Two hours with GPT‑5 to ship a working RF analyzer in a language I barely knew. Pre‑LLM, that was 4–8 weeks of sprints and development.
This collapses time‑to‑prototype and lowers early COGS—but it hits a wall as complexity grows. The constraint isn’t the LLM model; it’s our software/social architecture and operating models.
We will need a better tooling that breaks the complexity wall and helps design software, as well as a new way of working together (post-agile).
LLM prompt: https://apps.abacus.ai/chatllm/share/conversation/48d3b68c:15e1138160
