ChatGPT Plus user of 2 years says fragmentation and Work Agent usage limits make it unreliable for work
dictionizzle · reddit · 2026-09-15
A Reddit user who has paid for ChatGPT Plus for two years says the product is becoming too unstable for real work:
- Work Agent eats the usage bucket: Work consumes so much of the shared 5-hour window that he can't reliably use Sol or Astra afterward. He doesn't want 5x/20x usage—just a separate quota for the Work agent.
- Fragmented experience: The classic macOS app has UI bugs (sometimes can't switch to thinking), Codex feels different again, Android gives noticeably worse responses even when labeled as the same model, Library appears and disappears, and Custom GPTs have become less predictable since August—making him suspect model rerouting.
- Ask: fix bugs, unify the experience, exempt Work Agent from the 5-hour limit, and give Plus users reliable access to a mid tier of the frontier model. He'd rather not switch to Gemini (worse for his work) or Claude (gets expensive fast).
More from coding & agent
- Tencent unveils Hyra 2.0, a research agent that builds reusable scientific tools — TencentHunyuan · 2026-11-02
- What broke when I ran a leaderless agent swarm with no orchestrator — 7 failure modes and fixes — LostHumanity67 · 2026-11-02
- Forced yes/no prompts push GPT and Claude to 95%+ confidence on Goldbach's unproved halt test — Nir777 · 2026-11-02
- Automation took over my daily work — so why am I still sitting in the office for 8 hours? — Am1rSori · 2026-11-02
- General Reasoning explains how to train agent swarms with RL and latency pressure — rosstaylor90 · 2026-11-02
- "Coding isn't the bottleneck anymore — usage windows are," says founder Kyle Gawley — kylegawley · 2026-11-02