Essays you can run.

Long-form, interactive notes from building agentic systems in production. Every diagram is live code: drag it, break it, crash it.

• E.01

  Durable agents with Effect TS

  Typed failures, retries, timeouts, fallbacks and crash-proof steps for agents that never silently die.

  Effect v4Agents

  September 30, 2026 · 14 min
• E.02

  Streaming generative UI on TanStack Start

  Server functions, AI SDK streams and components that render while the model thinks.

  AI SDKTanStack

  September 30, 2026 · 11 min
• E.03

  Traces, not vibes: observability for LLM pipelines

  Spans, token budgets and evals wired into the same telemetry as the rest of your stack.

  OpenTelemetryEvals

  September 30, 2026 · 13 min
• E.04

  Let it crash: what Elixir supervisors teach AI agents

  Supervision trees as a mental model for fault-tolerant, multi-agent systems.

  ElixirArchitecture

  September 30, 2026 · 12 min