←
Zaevlad1,435 posts
Zaevlad avatar

Zaevlad

@RightNowIn

Ex-smart contract auditor (4+ years) advancing blockchain security via machine learning. Currently mastering linear algebra and matrix operations to build AI-na

◷ Joined December 2009▧ 312 Photos and videos
Zaevlad avatar
Zaevlad@RightNowIn·Sep 17

I guess Jev can also be very useful in web3 security agents, as well as onchain guards!

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

2:56
Zaevlad avatar
Zaevlad@RightNowIn·Sep 8

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

Zaevlad avatar
Zaevlad@RightNowIn·Sep 8

After the LLM, there's another compression step: the program automatically shortens variable names and removes unnecessary spaces and characters. The result is a compact version of the same contract

Zaevlad avatar
Zaevlad@RightNowIn·Sep 8

For example, depositPaused can become v1, while several lines of checks can be reduced to a single short construction. The meaning and logic of the contract remain unchanged

Zaevlad avatar
Zaevlad@RightNowIn·Sep 8

In LazyAuditor, the smart contract is first compressed using an LLM. The model removes unnecessary parts, shortens constructions, and renames long variables while preserving the contract's logic and behavior

Zaevlad avatar
Zaevlad@RightNowIn·Sep 8

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: ⤵️

Zaevlad avatar
Zaevlad@RightNowIn·Sep 1

It comes with limits reset!

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.

Zaevlad avatar
Zaevlad@RightNowIn·Jul 17

2. Users are trying to integrate Sol into Claude Code CLI because it's still better than Codex for developers. 3. Users also create workflows where Fable orchestrates Sol, but not vice versa. 4. Moreover GPT models still super lazy to follow long instructions.

Zaevlad avatar
Zaevlad@RightNowIn·Jul 17

The main problem with GPT models is that the team focuses more on creating temporary hype in X than on making its models truly excellent. 1. Even after the latest release, it could not beat Claude in any category on LMARENA.

Zaevlad avatar
Zaevlad@RightNowIn·May 25

Also added a Monaco-based editor integrated with the agent. Responses containing path:line references are clickable, opening the exact workspace file and scrolling directly to the referenced code location.

Zaevlad avatar
Zaevlad@RightNowIn·May 25

Visualization support was heavily expanded. The agent can generate interactive HTML/SVG widgets inside sandboxed iframes, attach follow-up prompts and build dynamic visualizations directly inside the chat interface.

Zaevlad avatar
Zaevlad@RightNowIn·May 25

Pipeline mutations are fully integrated into the agent workflow. You can ask the assistant to create agents, connect nodes or restructure flows and every canvas action is streamed live into the UI with audit log and undo support.

Post attachment