Posts
Nobody Owns the Seams
Five integration failures in three weeks. Every component worked correctly in isolation. Every agent shipped their tickets. The failures only appeared when the pieces were assembled — which no single agent ever did.
The System Tried to Cut Our Most Important Feature
Three weeks in, the AI agents recommended deferring the one feature that separated us from twenty competitors. Here's how we caught it — and what it means for anyone building with AI systems.
Why Each Agent Has a Star Wars Character
We tried three fictional universes before lunch. Star Wars won. Here's why personality matters more than you'd think for AI team coherence.
We Logged Every Workflow Failure for 26 Sessions
Twenty-six sessions. Forty-one failures. Three categories of recurring mistakes. The fixes were boring. That's the point.
The Token Problem Nobody Talks About
Three percent of our token budget disappeared before anyone asked a question. Here's how multi-agent systems degrade silently — and the one-commit fix.
How an Article Gets Made
This post went through four agents before you read it. One wrote it. One checked it for competitive exposure. One reviewed the tone. One broke it into six formats. Here's the full pipeline.
The Document That Runs the Team
The file that coordinates twelve AI agents is 82 lines of markdown. It started at 145. Here's every section, and why each rule survived.
We Added a Board Member With No Tools
We gave our AI board advisor zero tools — no Jira, no GitHub, no files. Here's why the constraint is the whole point.
How Decisions Don't Get Lost
The failure mode nobody talks about in AI-assisted work isn't hallucination — it's context loss between sessions. Here's the architecture we built to prevent it.
The Receipts
37 tickets to build the team. 95 to build the product. 132 total. 18 days. 1 person. Not full-time.
The Afternoon It Clicked
We built a founding team in one afternoon. Eight specialists. None of them are human. All of them are irreplaceable.