AI gives us the hypotheses. zeb labs is how we chase the ones worth chasing.
Build small.
Scale is a problem for later.
We start with a toy model and let the idea evolve, unconstrained by production limits.
How zeb labs works
- 01Start with a toy model
- 02Treat scale as a problem for later
- 03Don't limit innovation with production constraints
- 04Let it evolve and change
- 05…then make it conducive to applied use
Production constraints come in only when it ships, never before.
Substrate is the playground.
01
Curates R&D
Hands us hypotheses02
Tests weird ideas
Modular enough to run anything03
Parallelized learning
Learns from every experience04
Proactive applicability
Shows up before the askA bi-directional
research loop.
Curate
Substrate surfaces hypotheses and unexplored areas, the questions worth chasing.
Experiment
We take an idea and move fast: a toy model first, scale as a problem for later.
Compound
Every experiment feeds Substrate. The learning compounds and shows up proactively in engagements.
Working on problems
adjacent to ours?
Every zeb engineer rotates through zeb labs. We're open to collaborations.