doubleAI

1/
What if an LLM could not just edit a model's weights, but re-architect it as it trains?

We tried just that. The resulting architectures exit the known territory of transformers: the coding agents spliced code directly into the network.

On nanochat, the loss fell below ordinary training after each new code layer was introduced. On algorithmic reasoning, accuracy rose from 55% to 84%. We even let DeepSeek-V4.1-Flash act as the agent modifying itself, raising its AIME and HMMT 2026 score from 64% to 71%.

To do all this, the agents built a set of tools we’ve never seen before, which we call mechanistic intervention!

Blog: doubleai.s.gy/llm-redesign

1/ Together with @ltx_io, we used Maxwell, our Code Execution Model, to make video generation in LTX-2.5 1.87x faster.

With Maxwell, generating a five-second 4K clip on 2xB200s went from 250 seconds to only 134. No quantization or loss of quality at all. The code is open source.

What if code could think while it runs?

Meet Maxwell: the world’s first Code Execution Model.

A new species of AI that lives inside the machine.

We believe it’s the key to unlocking 100% of compute.

Requests for early access are open, link below.

1/ What is the maximum number of legal moves in a chess position where every one of them gives check? The record on that question had stood since 1968. Double Agent, our AI research system, just broke it.

How many checks can you count here?

1/ Double Agent, our agentic system, has achieved a breakthrough result on one of the most central problems in extremal graph theory. It constructed explicit regular graphs with girth (8/5) log n, the first improvement in 38 years, and a leap towards the theoretical maximum.