No. 79
2026-10-06
the AI pulse — brewed, not raw
 AI CRITIQUE

Simon Willison releases pwasm 0.2a0

Read the article    English (automatic translation)

Summary

Simon Willison released pwasm 0.2a0. This pure Python-based WebAssembly engine was created using 'vibe-coding' methods in January 2026. Subsequently, Claude Opus 5.5 improved the code with 42 commits, covering nearly the entire WASM specification. PyPI packages also include working builds of MicroPython and QuickJS.

Why it matters

This provides a small but concrete data point on whether today's large language models (LLMs) can meaningfully iterate on code generated by themselves or their previous versions ten months ago. Willison describes this as a 'sanity project' and explicitly states he does not trust it. However, the before-and-after process is read as an informal, real-world benchmark of model progress. Additionally, the continuous blurring of the line between toys and tools in the 'vibe-coding' debate contributes to this topic.

Highlights

  • Claude Opus 5.5's 42 commits with minimal prompts brought pwasm to near-complete WASM specification coverage
  • The PyPI wheel now includes working WASM builds of MicroPython, QuickJS, and Micro QuickJS
  • Willison explicitly labels the release as alpha and says 'I wouldn't trust it at all'

Critical take

Without diff analysis, test suites, or performance data, the phrases '42 commits' and 'minimal prompts' carry heavy rhetorical weight. It is unclear whether the engine is actually more correct or just superficially more complete. The post reads more like a showcase of model improvement than an engineering update. This is not a problem, but it offers very little for a developer to build upon.

Bu kritik, yerel yapay zeka modeli Qwen3.8-27B tarafından yazılmıştır. Haberi ben seçtim; kritik metnini yayımlanmadan önce tek tek denetlemiyorum.This critique was written by the local AI model Qwen3.8-27B. I chose the story; I do not review each critique before it is published.
Back to the issue