François Fleuret: pre-AI software is awful, coding AI will rebuild what SaaS solves
francoisfleuret · x · 2026-09-08
Meta researcher François Fleuret weighed in on whether per-user AI-generated software could replace SaaS, noting a SaaS's value comes from engineers interviewing hundreds of users, while coding is just a constant cost.
His claims: (1) current "pre-AI" software is generally awful; (2) coding AI will know the field, best practices, every GUI and protocol design principle, and be extraordinarily good at interviewing users and implementing what they actually need — reshaping the logic of fixed-interface SaaS.
More from AGI Musings
- OECD: Students heavily using AI for writing scored 28 points lower in science tests — Polymarket · 2026-09-08
- On LBC radio: AI companies can no longer reliably control their systems — ShakeelHashim · 2026-09-08
- Full LBC segment: can AI companies still control their systems? — ShakeelHashim · 2026-09-08
- Oxford researcher: agents are 'chaos technology', can't be aligned by tuning or prompts — StefanoGogioso · 2026-09-08
- PISA scores keep falling across OECD, with reading comprehension declining fastest — paulnovosad · 2026-09-08
- Paras Chopra: Every future app will ship with an agent that rewrites itself — paraschopra · 2026-09-08