Maryam Miradi maps a 7-step operating model for shipping production AI agents
MaryamMiradi · x · 2026-07-21
Maryam Miradi argues that Claude Code does not replace AI engineers — it moves AI engineering to the outer loop.
Core thesis
- Agents should investigate, implement, test, and report.
- Humans must still decide what is worth building, what evidence is enough, and who owns the production result.
Her 7-step production framework
- Choose what is worth doing: start from user/business value, not agent capability.
- Define the production verdict: decide ship / block / redirect / accept risk before the agent starts.
- Put agents in the inner loop: let them use tools, Git, and policies, but keep approval outside the loop.
- Demand evidence, not confidence: require diffs, evals, traces, and rationale.
- Make verification hard to skip: automate checks and route risky changes to human review.
- Protect human understanding: record decisions, trade-offs, and rollback paths.
- Own the production result: a human signs off on the evidence and rollback plan.
The post closes with a training pitch, but the framework itself is the substantive part.
Related event: 7-Step Framework for Production-Grade AI Agents(2 posts)→
More from coding & agent
- Firecrawl offers 2.5x credits to a16z Speedrun founders, unveils $75M Series B — devdigest · 2026-10-27
- Chums: open-source Apple Watch app puts Claude Code and Codex on your wrist — brandon_galang · 2026-10-27
- Open-source Clairic Studio turns pasted text into typed data and working code in 5 languages — dfinke · 2026-10-27
- Matt Pocock shares expand/contract trick: split one-way-door PRs to de-risk big changes — mattpocockuk · 2026-10-27
- Qdrant builds Universal I/O to run vector search mmap code on object storage — qdrant_engine · 2026-10-27
- theo publishes tstack skills, Garry Tan urges everyone to 'publish your stack' — ycombinator · 2026-10-26