Translation

SWE-1.7 Reach Near GPT 5.5 and Opus Intelligence

Cognition has released SWE-1.7, a new version of its AI coding agent that they claim approaches near GPT-5.5 and Opus levels of intelligence, marking a significant advancement in autonomous software engineering capabilities.

Background

- Cognition Labs is a US AI startup founded in 2023 by ex-Palantir engineers, known for building Devin, an AI coding agent that can autonomously write code, fix bugs, and manage GitHub pull requests. - SWE-1.7 is the latest version of the company's internal AI model. "SWE" stands for software engineering. - "GPT 5.5" and "Opus" refer to rumored or coded names for future large language models from OpenAI and Anthropic (Claude). The claim is that SWE-1.7 performs at a level close to the next generation of frontier models from those companies. - SWE-bench is a widely used benchmark that tests whether an AI can autonomously resolve real GitHub software bugs. A high score suggests a model can perform practical coding tasks, not just answer theoretical questions. - The blog post positions Cognition's model as competitive with the biggest AI labs, suggesting a smaller startup may have closed the gap in specialized coding ability.

RT Max Weinbach: I had early access to this model and I'll post more Thursday, but I think the most impressive part is this model NEVER GIVES UP If yo...
9.0
Max Weinbach says he had early access to OpenAI's new model GPT-5.6 Sol, calling it his favorite model by far. He highlights that it never gives up and will keep reasoning until it's done. OpenAI announced that GPT-5.6 Sol, along with Terra and Luna, will launch publicly on Thursday, with preview access expanding globally now.
Better Models: Worse Tools
7.5
Newer Claude models sometimes invent extra keys in tool call arguments, breaking validation in Pi's edit tool. The author suspects post-training for Claude Code's forgiving harness makes alternative schemas fail. This suggests closed RL training can degrade general tool-use reliability.
Writing an LLM from scratch, part 34b -- from bigrams to GPT-2, one component at a time (in JAX)
3.0
The author builds and trains a GPT-2 small model from scratch in JAX, starting from a basic bigram-style model and incrementally adding components like LayerNorm and Transformer blocks. Achieved a final loss of 3.418, beating their PyTorch version (3.538) and original GPT-2 small (3.499) on the same test dataset.
Flax debugging: making a hash of things
2.0
The author debugged a Flax NNX training loop where the loss was stuck at 10.82, indicating random guessing. By hashing the model parameters and comparing hashes across steps, they discovered the parameters weren't changing. The root cause was using @jax.jit instead of @nnx.jit, which is needed for proper in-place state propagation of parameter updates in NNX.
Writing an LLM from scratch, part 34a -- building a JAX training loop for an LLM training run
1.0
Building a JAX training loop for an LLM from scratch using Flax NNX and Optax. First validated the harness with a minimal "A-to-A" model (embedding then projection) before adding Transformer layers. Key challenges included fixing JAX's GPU memory default and slow data iteration by committing data to CPU memory.

SWE-1.7 Reach Near GPT 5.5 and Opus Intelligence

Background

Related stories

RT Max Weinbach: I had early access to this model and I'll post more Thursday, but I think the most impressive part is this model NEVER GIVES UP If yo...

Better Models: Worse Tools

Writing an LLM from scratch, part 34b -- from bigrams to GPT-2, one component at a time (in JAX)

Flax debugging: making a hash of things

Writing an LLM from scratch, part 34a -- building a JAX training loop for an LLM training run