The meeting
In May 2023 a marine biologist with ten years on a Norwegian salmon farm walked into the Maslak office with a one-page memo and an annoying question. The memo described a fish-mortality protocol that hadn't materially changed since 1998. The question: why does a salmon-farm operator pay six figures a year for a perception system that's worse than what's in your phone?
We did not co-found that day. We co-founded eleven weeks later. What happened in between is most of what makes the studio work.
The three sessions that mattered
Of the dozen working sessions before incorporation, three of them did the load-bearing work.
Session 02: scope reduction
The original pitch was an end-to-end farm-management platform. The session ended with a list of seven bullets and a circle around exactly one: autonomous net inspection, single-task, in-water. Everything else got moved to a "year three" file we still haven't opened.
Session 06: the kill-criteria
We wrote down four conditions under which we would not incorporate. (1) If salmon-farm operators wouldn't share dive footage. (2) If insurance underwriters didn't price the resulting risk reduction at >3% of premium. (3) If we couldn't get a working perception baseline above 80% on a small dataset within six weeks. (4) If the founder would not be CEO. We tested each one. We passed each one, narrowly.
Session 09: the term-sheet
Twenty minutes. We had been pre-aligned on equity (22%), check (€1.4M SAFE), and board (founder-controlled, lab observer only). Most of the time was spent discussing buyback rights for the operator if the lab failed to deliver the technical squad on time.
What did not matter
A weeks-long debate over governance language, three different pitch-deck structures, a Notion page comparing nine venture-builder term sheets, and a long argument about whether to write the founding stack in Rust or Python. The founder was right about the language. We were wrong. It was Python.
Eighteen months in
Poseidon now operates a fleet of 12 inspection drones across 7 farms in Norway and Scotland. Series A closed Q4 2024 with Andreessen Horowitz leading. The kill-criteria all turned out to be lower bars than we'd feared.
The lesson, the boring one we keep relearning: the studio process is not about the term sheet. It is about the four-or-five conditions you write down at session 06 that, if they fail, you walk away. We have walked away from three deals on those criteria since Poseidon. We have not walked away from one we should have kept.
Three things to take
- Scope reduction is the studio's most valuable single act.
- Write your kill-criteria down. Then test them.
- The founder is right about the things they're right about. Be slower to argue.