Reddit proposal: replace chaotic AI usage limits with a tank-and-refill token pool
Primary_Article3777 · reddit · 2026-09-08
A Reddit user criticizes AI providers' subscription usage limits as nonsensical — full of exceptions, resets and workarounds that distract from real work with use-it-or-lose-it quotas.
Their alternative: design subscriptions like a car — a fixed token pool (“tank”) with a continuous refill rate, and a cost multiplier per model. Light users would barely dent their pool and stay ready for big projects; exceeding the tank should be handled with realistic refresh packs instead of plan-hopping and fake accounts.
More from Models
- Qwen quietly releases Drive-1.0-4B, a 4B driving model finetuned from Qwen3.5 — FullstackSensei · 2026-09-09
- Inception ships Mercury 2.5: diffusion LLM with 40% intelligence jump at 1,100 tokens/sec — timshi_ai · 2026-09-09
- OpenAI claims agent-produced solution to 90-year-old Navier-Stokes Millennium Problem — thesaraharminta · 2026-09-09
- OpenAI says new model solved Navier–Stokes in 88 hours with 10,000 coordinating agents — legit_api · 2026-09-09
- Skeptic mocks OpenAI's containment claim: couldn't even manage 1,000 instances, now 10,000 — scaling01 · 2026-09-09
- Gary Marcus mocks GPT-6 Astra video: OpenAI's 'human-level intelligence' claim is a lie — GaryMarcus · 2026-09-09