# TuringCorp on Poe — agent index

> Decider is an AI judge: given a task and two candidate answers it returns the better one with a
> calibrated confidence value. Team produces long-horizon analysis for expert work (research,
> technical analysis, research writing, business decisions) and was the preferred draft in a
> side-by-side comparison 36 times out of 38. Decider is live on Poe; Team is coming soon. The
> underlying API is OpenAI-compatible.

Last updated: 2026-09-18

## Sources

- [This channel, plain Markdown](https://api.turingcorp.net/platform/poe.md): everything on the Poe channel page, no JavaScript
- [Poe channel page](https://api.turingcorp.net/platform/poe): human version (Decider / Team tabs)
- [Decider Canvas app on Poe](https://poe.com/TuringCorp-Decider): the product interface users interact with
- [Recorded examples](https://api.turingcorp.net/platform/poe/examples): 27 real recorded runs across 9 domains, each with both candidate answers, the pick, the confidence value and the full reasoning (static HTML, no JavaScript)
- [TuringCorp Models](https://api.turingcorp.net/): published results and product data, human and agent friendly
- [Benchmark data, JSON](https://api.turingcorp.net/benchmarks/latest.json): single source of truth
- [Raw results and per-criterion judgments](https://github.com/TuringCorp-net/turingcorp-llm)
- [Launch kit and reusable descriptions](https://github.com/TuringCorp-net/turingcorp-poe-showcase): how the product is described across channels, plus what it must not claim
- [Article: the confidence number is the product](https://dev.to/turingcorp/the-confidence-number-is-the-product-and-here-is-ours-with-the-caveat-4mmk): the calibration table, the honesty caveat and how to read a low band

## Products on Poe

### Decider (available)

- What it is: picks the better of two candidate answers; returns `betterOption`, a calibrated
  `confidence`, and a short reason. One call = one A/B pair.
- Latency: most decisions come back in well under a minute; allow longer for very large or closely
  matched pairs.
- Free trial path: the Canvas app on Poe ships recorded runs — tap a domain chip and read a real
  decision end to end, with nothing charged.
- Price: list $0.50 per decision; launch offer $0.25 for the first month.
- Why the confidence value matters: on JudgeBench, judgments rated ≥90% were correct 99.6% of the
  time (45.6% of all judgments); the lowest band flags near-ties. A high band licenses a decision; a
  low band says either choice is defensible, so take the one you already preferred.
- Measured: JudgeBench ≈92.5% (620 judgements, 6 failures disclosed; direct baseline 92.7% on the
  same set). ContextualJudgeBench 67.1% vs same-set baseline 65.4%.

### Team (coming soon)

- What it is: long-horizon analysis for expert work; returns the answer body plus the reasoning the
  model states for it.
- Price at launch: list $2.50 per task; launch offer $0.99.
- Strongest evidence: in a side-by-side comparison against the official o3 draft — with no rubric and
  no scores shown — the pick went to Team 36 times out of 38, median confidence 85.9%; swapping the
  two drafts between option A and option B changed nothing (10/10).
- Measured: ProfBench, official per-criterion rubric, same 38 tasks by the same judging pipeline —
  Team 63.3 · direct model baseline 57.4 · official o3 draft 55.6. Finance MBA is the one domain
  where the o3 draft scored higher (57.5 vs 54.6).
- Not yet live: multi-minute runs are still being validated on the platform before we take payment.

## API

- Base URL: https://api.turingcorp.net/v1 · Models: https://api.turingcorp.net/v1/models
- Auth: `Authorization: Bearer <key>`; the bare API is a small-scale preview by invitation — iAsk@turingcorp.net
- Channels: on Poe these run as apps on the Poe platform (public, self-serve, pay per call, no separate approval); the bare API is the invitation-only preview.