# Dossier — Write the Answer First: Generating Eval Datasets for Agents With No Benchmark

- **Session ID**: `c4ba0138-6f6d-4780-93a1-4d8f0c478339`
- **Talk Transcript**: [transcript.md](transcript.md)
- **Recording**: `2026-10-13_1104_Air-Canada`
- **Freshness**: `settling` (last checked 2026-11-14T12:15:30-05:00)

## Speakers

- **Nathanael Weill** (Air Canada) · status: `partial`
  - Site: https://ai.engineer/nyc/2026/schedule
  - Evidence: *"Problem when we tackle some great vision for a task that is so specific, we don't have any particular benchmark we can rely on."*

## Companies

- **Air Canada** (https://ai.engineer/nyc/2026/schedule) — Organization represented at AI Engineer New York 2026 (Write the Answer First: Generating Eval Datasets for Agents With No Benchmark).

## Verified Resources

- **[Schedule | AI Engineer NYC 2026 — Write the Answer First: Generating Eval Datasets for Agents With No Benchmark](https://ai.engineer/nyc/2026/schedule#session-c4ba0138-6f6d-4780-93a1-4d8f0c478339)** (`product`, first seen 2026-10-26T10:41:07-04:00, post-event publication — added 13 days after the talk)
  - Why: Official AI Engineer NYC 2026 session entry and abstract for "Write the Answer First: Generating Eval Datasets for Agents With No Benchmark".
  - Evidence: *"Problem when we tackle some great vision for a task that is so specific, we don't have any particular benchmark we can rely on."*
- **[How to Test AI Agents When No Benchmark Exists — Nathanael Weill, Air Canada](https://www.youtube.com/watch?v=c_SPW__jF9w)** (`youtube`, first seen 2026-11-14T12:15:30-05:00, post-event publication — added 32 days after the talk)
  - Why: Official recording of this talk on the AI Engineer YouTube channel, published after the event.
  - Evidence: page title *"How to Test AI Agents When No Benchmark Exists — Nathanael Weill, Air Canada"*

## Coverage & Gaps

- Post-event settling window (12 days after 2026-10-14): standalone slide deck not yet published on the session page.