About
Software that finishes the job
Most tools help you work. Helix is built to do the work and show you it was done.
Why we built it
Assistants that only talk leave you holding the whole job: you still plan it, run it, check it and remember what happened. Helix is built the other way round. You state the outcome, and the platform takes responsibility for the plan, the execution and the evidence.
That reframes every design decision. A run needs its own machine, because real work is not safe in yours. It needs to record what it did, because a claim you cannot check is not worth much. And it needs a ceiling you set, because autonomy without a limit is not a feature.
What we hold to
Evidence over confidence
A run that cannot prove its work says so, in the record you read.
Isolation is not optional
Real work needs a real boundary, so every run gets one.
You set the ceiling
Autonomy is a level you choose, and the level is enforced.
The record is the product
If you cannot replay what happened, it did not really happen.
By the numbers
Measured, with the source beside each figure.
Start with one real task
Pick something you would otherwise do yourself. That is the whole test.

