Internal Coding Reasoning Compute Surges 100x
eliebakouch · x · 2026-07-10
Cited content reveals that over the past six months, the share of research compute allocated to internal coding reasoning has grown 100x, and internal agentic token usage has surged roughly 22x.
The poster speculates that compute consumption per token may have increased about 4x within half a year, reflecting a rapid shift in internal AI usage toward coding and agentic scenarios.
Related event: Frontier AI Models Consume More Tokens Despite Efficiency Gains(5 posts)→
More from coding & agent
- He benchmarks new models by having them build their own copilot — $28 of compute — bucolucas · 2026-10-27
- Should agents be allowed to change plans mid-task? Reddit debates the drift boundary — wistfulcracker84 · 2026-10-27
- HeyGen + Canva MCP turns slide speaker notes into an avatar presentation video — HeyGen · 2026-10-27
- Local beats cloud: Qwen 3.8 FN handles private student data as well as frontier models — dr_lm · 2026-10-27
- Cloudflare Computer adds agent interruption recovery and approval pauses in v0.7.0 — irvinebroque · 2026-10-27
- LLM Tails cuts agent input tokens 80% by making agents read before they write — savadks1818 · 2026-10-27