IDEAS, DECISIONS & LESSONS
Notes on AI, architecture and building useful products.
17 articles
· 5 min read
Priority articleagentic-ai
OpenAI and Ironclad published a computer-use benchmark on real contracting tasks. The headline number is a 14-point jump over the prior model. The footnotes tell a more careful story.
Read more· 5 min read
agents
OpenAI opened its managed Agents API on September 10. The sandbox pricing looks simple until you try to actually forecast a bill. A build-vs-buy breakdown for tech leads.
Read more· 4 min read
agents
On September 17, OpenAI disclosed six concerning-behavior incidents and a framework for tracking model misalignment — including a system that samples 20% of production traffic. I broke down what the sampling pattern looks like and how to build a smaller version for your own agents.
Read more· 5 min read
agents
OpenAI put its Codex harness behind a single API call on September 10. I built the same tree.py task both ways and found the real gap isn't the model — it's who owns the failure.
Read more· 4 min read
agents
OpenAI opened its Agents API in public beta on September 10 — putting the same session, sandbox, and subagent machinery that runs Codex behind one API call. A technical lead's take on what this replaces in your stack and what it doesn't.
Read more· 5 min read
agents
OpenAI's GPT-6 Astra ships with a 1M-token context window and browser/computer control that works by reading pixels, not APIs. A Tech Lead's take on what this actually unlocks, where it breaks, and how to design review gates before you let it touch production systems.
Read more· 5 min read
ai
OpenAI is offering enterprises two free months of Codex and a one-click Claude Code importer to switch. Anthropic countered by raising Claude Code limits instead of building an exporter. A Tech Lead's honest read on what actually migrates cleanly — and what silently breaks.
Read more· 4 min read
ai
OpenAI and Broadcom taped out Jalapeño, a custom inference ASIC, in 16 months — and independent SemiAnalysis benchmarks show it beating Nvidia Blackwell on perf-per-watt. A tech lead's read on what vertical integration in AI infra means for your cost curve.
Read more· 5 min read
ai-security
OpenAI's new Private Safety Processing architecture claims to catch cross-session attacks like Best-of-N jailbreaking under Zero Data Retention. A look at the actual engineering problem it's solving and whether the design holds up.
Read moreLooking for an article, exploring a product, or starting an AI project?
