IDEAS, DECISIONS & LESSONS
Notes on AI, architecture and building useful products.
85 articles
13 Sep 2026 · 5 min read
agentic coding
Claude Code v2.1.269 adds claude plugin eval — six grader types, a no-plugin baseline, and CI flags like --threshold and --max-cost-usd. A Tech Lead's walkthrough of what it fixes and how to wire it into a release gate.
Read more11 Sep 2026 · 7 min read
agentic coding
Backend, parent dashboard, ops console and companion app built simultaneously against one written API contract -- and everything I re-verified rather than trusted.
Read more09 Sep 2026 · 5 min read
ai
HydraFusion picks a per-task execution plan across model providers — cascade, critique, or single-shot — and reportedly beats Claude Opus 5 quality on TerminalBench 2.1 at 36-67% lower cost. A Tech Lead's read on whether this replaces your model-routing logic or just moves it.
Read more09 Sep 2026 · 4 min read
ai
The PyTorch Foundation added Alibaba Cloud and Cambricon as Platinum members (with governing board seats) and Ant Group as Gold at PyTorch Conference China. A Tech Lead's read on what non-CUDA silicon getting a direct vote in PyTorch's roadmap means for anyone betting on the framework's neutrality.
Read more09 Sep 2026 · 5 min read
agentic-ai
Anthropic's September Claude Code update quietly raised bashOutputMaxChars and taskOutputMaxChars to 128K and added a file-based subagent system prompt flag. A Tech Lead's read on why this unglamorous change matters more than most model announcements for anyone running fan-out subagent pipelines.
Read more09 Sep 2026 · 5 min read
ai
Anthropic's Claude produced a 13M-line, fully machine-verified Lean proof of Fermat's Last Theorem in 11 days via the open-source Prove2Me orchestrator. A Tech Lead's read on what the multi-agent coordination pattern behind it means for any team building long-horizon autonomous agents.
Read more07 Sep 2026 · 6 min read
ai-agents
A single Git performance setting, core.fsmonitor, lets a malicious repository run arbitrary code the moment an AI coding agent looks at it — before you've typed a prompt or clicked 'trust this folder.' A Tech Lead's breakdown of the exploit and the fix.
Read more07 Sep 2026 · 5 min read
ai
Nemotron 3 Ultra pairs Mamba state-space layers with transformer attention and a mixture-of-experts router to hit a 1M-token context at 5x the inference throughput of comparable dense models. A Tech Lead's read on the architecture and where it changes agent deployment math.
Read more07 Sep 2026 · 5 min read
ai
NVIDIA and Hugging Face published Open-SWE-Traces, SWE-Zero, and SWE-Hero — half a million open agent trajectories with tool calls, patches, and failure cases. A Tech Lead's guide to reusing this format for internal agent evals, not just model fine-tuning.
Read more
