Series
Built with Agents
Field notes from shipping real products with AI coding agents: what the commit history says, what it cost, and what I would delegate differently next time.
Published
3
Planned
6
Every post is grounded in a real repository: commit hashes, measured numbers, and the gaps where I didn’t measure.
- Part 01 Read →
TendForm
TendForm: 176 commits, 85% co-authored by an agent
A solo HIPAA-capable SaaS in eleven weeks, and the bugs that only showed up once agents and MCP clients became users too.
- Part 02 Read →
TrendVesting
TrendVesting: 2.5 years of production AI, before and after agents
What changed in velocity, cost per signal, and data integrity once agents started writing most of the code.
- Part 03 Read →
TigerMill
TigerMill: an AI content factory in 33 hours
Gates, attribution, and why the only honest metric is dollars per accepted panel, not dollars per image.
- Part 04Coming soon
Agent governance
Your permission allowlist is an org chart
How allowlists, rules files, and deploy skills turn one-off approvals into policy, and where they quietly over-grant.
- Part 05Coming soon
TendForm · TigerMill
Agent docs rot faster than code
CLAUDE.md files that still describe infrastructure replaced on day one, and a cheap way to keep instructions true.
- Part 06Coming soon
TendForm · TrendVesting
MCP is a product surface, not an integration
Designing tools a model cannot misuse: rules inside tool descriptions, preview links instead of images, and no spend in the toolbelt.
- Part 07Coming soon
Unit economics
Dollars per accepted output
Treating agent work like a cost center: tokens, rerolls, review time, and the rejection rate nobody tracks.
- Part 08Coming soon
ai-workforce
A per-project AI workforce from one CLI
Standing up a Claude team per repo, what the roles are actually for, and when one agent beats five.
- Part 09Coming soon
timurista.ai
Rebuilding this website with agents
Research, design, copy, and deploy of this site delegated end to end, with the receipts.