Our manifesto
Government is how a country takes care of its people. When it works, contracts are awarded fairly, rules make sense, benefits reach the people who need them and emergencies get answered. When it doesn’t, everyone feels it. Our goal is to make it work better.
AI isn’t going away. Agencies, contractors and the labs that serve them are already putting it to work. Ignoring that won’t slow it down. It only means the models learn without the people who know the work best.
Models are fast, tireless and good with words. They don’t know why a source selection gets protested, how a rule survives a challenge or what an inspector general will ask. That judgment lives in the people who did the work.
Neither is enough alone. People can’t keep pace with everything government is asked to do, and models can’t be trusted to do it unchecked. Together they can move faster and get it right.
So we bring the two together. Experts who served in government grade what models get wrong and write out the reasoning they’re missing. The models learn from that reasoning, and the experts are paid for what they know.
That collaboration is the way forward to a government that works better. The people who serve stay at the center, and the tools they use learn from them.
What we hold ourselves to
People at the center
Experts decide what good work looks like. Models learn from them, never the other way round.
Credentials on the record
Every answer comes with the record of the expert who wrote it, so you can see whose judgment you’re relying on.
Fair to the experts
Pay is confirmed before anyone starts, taking part is voluntary, and an employer’s and agency’s rules always apply.
Ethics before
paid work
A current federal employee is never sent paid work unless their agency’s ethics office has approved it in writing. Without that approval they can take part in unpaid work only, and if their agency declines, we send them nothing.
- No override and no exceptions
- Checked again before every send and every submit
- Approval recorded with a date to check it again
Three rules for how we score.
Aegis Eval, the scoring engine underneath Your Roster, follows three rules and is tested against all three. A change that breaks one doesn’t ship.
No magic number
Each dimension gets its own result. We never combine them into one trust score, because how much each one matters depends on your mission, and that’s your call.
Unknown is a valid result
Anything we didn’t test is marked NOT EVALUATED. A run that couldn’t reach a judgment never counts as a pass.
Evidence over claims
Every score links back to the question, the source material, the model’s output, the expert’s reasoning and each grading decision, plus a sha256 fingerprint of the stored record, so any later change shows.
Help build a government that works better.
Join as an expert, or start with a call if you’re building a model.
Join