Debugged an articulation model with real speech data — 305.6k tokens

Debugged an articulation model with real speech data

aai

First session with sonnet-5

First build using this model

305.6ktokens

16prompts

64:21:55time

Claude Codeclaude-sonnet-5$3.98

mostly thinking

Tested the model on authentic speech audio and generated trajectory visualizations with correlation metrics.

Tokens / prompt19.1k

Cache hit97%

Burn rate$0.06/hr

Derived from this session's token and cost data. Not shown on the feed.

Building 18%Verifying 8%Thinking 74%