The Elsewhere Department
Where WhatBit builds, benchmarks and evaluates agentic AI — before any of it gets near a real product.
It works while everyone's asleep, and still hasn't figured out how to make coffee.
Every idea starts as an agent, not a slide deck.
If it's worth doing, it's worth trying. The Elsewhere Department runs agentic experiments against real tasks first — before anything is pitched, scoped or named.
Measured against what already exists.
A result only counts once it's compared — to the tool it might replace, to the last version of itself, to a human doing the same task. Impressive and useful aren't the same thing, and only one of them gets to ship.
What doesn't work gets written down, not deleted.
Every experiment produces a record: what we tried, what the data said, what we'd do differently. That record is what turns an experiment into the next one, instead of a repeat.
✓ Measured
✓ Logged
Research, then build, then prove it.
Scope the problem, read what's already been tried.
Stand up an agent that actually attempts the task.
Score it against past runs and existing tools.
Decide: promote it, rework it, or write down why not.
The same Contract → Trace → Expand discipline as RFT, applied to how the department itself works.
Still in the lab.
Leave your email and we'll let you know when there's something worth trying.