Small teams.Long problems.
Projects run for months, not sprints. We protect deep work: no meetings before noon, ever.
If it is safe to share, we share it. Everyone here has their name on a paper or a release.
Most teams are three to six people with one clear owner and no layers in between.
Two offices and eleven time zones. We write things down so nobody has to be awake at 3 a.m.
Every researcher gets a guaranteed GPU budget. No queueing for weeks to test an idea.
Salaries are published on every role. Equity for everyone, and 26 days off plus local holidays.
Open roles
Eight seats at the table.
Research scientist, reasoning
Research
London
£160k–£230k + equity
Research engineer, evaluation
Research
Montréal
C$170k–C$240k + equity
Inference engineer
Infrastructure
Remote, Europe
€140k–€200k + equity
Member of technical staff, safety
Safety
London
£150k–£220k + equity
Developer relations lead
Product
Remote, North America
$170k–$220k + equity
Product designer, API and console
Product
London
£110k–£150k + equity
GPU cluster engineer
Infrastructure
Montréal
C$150k–C$210k + equity
Recruiting partner, research
Operations
London
£80k–£110k + equity
Three years,one question.
Four researchers leave a large lab to work on one problem: models that check their own work.
'Search before you speak' shows small models beating large ones when given time to think.
Private preview with 40 research groups. Median task length grows from 20 s to 6 min.
Public launch with a 142-page system card and an open 8B distill.
Longer horizons: models that work on a problem for a day, and show every step.