Dimillian: as models get smarter, intent-based single-thread prompting beats agent orchestration
Dimillian · x · 2026-10-05
iOS developer Dimillian (Thomas Ricouard) questions whether we overcomplicate AI coding workflows as models improve: he now favors an intent-based approach, running a single thread and queueing a whole project via prompts — and it works fine. Months ago devs celebrated agents moving Linear tickets from todo to in-progress; now tools add new orchestration primitives he isn't sure he needs. He admits he likes staying close to the prompt box, micromanaging and over-testing everything, wondering if he's "just a glorified QA engineer" — and whether frontend vs backend work explains it. His takeaway: no point building for building's sake, and process should be re-evaluated against model capability.
More from coding & agent
- xAI launches Grok Bot: AI teammates that log into your tools and finish real work — aitrendz_xyz · 2026-10-05
- Grok Bot as a real personal assistant: 12 hands-free use cases from food ordering to job hunting — aitrendz_xyz · 2026-10-05
- Dev says telling his coding agent to stop using subagents made everything work better — QuixiAI · 2026-10-05
- Permission granted but action wrong: the agent authorization problem — Invisible_act1988 · 2026-10-05
- AI agents given one prompt and $1 each to produce 30s videos: Opus 5.5 vs GPT-6 Astra — no3us · 2026-10-05
- Agentic Field Guide launches with 7 hands-on Kubernetes agent labs, 9 more coming — davemccollough · 2026-10-05