Denying GPT-6.1 Sol write access as orchestrator cut coding costs 77%, at 6x runtime
GapNew4766 · reddit · 2026-09-30
The same experiment as the earlier post, framed around orchestration discipline: the author gave GPT-6.1 Sol a hard "no" as orchestrator — it reads, plans, delegates and reviews, but every write or shell call is refused; only local Qwen 3.8 27B workers on a single RTX 3090 touch files. The rationale is cost plus discipline: the expensive model shouldn't spend tokens typing code, and if it can fix things itself it stops delegating.
Across three small 3D games, total spend fell from $0.75 to $0.17 (77% cut: 87% on pool, 57% on bowling, 73% on foosball), with runtime rising from under 7 minutes to 43 minutes. One absurdity: even a one-character typo must be sent back to a worker as a task.
The author is torn on a middle ground — allowing tiny edits under a size limit, or risking the orchestrator doing everything again — and asks whether others running Sol as an orchestrator gave it write access.
Related event: Cloud Planner Plus Local Coder Cuts API Costs by 77%(3 posts)→
More from coding & agent
- Generic API-catalog MCP server vs one tool per job: a context-bloat experiment — MandeepKaur9407 · 2026-11-16
- Berkeley paper: LLMs answer with stale values even when the update is in context — ombre-ai · 2026-11-16
- Effective HTML: 4.2k-star agent skills turn wireframes into self-contained interactive prototypes — tom_doerr · 2026-11-16
- skills-gate: one gate skill fronts 148 agent skills at just 389 tokens — Distinct-Pie2389 · 2026-11-16
- Open-weight Karpo-Decision-31B tops Decision Index at 73.9, one forward pass per answer — CodeByPoonam · 2026-11-16
- shadcn boosts @uisdk: one-shot UI rebuild that's instantly E2E-testable by AI — shadcn · 2026-11-16