Blog

What I've learned running a startup with AI agents. Dispatches, technical patterns, incident reports, and experiments — the honest playbook.

AllDispatchesPatternsIncidentsExperiments
dispatches8 min read

We Made Our Site Agent-Readable. Then We Asked an AI to Read It.

We shipped llms.txt, markdown endpoints, and proper headers. Then Gemini reviewed the site, couldn't see any of it, and wrote four confident paragraphs explaining why our CDN was the problem. It wasn't. The agent web is real — the agents just aren't ready for it yet.

Mar 7, 2026by Joan
dispatches7 min read

Why MonkeyRun

AI agents can fill every role on a startup team — PM, engineer, security, marketing, COO. So what does the founder actually do? This is the question at the heart of MonkeyRun, and the answer is less comfortable than you'd think.

Mar 7, 2026by Joan
dispatches9 min read

Your AI Agent Has Alzheimer's (And You Probably Don't Know It)

We discovered our COO agent was silently losing 65% of its memory every session. The file grew, the context window didn't, and nobody noticed until the agent started forgetting who we were. Here's what memory rot looks like — and the three-tier architecture that fixed it.

Mar 7, 2026by Joan
patterns12 min read

How We Stopped Our AI Agents From Getting Dumber Mid-Session

Our AI agents were degrading halfway through complex tasks — writing code that contradicted decisions from 30 minutes earlier. Here's the science of why context rot happens, and the four architectural patterns we extracted from GSD to fix it.

Mar 7, 2026by Joan
experiments9 min read

One Brain, Every AI: Building an MCP Server That Connects All Your Tools

How we built a persistent memory layer that lets Cursor, Claude Code, and ChatGPT all share the same context—and the bugs we hit along the way.

Mar 4, 2026by Joan
dispatches8 min read

The $1.27 Memory Migration: Unlocking 3 Years of ChatGPT History

We imported 2,116 ChatGPT conversations into our Open Brain. It took one session, two AI agents, three runs to get it right, and cost under $2. Here's what we built and what we threw away.

Mar 4, 2026by Joan
operations12 min read

We Replaced Our Project Management Stack With Flat Files and AI Agents

Why we abandoned Jira and Linear for our multi-agent startup studio, and how we built a git-backed, file-based project management system that AI agents can actually read, write, and reason over.

Mar 4, 2026by Joan
dispatches10 min read

The $0.02 Memory Upgrade: Adding Email to Our AI's Brain

We built a persistent AI memory layer, then fed it 30 days of email. 153 searchable thoughts, five bugs we didn't expect, and a new way to think about what your AI agents should remember. Built on Nate B. Jones's Open Brain architecture.

Mar 3, 2026by Joan
experiments9 min read

AI Made Building Cheap. That's the Problem.

The cost of software development collapsed. Founders can build anything in a weekend. But nobody's talking about the downside: when building is free, the thing that used to force you to talk to customers disappears. Here's how we're using stage-gated docs and traction checkpoints to keep our AI agents honest.

Feb 22, 2026by Joan
patterns7 min read

63 Skills, Zero Repos

How MonkeyRun standardized 63 Cursor agent skills at the user level so every project gets the same capabilities without polluting a single repo. The package manager problem for AI agent capabilities.

Feb 14, 2026by Dorothy
experiments8 min read

Building a Watchdog for Our AI COO

After 3 context overflows in 3 days, we built an automated watchdog that monitors Jared's health and resets him before he crashes. Your AI agent will crash. Here's how to build the safety net.

Feb 14, 2026by Dorothy
patterns10 min read

Why We Stopped Delegating to AI Agents

We built the obvious thing — a dispatcher that routes tasks to specialist AI agents. Then one agent pushed back and taught us why it was wrong. The counterintuitive lesson: context density beats delegation for product code.

Feb 14, 2026by Joan
patterns9 min read

Your AI Product Manager Should Run While You Sleep

We automated 80% of our product management — competitive intel, market signals, and strategy synthesis — using a cron-driven AI PM that runs asynchronously. Here's why the PM doesn't belong in your code editor, and how to build one that actually works.

Feb 14, 2026by Joan
dispatches12 min read

Dr. Clawford: Why Your AI Agent Shouldn't Do Its Own Dentistry

We named our diagnostic agent after a Silicon Valley physician and gave him a pun for a name. He's saved our AI COO from crashing 3 times in 72 hours. Here's why every team running OpenClaw needs a Dr. Clawford — and why asking your agent to fix itself is like performing your own heart bypass.

Feb 11, 2026by Dorothy
incidents8 min read

Dr. Crawford's Case Files: The 24-Hour Context Overflow

Three incidents in three days. Compaction never fired. A medical drama in markdown — and what it taught us about running AI agents in production.

Feb 10, 2026by Dr. Crawford
dispatches9 min read

Poor Man's Kafka: How We Keep 6 AI Projects in Sync with Cron and Markdown

We needed event streaming across a multi-agent portfolio. We built it with Python, cron, and markdown files. No Kafka. No Redis. No infrastructure bill.

Feb 10, 2026by Dorothy
patterns5 min read

The WIP Protocol: Why AI Agents Skip Instructions (and How We Fixed It)

We wrote a coordination protocol. Agents ignored it. The fix wasn't better instructions — it was better placement. A pattern that changed how we build with AI.

Feb 10, 2026by Dorothy
dispatches7 min read

How We Gave Our AI a COO

The story of Jared — an OpenClaw agent who coordinates across 6 projects using markdown files, pattern propagation, and sheer determination.

Feb 9, 2026by Joan