$30k of inference cost only $13.25 billable — subscriptions buying 50-100x face value
arthurcolle · x · 2026-09-15
Follow-up from arthurcolle: of $30k total model usage, local records show only $13.25 was billable (non-subscription dsco lanes); the rest rode ChatGPT/Codex and Claude Max subscriptions — paid subscription fees are effectively buying inference at roughly 50–100x face value. Dedupe cross-check found only 2 duplicate records between the dsco ledger and Codex rollups.
Related event: Developer's AI Bill Shows $30K Nominal Usage Cost Only $13 Extra(4 posts)→
More from Infra
- Four dev boards hooked to the internet: test AI-written firmware on real silicon via HTTPS — SelfishlyWandering · 2026-09-15
- Grouped Value Attention shrinks KV cache by reconstructing keys on demand — Vishesh Tripathi · 2026-09-15
- jinfer brings native AI inference to the JVM, matching llama.cpp on CPU with zero Python — mukel90 · 2026-09-15
- Wan 2.2 on one RTX 5090: frame count doesn't touch VRAM, but resolution drops it by 10GB — Realistic-Fennel-190 · 2026-09-15
- Stateless LLM failover preserves ~0% context; ContinuityBench proxy hits 99.20% CPR — its_vayishu · 2026-09-15
- 500 trillion tokens in a week: chart points to surging LLM usage — gajesh · 2026-09-15