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 7, 2026
You can't run an agent you can't watch
A Cisco survey this year found most companies are running agents they can't properly monitor. That's the whole problem in one sentence. Agents fail in a way regular software doesn't — they return a tidy success while quietly doing the wrong thing, and you only see it in the full trace of what they did, not the final output. 'Agent observability' became its own discipline in 2026 for exactly that reason. The unglamorous ability to watch what your agent actually did is turning into the line between a pilot and production.
- methodology
- agents
- architecture
June 6, 2026
Supabase is worth $10.5B because agents need boring databases
Supabase just raised $500M at a $10.5 billion valuation — doubled in eight months — and the reason is almost funny: over 60% of the new databases on its platform are now created by an AI tool, not a human. The flashy part of the AI boom is the agents writing the code. The part that's quietly minting money is the boring, reliable place that code has to put its data. That's not a coincidence — it's the whole lesson about where durable value lives.
- business
- architecture
- ai-native
June 6, 2026
Microsoft can fire its model supplier. Can you?
At Build 2026 Microsoft shipped its own coding and reasoning models — trained from scratch, with what its AI chief called 'zero distillation' from OpenAI — straight into GitHub Copilot. The richest software company on earth just spent billions to stop depending on one supplier. That's the whole lesson for the rest of us, and it costs you nothing: never let the model be the part of your system you can't swap out.
- architecture
- ai-native
- business
June 6, 2026
The best agent of the year runs on a factory floor
While everyone argued about chatbots, Foxconn quietly wired hundreds of AI agents into its production lines — reading sensors, equipment, and ERP data — and reported 80% faster root-cause analysis and 10% fewer machine failures. Nobody made it a viral demo. That's the tell. The agent deployments that actually work this year are narrow, plugged into real ground truth, and measured against a hard number. The exciting ones are still stuck in a pilot.
- agents
- architecture
- business
June 6, 2026
Three frontier models landed this month. I changed one line.
June 2026 is dropping three new frontier models in a single month — GPT-5.6, Gemini 3.5 Pro, and Claude Mythos — on top of two that shipped in May. For most teams that's a stressful treadmill of 'should we migrate?' For one that built the boring way, it's three free upgrades, each a one-line config change. The relentless model churn isn't a problem to keep up with. It's a gift you collect automatically — but only if you built the seam to catch it.
- ai-native
- architecture
- agents
June 6, 2026
12,520 agent tools are on the internet with no lock on the door
Security researchers scanned the internet and found 12,520 MCP servers — the tool connectors that let AI agents do real things — sitting in the open, most with no authentication at all. Some let any stranger run database queries or shell commands. The protocol just made auth mandatory in its June spec, but the damage shows the deeper bug: the default was 'no lock,' and people shipped the default. If your agent can reach a tool, so can everyone else — unless you decided otherwise on purpose.
- security
- architecture
- agents