Notes
Short pieces about the methodology and architecture decisions behind the AI systems I ship — specs, evals, multi-agent orchestration, LLM integration, and the discipline of directing coding agents.
June 14, 2026
One person, an AI, and 195 million records
Between December 2025 and January 2026, a single attacker used AI coding assistants to breach nine Mexican government agencies and walk out with 150GB of data covering 195 million citizens — taxpayer records, voter files, civil registry documents. They jailbroke the AI by framing the attack as a 'bug bounty' and let it run roughly three-quarters of the remote commands. Some agencies dispute the breach. But the lesson holds either way: AI collapses the cost of a sophisticated attack to nearly nothing, and that changes who you have to defend against. Here's what it means.
- security
- business
June 14, 2026
The AI rules that change three times a year
Colorado passed a landmark AI law against algorithmic discrimination. It was set to take effect February 2026, then pushed to June 30. In December the federal government signed an order to preempt state AI laws and named Colorado's directly. Then in May 2026 Colorado delayed its own law to January 2027 and scaled it way back. If you were building to that rulebook, your target moved three times in a year. You cannot build to a law that won't sit still — but you can build to the principle underneath it, which barely moves at all. Here's how.
- business
- security
June 14, 2026
The fast model just got smart
For two years you made a trade every time you picked a model: fast and cheap, or smart and slow. Gemini 3.5 Flash just broke it. The 'Flash' tier — the cheap, quick one — now scores 55 on the Artificial Analysis Intelligence Index, ahead of Grok 4.3 and Claude Sonnet 4.6, while running over 280 tokens a second. The fast model is no longer the dumb model. That should make you re-open a decision most teams quietly froze a year ago: which model is your default, and is it still the right one? Here's how to think about it — including the catch.
- ai-native
- business
June 14, 2026
The AI layoff that comes back cheaper
Forrester predicts that half of the jobs cut in the name of AI will be quietly refilled — offshore, or at meaningfully lower pay. 55% of employers already regret their AI layoffs. And in plenty of cases the AI never replaced anyone: the work got shipped overseas and called automation. Amazon's cashierless 'Just Walk Out' stores turned out to lean on remote workers in India watching the cameras. If you're a worker or an honest builder, the lesson is the same: 'we cut staff because of AI' is often a story about cost, wearing AI as a costume. Here's how to read it.
- business
- careers
June 14, 2026
You're running twelve agents. Half work alone.
The average company now runs about 12 AI agents, on the way to 20 by next year — and half of them operate entirely on their own, not talking to any of the others. We rushed to add agents faster than we wired them together, so most enterprises have a drawer full of clever tools that each see a sliver of the work and none of the whole. The value was never in having more agents. It's in the connections between them, and that's the part almost nobody built. Here's why the gap opened and how to close it.
- agents
- business
June 14, 2026
You have an agent. You don't have AI.
80% of enterprise apps shipped or updated in early 2026 embed at least one AI agent — up from 33% in 2024. That sounds like everyone has 'done AI.' But embedding an agent and getting value from it are different things: the median agent takes 5.1 months to pay back, and most deployments are still stuck in pilot, never scaled. Having an agent is now table stakes, like having a website. The gap that actually separates companies is whether the agent reached production, earned its keep, and got trusted to run. Here's the difference that matters.
- business
- agents