Run 21159ad5
$0.000717
total cost
6,173
input tokens
41%
cached
2
turns
active
status
Total cost
| component | share | usd | |
|---|---|---|---|
| LLM measured | billed by the provider, reconcilable against the gateway log | 10.4% | $0.000717 |
| TTS estimated | 221 chars the interviewer would speak | 48.2% | $0.003315 |
| STT estimated | 22s of candidate audio · Cloudflare Workers AI — @cf/openai/whisper | 41.4% | $0.002846 |
| total | $0.006877 |
TTS $15/1M chars · STT $0.00762/audio-min · speech figures are ESTIMATES from the transcript — no audio was produced or transcribed
STT rate source: neuron rate inherited from probe-atlas; neurons-per-minute is an unverified guess
Transcript & scoring
turn 1 · pull · motivation · core
Q. What's drawing you to this role?
That's a really clear way to put it—liking that the failure was concrete.
A. I moved into backend work after a data migration went badly on my second project. We lost two days because nobody had checked the encoding on the source files. I ended up owning the import path after that, and I liked that the failure was concrete - you could see exactly which row broke and why.
turn 2 · evidence · self_awareness · core
Q. What's an area you're actively trying to improve in your work?
A. (unanswered)
Cost by stage
| stage | calls | wasted | in | cached | think | out | cost |
|---|---|---|---|---|---|---|---|
| analyze_turn | 1 | 1938 | 0 | 0 | 182 | $0.000322 | |
| generate_questions | 1 | 2376 | 2304 | 0 | 492 | $0.000154 | |
| interviewer_turn | 1 | 1859 | 256 | 0 | 53 | $0.000240 |