AI should answer to the people who use it.
Nudge brings accountability and transparency to AI work: every claim checked, every cost in the open, and the right model for each job, whichever company made it.
Why we built it
We build software with AI agents every day, across many products at once. The work got faster, and the surprises grew with it: work reported as done that was not, costs that crept up without anyone choosing them, and no plain record of what had actually happened.
AI tools are powerful, but they are not yet answerable. Nobody checks their claims, few people see what they cost, and the choice of model is usually made for you. Nudge exists to change that, one practical step at a time.
What accountability means in Nudge
-
Claims are checked
"Done" is not taken on trust. Nudge runs the test itself, sends failed work back, and hands the job to Claude when it fails twice. Work that failed its test never reaches you.
Live today -
Everything is in the open
A plain record of what each agent did, what passed, what failed and what it cost. What Nudge itself collects is written down, and you can see it before it is sent.
Live today -
The best model for the job
Advice comes from measured results on real work, not from one company's catalogue. You are never tied to one provider's models because that is where the tools happen to point.
Live today -
People stay in control
Every switch is yours. Nudge advises, checks and records; it never takes a decision away from you, and it keeps the work's history with you, not with a vendor.
Next: continuity, resourcing, proof and feedback
Why this has to come from outside the AI companies
Each AI company builds tools that work best with its own models, and that is natural. It also means accountability will not come naturally to them: none has a reason to tell you when another company's model would do a job better or for less, to check its own work independently, or to make leaving easy.
Checking results across providers, choosing the best model for each job and keeping your speed, cost and outcomes at the centre all need someone who sits on your side of the table and works with every model. That is the place Nudge takes.
What we promise
- We work for you
- No model provider pays us, and none can buy a better place in Nudge's advice. We do not earn more when you spend more.
- Measured, not claimed
- Every figure we show comes from real work, with the method written down, so you can check it.
- Open to inspection
- Nudge is open source. Your code stays on your machine, and what we collect is written down plainly.
- Your choice, always
- Turn any part off with one word, and see exactly what is sent before it goes.
Where we start
Checked delegation is the first part. Each part that follows will save cost, improve outcomes, remove bottlenecks or make the work more transparent, and each will be judged by the same measure: does it keep AI answerable to the people who use it?