OctaPulse (YC Winter 2026): vision and robotics for fish farm quality inspection
Skoor 30 · n=3 signals · confidence 0.7 · sector tag: energy
What they built
OctaPulse describes itself as "CV and robotics to automate quality inspection in fish farms." [source]
The company says its AI vision platform automates hatchery QA, starting with broodstock phenotyping and juvenile deformity inspection, using off-the-shelf cameras. [source]
It says it plans to expand into feeding, health monitoring and processing. [source]
Founders' names and backgrounds: not in the evidence.
Record
Measured (n=3, all from one page observed 2026-10-14):
The YC page lists status Active, stage Early, and not hiring. [source]
Self-reported, not independently verified:
Repos, funding and outside hiring data: not in the evidence.
Inference: All three signals measure YC membership and team size. None measures traction.
Why now
The company claims aquaculture is the fastest-growing food sector and has passed commercial fishing in seafood production. [source]
It claims farms spend more than $200K a year on technicians and geneticists for these steps. [source]
Inference: Using off-the-shelf cameras lowers hardware cost per deployment. This depends on the software generalizing across species, which is not shown.
Unknowns
- Who the founders are and what technical work they have done: not in the evidence.
- Whether the 2 early-2026 deployments happened: not in the evidence.
- How "90 percent accuracy" is defined, and what it is measured against: not in the evidence.
- Inference: The "energy" sector tag does not match an aquaculture company. YC lists it under Industrials and Climate. [source]
- The term "robotics" appears in the one-liner, but the description covers only vision. [source]
Questions for the room
- What does "above 90 percent accuracy" measure? Is it per-fish deformity classification against a geneticist's labels, and on how many fish?
- Is the trout pilot paid in cash or contingent on results, and what does it take to convert to a contract?
- Are the 2 additional farms live today? Are they different species, and how did the model perform without retraining?
What would move the Skoor
It goes up with a named paying customer, the 2 extra farms confirmed live, or a founder record in vision or robotics. It goes down if the pilot is unpaid or stalled, the accuracy comes from a small internal test set, or the extra deployments did not happen.