# The exact tools used to port a massive codebase in days #programming #typescript #dev

Source: [AI Engineer](https://www.youtube.com/shorts/AJhjm4_UJIk)  
Feed7 permalink: https://feed7.dev/p/ajhjm4-ujik-1ei39n8  
Published: 2026-09-09T20:07:49.000Z  
Trust: Source Linked (source_linked)

## Why Included

A reported weekend Python-to-TypeScript port shows how agent-driven iteration can make once-prohibitive migrations plausible, provided verification is part of the loop.

## Source Summary

Mike Krieger describes using an agent workflow to port **a couple hundred thousand lines** from **Python to TypeScript** over **one weekend**, motivated by a better Bun deployment path.

## Practical Implication

For large migrations, rethink the unit of work: let agents repeatedly port, verify, compare both implementations, and continue through failures. Treat deployment readiness and behavioral parity as explicit acceptance criteria.

## Agent-Ready Context

Mike Krieger describes using an agent workflow to port **a couple hundred thousand lines** from **Python to TypeScript** over **one weekend**, motivated by a better Bun deployment path.

For large migrations, rethink the unit of work: let agents repeatedly port, verify, compare both implementations, and continue through failures. Treat deployment readiness and behavioral parity as explicit acceptance criteria.

This is one retrospective account, not a reproducible benchmark. The transcript provides no test coverage, defect rate, staffing detail, workload breakdown, or evidence that the same pace generalizes to other codebases.

## Connected Context

Feed7 judgment across 732 accumulated Signals:

This turns goal-level delegation into a concrete migration loop: port, compare implementations, verify parity, and persist through failures until deployment criteria are met. It reinforces the prior emphasis on outcome evidence over generated diffs, but does not weaken the case for bounded review or safeguards; the reported weekend pace is an anecdote without enough quality or workload detail to serve as a planning benchmark.

- [How Anthropic Builds: Lessons from Labs — Mike Krieger, Anthropic](https://feed7.dev/p/how-anthropic-builds-lessons-from-labs-mike-krieger-anthropic-0ciws2c) — This supplies the specific Python-to-TypeScript migration behind the broader account of goal-level delegation, while preserving its emphasis on verification and bounded claims.
- [How to Kill the Code Review — Ankit Jain, Aviator](https://feed7.dev/p/how-to-kill-the-code-review-ankit-jain-aviator-0rku6kj) — Behavioral parity and deployment readiness are examples of the acceptance evidence humans can inspect when line-by-line review becomes impractical at migration scale.
- [Loop Engineering from First Principles — Kyle Mistele, HumanLayer](https://feed7.dev/p/loop-engineering-from-first-principles-kyle-mistele-humanlayer-1nuq7gf) — The large repeated migration loop contrasts with the candidate’s one-bounded-task-and-pause pattern, showing a more autonomous unit of work without establishing that it is generally safer.

## Context Map

- Layer: agent
- Domains: coding
- Topics: coding-agents, harness-engineering, agent-reliability

## Uncertainty

- This is one retrospective account, not a reproducible benchmark. The transcript provides no test coverage, defect rate, staffing detail, workload breakdown, or evidence that the same pace generalizes to other codebases.

## Agent Instruction

Use this item as source-backed context. Do not invent claims beyond the linked source. If this item conflicts with another source, call out the conflict.