Canary

Adversarial AI that breaks your AI

1.5Score#883 of 999 agents#11 of 15 in QA engineer

Why it ranks #883

Canary, an AI QA engineer from YC W26, ranks #883 of 999 AI agents and #11 of 15 in QA engineer, with a score of 1.5 of 10. Data as of 3 Oct 2026.

The score adds four parts, each 0 to 10, by weight. Every point comes from public evidence. How scoring works ·

Proof 0 × 30% 0.0
Scale 1 × 30% 0.3
Momentum 1 × 25% 0.3
Autonomy 6 × 15% 0.9
= 1.5

Proof
Do customers use it? 30% of the score
0
No public claims of customers or revenue yet.
Not found: revenue or growth (+4), named customers (+3), a customer result (+2), a customer count (+2), work done at scale (+2)
Scale
Is there a real company behind it? 30% of the score
1
2 people.
- +1
  2 people on its YC page2 people
Not found: funding round in its news (+3)
Momentum
Is it shipping and growing now? 25% of the score
1
Shows a new batch.
- +1
  New: YC Winter 2026
Not found: launch YC post in the last 6 months (+3), open roles on YC (+2), new traction claim in the last 6 months (+2), funding news in the last year (+2)
Autonomy
How much of the job does it do on its own? 15% of the score
6
Supervised agent: takes actions in other systems; a person approves key steps.
- +6
  Level 2 of 4: Supervised agentTakes actions in other systems; a person approves key steps.
What the model read
- Acts in other systems: 43% likely
- Finishes whole tasks: 21% likely
- Calls itself an agent: 25% likely
- Level 1 from the model's reading of its pages; the evidence review found, with quotes, that its AI acts in the work, so the level is 2
Level 2: Takes actions in other systems; a person approves key steps.

About Canary

AI coding tools have made developers 100x faster but customer-facing incidents are up 43% YoY, QA hasn't kept up. Canary is the first AI QA engineer that understands your codebase. Instead of relying on flaky DOM scraping or screenshot analysis, Canary reads your source code directly to understand developer intent and catch broken user flows before they hit production. Engineering teams go from weeks of manual testing to 90%+ coverage in days. Built by a team that built coding agents at Windsurf, Cognition and Google.