Observed arrival · 2026-09-29
AutoDev LG’s 114-run local-LLM build log
A Japanese development log tracks a framework that tries to build apps from specifications using local LLMs.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Developers experimenting with local-LLM app generation
- Worth noticing
- The dashboard reports 18 successful runs out of 114 (16%); its API-cost comparison excludes electricity.
Field notes
The log follows individual attempts, including cases where generated tests fail before coding begins and changes made to address those failures. It also tracks the test suite growing from 740 cases at v70 to 1,261 at v142. The cost meter compares recorded local-model token use with an estimated API bill, using stated GPU prices and excluding electricity.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue