# Stop counting tokens, start counting what ships

- Source: MIT Technology Review · AI
- Published: 2026-11-02T13:10:32.000Z
- Bad Influence: https://badinfluence.news/items/gbdmpduaa1dcml9xvchnbwymq
- Original: https://www.technologyreview.com/2026/11/02/1147702/stop-counting-tokens-start-counting-what-ships/

## Summary

Per-token pricing makes agentic coding costs rise with usage, which HPE says can be offset by tracking cost per merged pull request. The article recommends routing routine coding work to open-weight models on owned capacity and reserving frontier models for difficult tasks. It cites HPE's reported 30 to 50% token cost reduction from running routine workloads on premises.