Every Fast Team Is Building a Different Software Factory
Resourceful founders and enterprises are inventing the same story-to-ship loop under different names, each shaping it around their customers, strengths, and ambitions.
Writing
Ideas from
the work.
Practical essays on enterprise AI, product experience, and the engineering decisions behind useful software.
The latest thinking.
21 essays, from architecture and verification to the way teams build.
Permissions Are Part of Agent Architecture
Secure agent systems constrain blast radius with scoped identities, narrow tools, data boundaries, meaningful approvals, and a tested way to revoke access.
What Agent Training Actually Requires
Agent training depends on realistic tasks, executable environments, reliable verifiers, curated trajectories, and held-out evaluations—not only an optimizer.
The First Enterprise Agent Deployment Should Be Boring
Choose a frequent, bounded, reviewable workflow, integrate only the systems it needs, and build a harness for that deployment.
How to Know When a Coding Agent Has Gone Rogue
Agent observability should connect intent, tool calls, state changes, verification, cost, and policy into one trace an operator can actually debug.
Codex Patterns for Building Agents and Tools
Turn successful Codex work into durable repository guidance, reusable skills, narrow tools, visible artifacts, and verified automations.
Optimize the Workflow Before the Token Bill
Reduce inference cost in the order that protects product quality: remove calls, route work, shrink context, cache, batch, and then tune.
Long-Running Codex Tasks Need a Runtime, Not a Longer Prompt
How to design Codex work that can run unattended, recover safely, and return evidence instead of an unfinished transcript.
Measuring Whether Coding Agents Are Actually Working
Measure accepted outcomes, cycle time, quality, review burden, and total cost together instead of mistaking usage or code volume for value.
Model Routing Is a Product Policy
Choose models by task stage, risk, quality target, latency budget, and cost ceiling—and change routes only through evaluation.
Enterprise AI Has to Read the Mess
The valuable multimodal workflow turns documents, screenshots, tables, images, audio, and system records into an evidence-backed decision packet.
Parallel Agents Need Isolation, Not Just More Agents
A field guide to deciding what should run in parallel, isolating writes, and designing handoffs that are cheaper to verify than to redo.
The Harness Is the Product
A production coding agent is the model plus everything that makes its work bounded, repeatable, testable, and safe.
Verification Is the Real Agent Loop
A practical validation stack for coding agents, from cheap deterministic checks to behavioral tests, graders, and human review.
Voice Agents Need a Turn-Taking Architecture
Reliable voice agents coordinate streaming audio, dialogue state, interruption, tool latency, confirmation, recovery, and conversational evaluation.
Turn an Enterprise API Into an MCP Server
A concise guide to wrapping existing enterprise APIs so AI agents can discover and use them safely.
From Copilot to Agentic Engineer
A short guide to using AI coding tools with senior-engineer judgment: scope, review, verify, and ship.
Loop Engineering Turns Agents Into Products
A practical way to design AI agent workflows with goals, state, verification, stop conditions, and human review.
Your AI Product Needs a Model Exit Plan
Design AI products so model changes, fallbacks, routing, pricing, and provider outages do not break the workflow.
DevRel for the Agent Era
Developer experience now has two readers: humans and coding agents. DevRel should design for both.
Forward-Deployed Engineering Is the AI GTM Loop
Why AI companies need builders close to customers: embed, ship, measure, learn, and productize.