The dominant scaling narrative in large language models has been straightforward: more parameters, more data, more compute. But there’s a
Continue readingLoopCoder-v2: Why Two Loops Beat Four in Test-Time Compute Scaling
Editor’s note, September 2026: GLM-5.2 is no longer the newest model in the family — GLM-5.3 shipped in August 2026
Continue readingGLM-5.2: The New #1 Open-Weight LLM and Why IndexShare Matters
The attention mechanism is the backbone of every transformer model, but it carries a brutal cost: quadratic complexity with respect
Continue readingHow MiniMax Sparse Attention Achieves 28x Compute Reduction at 1M Context Length
The GitHub trending page this week is dominated by AI agent tooling, but tucked between the skills and plugins are
Microsoft’s Build 2026 conference delivered a move that had been anticipated for months but still landed with weight: the company
AI agents have a skill problem. You give a language model a system prompt — or “skill” — and it
Continue readingSkillOpt: Training AI Agent Skills Like Neural Networks
Editor’s note (September 12, 2026): Four months on, this piece holds up well — both studies remain the reference points
Continue readingConstraint Decay: Why Your AI Agent Forgets the Rules (and What to Do About It)
Editor’s note (September 12, 2026): This is a May 2026 snapshot of a fast-moving ecosystem, and the trajectory since only
Editor’s note, September 2026: Written at launch in May 2026, when pricing hadn’t been published — that has since changed:
Continue readingQwen3.7-Max: Built for the Agent Era, Not the Chat Era
The first version of any RAG pipeline usually looks the same: embed a query, search a vector store, stuff the
Continue readingBeyond Naive RAG: 4 Advanced Patterns That Actually Work in Production