Independent daily intelligence
Signal Archive
47 editions on record
- 047StoreBench: a live-commerce environment for autonomous operator agents24 stories →
- 046StoreBench: a live-commerce environment for autonomous operator agents19 stories →
- 045World Labs launches a public API for generating explorable 3D worlds22 stories →
- 044OpenTPU: an open-source AI accelerator20 stories →
- 043Beam: Reflection's 501B open-weight model24 stories →
- 042Seeing, Saying, but Not Using: From Reportable Spatial Facts to Usable States in Multimodal Large Language Models21 stories →
- 041Kolibri: A Sovereign Open-Weight Model19 stories →
- 040Announcing the World API22 stories →
- 039Clef: Open-weight decision models and a new RL fine-tuning platform22 stories →
- 038ArgGYM: A Procedural, Engine-Verified Benchmark for Structured Defeasible Reasoning23 stories →
- 037PSSA: A non-transformer language model written from scratch in Rust24 stories →
- 036Streaming 3DGS worlds on the web24 stories →
- 035The Price of Thought: Does Test-Time Reasoning Pay in LLM Trading?23 stories →
- 034Reverse-engineering the vintage Intel 8087 tangent algorithm: more than CORDIC23 stories →
- 033Thinking Leakage: A causal audit of NoThink post-training in hybrid reasoning models14 stories →
- 032AgenticCADedit: Stateful, tool-mediated multimodal CAD editing24 stories →
- 031COPE: continual LLM personalization from sparse feedback19 stories →
- 030The economics of open-weight inference21 stories →
- 029GVPO++: Group Variance Policy Optimization for LLM Post-Training and On-Policy Distillation18 stories →
- 028GVPO++: Group Variance Policy Optimization for LLM post-training and on-policy distillation20 stories →
- 027Saving another 100TB of RAM with math and Rust23 stories →
- 026Saving another 100TB of RAM with math and Rust23 stories →
- 025Training a 4B model to produce 81% faster query plans than Postgres24 stories →
- 024Announcing the World API22 stories →
- 023Distributed Systems Classics (2017)16 stories →
- 022Mapping the mind of a large language model23 stories →
- 021Anthropic maps concepts inside Claude Sonnet16 stories →
- 020Mapping the mind of a large language model24 stories →
- 019T1: Terminal Agent Reinforcement Learning for Long-Horizon Tasks24 stories →
- 018Streaming 3DGS worlds on the web17 stories →
- 017When memory helps tool-using LLM agents24 stories →
- 016Inside OpenAI’s research acceleration with coding agents22 stories →
- 015Extremely Sparse Supervision Incentivizes Reasoning Ability24 stories →
- 014Streaming 3DGS worlds on the web24 stories →
- 013World Labs launches a public World API17 stories →
- 012Speculative macro commit for faster tool-using agents18 stories →
- 011The efficient frontier of LLM inference18 stories →
- 010GPU World23 stories →
- 009How to build a diffusion language model20 stories →
- 008How to build a diffusion language model23 stories →
- 007GLM-5.3 is now open-weight21 stories →
- 006GLM-5.3 is now open-weight24 stories →
- 005GRAS: Guided Reduced-Variance Proposals and Adaptive Selection for Training-Free Reward Alignment in Discrete Diffusion24 stories →
- 004SHIFT-LLM: correcting distribution shift after depth pruning22 stories →
- 003Evaluating LLMs as calibrated causal-edge classifiers23 stories →
- 002Executable Is a SQLite Database23 stories →
- 001Qwen 3.8 27B is excellent, but it defaults to wildly overthinking things23 stories →