Post by @jasongorman@mstdn.business

2026-10-03 13:29 UTC

Pooled together the best available research and added it to a new section &quot;Code Maintainability &amp; LLM Performance&quot; The 101? Don&#39;t listen to the folks telling you it doesn&#39;t matter if AI-generated code is maintainable by humans. The data says it really does matter. https://codemanship.wordpress.com/2026/08/12/ai-software-development-what-does-the-data-say/

Replies (2)

@nice-nigger@nicecrew.digital 2026-10-03 13:30

muh fuh guilteh geah muhfuckin shounen anime make me basenrehpill gnawmean I giddat tampon bakpak nigguh wuhchoomean I ain no girlboss bitch I be duh baddiss bitch on duh blawk I got dat nine to five nigga I got dat cash money nigga I show fediverse admins muh tee-tees ain't nobody scurr yo reddit ass muhfuggih schizo shit bitch daaaayuuuuuuum mercedes dealership n sheeit sail of foam I got dat ifoam plus main on gawwwww :peachcat_spank:

Open ##5338690
@laurentgiroud@retrorewind.social 2026-10-03 17:03

@jasongorman@mstdn.business @48kRAM@retrorewind.social Holy carrot, this is damning. This matches my experience with LLM coding: reliability is an issue and bringing it to the level I expect of my team members required significant investment which 1) negates the benefit and 2) is fairly logical/predictable (reduced input implies low information density which necessarily increases the imprecision of the output. Human coders bring an enormous dataset of experience embedded which compensates for it but LLMs haven’t).

Open ##5338694