/0Qzyt1YEAV
Fable 5.1 is now live in Claude Code and the Claude Platform.
It's priced the same as Fable 5, with 75% cheaper API cache reads. It gets a lot further into a long task before it needs your input, is better at telling you when it's stuck, and its writing style is more natural.
@OpenRouter has done more for the crypto community than most modern projects, simply by allowing their users to pay with crypto.
More platforms should enable crypto payments too. Tag companies in the comments that should offer crypto payments to their users
I'll start 🔻
@thsottiaux Yes, but that's only true for short, one-off tasks. GPT models are extremely lazy, and while they spend fewer tokens on a supposedly completed task, I end up having to explain to GPT over and over what I need and ensure the task is actually finished