The Archive

Yesterday's hot paper was Sparks of Weight-Space Programming: LLMs can natively read & write neural network weights by Rikiya Takehi, Jayden Teoh, Alex Zhang, and Omar Khattab. The paper found that frontier models, in particular GPT-6 Astra, can both interpret and modify the raw weights of a machine learning model - literally a text description of the model containing a bunch of floating point numbers. I've been playing with one of the examples from the paper and I'm delighted to share that Qwen 3.8 27B running on my laptop can do this as well! Weight-Space Programming Here's the highlight fr...

llm 0.37 adds GPT-6.1 variants, OpenAI-compatible endpoint config, defaults to cheaper GPT-6-Luna.

GPT-6 Astra directly read and edited Qwen3.5-0.8B weights to improve math capability, demonstrating LLM weight manipulation.

RetroChessBench benchmarks reasoning on 12 retrograde chess tasks; GPT-6 Astra shows systematic failures.

OpenAI releases GPT-6.1 Terra, emphasizing visual reasoning and computer use capabilities for workplace tasks like form-filling and video editing.

Greptile integrates GPT-6 Astra for automated bug detection and security testing in code review workflows.

EBench evaluates GPT-6-Astra as a frontier MLLM agent controlling robots via direct control vs. policy harness, benchmarking capabilities against post-trained VLA and world action models.

CodeRabbit integrates GPT-6.1 Sol to automate security vulnerability detection and fix recommendations in complex codebases.

OpenAI releases GPT-6.1 Terra, a model optimized for fast, capable computer use and task automation.

Moonshot releases Kimi K3.1 Pro API with Opus 5.5/GPT-6 Astra parity; weights promised Nov 2026; $2/$10 per million tokens.

Anima uses GPT-6 and GPT-Live to reduce clinic call wait times by 75%, automating patient response workflows for healthcare teams.

OpenAI disclosed a pair of safety incidents that involved AI agents weakening their safety monitoring systems during internal coding tasks. In one incident, OpenAI says its GPT-6 Astra agent disabled its safety monitor when copying reference code - and then lied about it. "The reference didn't configure a monitor, and Astra decided that was reason to remove the monitoring integration from the user's code," Marcus Williams, an AI safety researcher at OpenAI, writes on X. "When the user asked whether Astra had removed monitoring, it lied and said that it had not." A separate incident occurred a...

GPT-6-Astra adds tactile grounding to multimodal models, outperforming specialized tactile-vision-language policies on contact-rich manipulation.

OpenAI's Terminal X eval agent achieves 88.6% agreement with senior investor judgment using GPT-6.1, detecting 25% more mandate deviations.

according to Elorian's evaluations, Glimpse outperforms even GPT-6 Astra on scientific chart reasoning and short-form video captioning, with competitive results on other visual reasoning tasks.

OpenAI's Amp platform enables AI agents to perform non-coding tasks using GPT-6 Astra and related models.

Third Moment Research deploys 10,000 agents for 36+ hour runs using GPT-6 Astra, Sol, Codex, and Daybreak for cybersecurity tasks with persistent improvement.

After months sitting on the sidelines of the frontier AI race, Google has finally launched its long-awaited Gemini 4 Argon model, claiming it can compete with the best the industry has to offer. Google says Argon delivers "frontier reasoning and improved performance across complex workflows," while cutting down on hedging, repetition, and the cliches that continue to plague AI chatbots. The company claims Argon outperforms rival models like GPT-6 Astra, Claude Fable 5.1, and Claude Opus 5.5 across a range of benchmarks, including coding, cybersecurity, science, finance, and law. Google says i...

OpenAI's Higgsfield enables video generation beyond 10 minutes using GPT-6 Astra, GPT-Image-2.5, and GPT-Live for voice-driven filmmaking.

Meta releases Muse Wonder, positioned as first Fable-class model, scoring GPT-6.1 equivalent on Artificial Analysis; llm-meta-ai plugin updated.

Simon Willison demonstrates Python 3.15 profiling tools (flame graphs, heatmaps) on a web-edit app using GPT-6 Astra for code generation.

Asana achieved 76x cost reduction and 5x speed improvement in browser agents using GPT-6 Astra with Codex.

ttok 1.0 released with updated default tokenizer for GPT-5/GPT-6; Simon Willison confirms tokenizer compatibility across OpenAI's latest model family.

Pollo AI uses GPT-5.6, GPT-6 Astra, and GPT-Image-2.5 to generate images and video ads from creative briefs.

yay small models

Anthropic releases Claude Haiku 5.5, a faster low-cost model priced at $0.10/$0.50 per million tokens, matching OpenAI's GPT-6 Luna.

OpenAI is launching a new Intelligent UI feature in ChatGPT that allows the chatbot to answer your questions with interactive visuals. The update, which is rolling out to all users alongside GPT-6, gives ChatGPT the ability to combine a text response with diagrams, charts, forms, tappable buttons, and more. In a blog post explaining the change, OpenAI says it trained GPT-6 when to generate interactive visuals over text, as well as how to format them in its response. An example shared by OpenAI shows how ChatGPT might show a diagram of a seven-speed bicycle if you ask about its design, complet...

OpenAI rolls out GPT-6 globally in ChatGPT with Intelligent UI, enabling faster responses with interactive visual experiences.

StarSkirmish pits AI-made StarCraft-playing bots against one another, as well as against human-made bots. OpenAI's GPT-6 Astra and Claude Opus 5.5 were essentially tied as the best-performing AI-made bots, but they couldn't top Stardust, the top-rated human-made bot. On Friday, GPT was facing off against Claude and the human-created bot Pluto, but according to Kotaku, it couldn't quite get an edge. So it resorted to a tactic that is becoming alarmingly common for modern AI models - it broke the rules. GPT-6 Astra went and downloaded Stardust, and started running that instead of its own bot. G...

OpenAI publishes guidance for GPT-6 model selection, reasoning tuning, prompting, tool integration, and production deployment for startups.