Heavy user: GPT-6.2 Sol and Astra follow instructions better but lost implicit intent understanding
that_antimatter · reddit · 2026-11-21
A ChatGPT Plus user who estimates 3+ billion tokens consumed in one month argues GPT-6.2 Sol and Astra have a deeper problem than incomplete outputs: the models now execute literally what you said rather than inferring what you meant. Earlier models handled loosely structured prompts gracefully; the new ones often return tasks half-done or need extra instructions.
- Skilled prompt engineers can compensate with scoping, constraints, self-verification instructions — but ordinary users shouldn't need that expertise.
- The author isn't asking for the model to invent requirements, just to interpret intent, spot missing steps, make reasonable assumptions, and carry tasks to completion.
- His conclusion: model capability is rising while the skill required to extract it is also rising — a usability problem, not a prompting problem.
More from Models
- GPT-6.2 Astra writing success rate falls to 85.9% from 91.2% — arena · 2026-11-22
- GPT-6.2 Astra gains +50 in coding, drops -54 in chat on Agent Arena — arena · 2026-11-22
- Apple testing daily limits on advanced Siri AI requests in iOS 27.3 beta, per leak — alvelda · 2026-11-22
- Muse plays GTA6, suggesting guardrails don't cover game playing — Dave_Maynor · 2026-11-22
- Paying Google Workspace users locked out of Gemini over custom-domain account restrictions — film_girl · 2026-11-22
- Lean Community Verifies 17M Lines of Code, Confirms Claude-Assisted Riemann Hypothesis Proof — Polymarket · 2026-11-22