Field notes

Notes from
building with agents.

What worked, what broke, and what we'd do differently, written while shipping, not after.

ai-agents
multi-agent
bellclub
production
FeaturedOct 2026Zerolve Team

Five Agents, One App, Twelve Days to the App Store

How five parallel agents took Bell Club from fork to App Review in twelve days: 288 merged PRs, 4,297 tests, and the five rules that made it work. Plus everything that broke.

Read the post →
Oct 2026Bell Club Is on the App Store: Train Solo, Get Stronger TogetherOur sixth app is live: kettlebell training on iPhone and Apple Watch. One bell, a plan that adapts week by week, and a private crew of up to twelve people who see that you showed up.bellclubAug 2026Waiting 4 Is on the App Store: The Anti-Productivity AppOur fifth app is live: a single-screen iOS app that tracks what other people owe you. No projects, no streaks, no guilt. Just who, what, and how long it's been.waiting4May 2026How We Built AI Agents That Improve ThemselvesInside our journey from a specialized agent squad to a self-improving system. Real production learnings, honest failures, and the architecture that makes agents better over time.ai-agentsApr 2026240 Tests in 7 Days: Building Testing Guardrails from Day OneHow a UI/UX redesign triggered a full testing overhaul: designing test suites from scratch, integrating AI agents, and learning why testing guardrails are fundamental from the start.testingFeb 2026Introducing Agent Squad v1.1: Our AI-Powered Development TeamMeet the five specialized AI agents that power zerolve's development: Pi the orchestrator, Scout the researcher, Daedalus the designer, Muse the design strategist, and Forge the builder.ai-agentsNov 2025Use Claude Code with Your GitHub Copilot SubscriptionStop paying for multiple AI subscriptions. Use your existing GitHub Copilot subscription to access Claude Sonnet 4.5, GPT-5, and more AI models in one unified development environment.claude

One post, when it's worth it.

or subscribe via RSS →