Tell HN: OpenAI brings back 5 hour limit for plus and business standard users
OpenAI has reinstated a 5‑hour usage window for its $20 Plus and standard business subscribers, prompting debate over fairness, usability, and the company’s path to profitability. Some users see the cap as a reasonable way to prevent people from burning through their weekly allowance too quickly and as an incentive to upgrade to higher tiers, while others argue it undermines serious or time‑sensitive work and feels like a bait‑and‑switch after a period of looser limits. The thread also touches on broader concerns about AI pricing, market competition, long‑term vendor lock‑in, and the viability of alternatives such as open‑weight models, third‑party inference providers, and local deployments.
Overall Reaction to the 5‑Hour Limit
- Many see the 5‑hour session cap as making Plus/standard less useful, especially for coding “sprints” and larger continuous tasks.
- Others argue it’s a reasonable way to ration a scarce resource, preventing users from burning an entire weekly allowance in a single binge session.
- Some like it because it naturally enforces breaks and helps spread usage across the week.
- There is frustration that weekly resets do not also reset the 5‑hour window, which feels “sloppy” or punitive.
- Frequent rule changes are criticized as “bait and switch,” undermining trust even if the economics are understandable.
How the 5‑Hour Window Works and User Workarounds
- The window starts when a user resumes usage after a lull; it’s a fixed 5‑hour block with its own token quota, then fully resets.
- Several posters discuss gaming the window via scheduled “ping” prompts before work hours to shift when the window falls.
- For larger or continuous workloads, people suggest switching to API usage or higher tiers.
Alternatives and Usage Strategies
- Users describe mixing providers: OpenAI Plus/Pro, Opencode Go, Hyper, OpenRouter, DeepSeek, Claude, Gemini, and various cheap or open models.
- Hyper in particular is highlighted as a better-value flat subscription with a daily dollar-equivalent quota and high speed, assuming you like its model lineup.
- Some rely on open or local models for most work, reserving frontier models like Astra/Sol for final reviews or particularly hard problems.
Pricing, Value, and Tiers
- The 5‑hour cap is widely seen as nudging users toward the $100+ plans, which currently lack that limit and can deliver token usage worth far more than their sticker price.
- Some call the base plans “trial‑level” and unsuitable for serious, high-iteration development work; others find them adequate for light or search-like use.
- Suggestions include: opt‑in caps instead of mandatory ones, time‑of‑day pricing to shift load, and an expensive “unlimited” tier for power users.
Economics, Competition, and Lock‑In
- Posters debate profitability: inference may be profitable per call, but overall businesses are still burning cash on training and infrastructure.
- Many see current prices as subsidized and unsustainable, expecting higher prices and/or tighter caps once growth slows or after IPOs.
- Others note strong competition (including open models and third‑party inference providers) and relatively low switching costs as limits on future “monopoly rents.”
- There is ongoing concern about lock‑in via proprietary “memory” features and data, with advice to keep architectures portable across models.
Social and Ethical Themes
- Some frame the pricing and limits as an “economic drug” strategy: hook users with cheap, powerful tools, then raise prices once dependence forms.
- There is concern that high‑end tiers are accessible mainly to wealthier users or well‑funded companies, making AI a “wealth inequality accelerator.”
- A counter‑view is that users should exploit today’s low prices while they last and be prepared to switch providers or models as conditions change.