# The Inner Loop

Exploring the 'Inner Loop' of AI-driven engineering from Tokyo. Principal Architect at Microsoft, veteran of Amazon & startups. Moving teams from 'vibe coding' to Hypervelocity Engineering—balancing system rigor with agentic orchestration.

Page: <https://rssamplifier.com/innerloopai-substack-com>  
Feed: <https://rssamplifier.com/innerloopai-substack-com.md>

---

## [I Killed My Agent Team](https://innerloopai.substack.com/p/i-killed-my-agent-team)

_2026-08-19 · Mike Lanzetta_

Sixteen agents, a year of tuning, and the week I noticed my code reviews had stopped arguing back.

## [A Juice Cleanse for Your Digital Brain](https://innerloopai.substack.com/p/a-juice-cleanse-for-your-digital)

_2026-08-05 · Mike Lanzetta_

Memory poisoning persists in AI agents. Hermes, Mem0, Letta, and Zep ship freshness hygiene; none ship decontamination. The case for a frontier-model audit.

## [Engineers Are Underwriting $4 Trillion of Debt and Don't Know It](https://innerloopai.substack.com/p/engineers-are-underwriting-4-trillion)

_2026-07-25 · Mike Lanzetta_

The AI build-out is $5.5 trillion of capex riding on depreciation schedules. The engineers shipping distillation tooling are the ones quietly setting them.

## [Sovereignty Is Not a Compute Bet](https://innerloopai.substack.com/p/sovereignty-is-not-a-compute-bet)

_2026-07-03 · Mike Lanzetta_

Japan put ¥1 trillion behind a homegrown model and 10 million robots by 2040. Everyone scoring it in GPUs is mispricing the bet.

## [Your LLM Is an Interpreter You Forgot to Compile](https://innerloopai.substack.com/p/your-llm-is-an-interpreter-you-forgot)

_2026-06-20 · Mike Lanzetta_

The case for compiling the model out of your runtime loop, and paying for live inference only where it earns its place.

## [There Are No New Engineering Disciplines](https://innerloopai.substack.com/p/there-are-no-new-engineering-disciplines)

_2026-05-25 · Mike Lanzetta_

In late 2023 I shipped a production RAG system.

## [Orchestrating an Agentic Crew at Scale](https://innerloopai.substack.com/p/orchestrating-an-agentic-crew-at)

_2026-05-24 · Mike Lanzetta_

A year ago, I proposed that engineering teams should “start slow” to build the foundations for AI integration.

## [Hypervelocity Engineering](https://innerloopai.substack.com/p/hypervelocity-engineering)

_2026-05-24 · Mike Lanzetta_

A Year on the Front Lines