Blog
- Parallel Coding Agents and Stacked PRs: What Breaks When You Merge
November 14, 2026 · Splitting a feature across parallel agents is the easy part. Getting their branches back into main is where it goes wrong, especially with squash merges and stacked pull requests.
- Our Cost Dashboard Said $0. The Loop Had Made 3,466 Model Calls.
November 11, 2026 · One feature looped through implement, review and verify twelve times in three hours. Here's what that cost, why our numbers undercounted it, and what actually drives the bill for agent pipelines.
- Spec-Driven Development, End to End: From a Ticket to a Merged PR
November 7, 2026 · What actually happens between "here's a ticket" and "it's merged" when every stage is an AI agent and the spec is the contract. The files each stage writes, and why they matter more than the prompts.
- Evals for Agent Pipelines: Scoring Every Handoff in Phoenix
November 4, 2026 · Unit tests can't tell you a multi-agent pipeline is quietly losing facts between stages. We score every handoff, report and loop on real runs and chart it next to cost and latency.
- Human-in-the-Loop That Doesn't Slow You Down
October 31, 2026 · Approval gates are what make AI agents safe to run on a team, and what makes people hate them. Three small UI changes turned our gates from friction into a 30-second decision.
- Your Expo App Crashes on iOS 27 at Launch: The UIScene Fix
October 28, 2026 · iOS 27 enforces the UIScene life cycle and crashes apps that still start from the app delegate. Here's the crash signature, why Expo apps hit it, and a 40-line config plugin that fixes it.
- Your Host Was Breached. Here's How to Rotate Every Secret Without Downtime.
October 24, 2026 · When a hosting provider says your environment variables may have been read, every secret has to change. Two of ours couldn't change safely, until we taught the app to accept the old and new value at once.
- Coding Agents Are Not Enough. Your Team Needs a Software Factory.
October 22, 2026 · AI made writing code cheap, not shipping software. Why teams need a software factory — specs, review and QA as stations, an orchestrator, deploy verification and a record — and not just a smarter agent.
- What Broke When I Let AI Agents Run My Software Process for a Month
October 21, 2026 · Twelve loops in three hours, summaries that forget commit hashes, and a token in my logs. Lessons from building an assembly line of coding agents with human gates.