Verify
Check a Result yourself
When someone says a Skill made their agent better, you shouldn’t have to take their word for it. Every Result on Techtree is a signed record you can download and check on your own computer, without trusting this site.
Check a Result now →How a Result is made
-
01
The agent works the same tasks twice
An agent run by Nous Research’s Hermes works a Climb’s fixed tasks once without the Skill and once with it. The model, the tools and the limits stay the same.
Hermes on GitHub ↗ -
02
Prime Intellect’s Verifiers scores every task
Prime Intellect’s Verifiers, an open-source library for testing AI agents, gives the agent each task and scores its answer by that task’s own rules.
Read Prime Intellect’s Verifiers ↗ -
03
The record is signed
The tasks, every score, the Skill and the settings go into one signed record: the Result. Changing any of it afterwards breaks the signature.
See published Results →
What a check tells you
Nothing changed after signing
Every part of the Result still matches the signature made when the run finished.
How this is checked →Only the Skill changed
Both runs used the same tasks in the same order, the same model and the same settings. The Skill is the only difference.
How this is checked →The verdict adds up
Better, worse or no different follows from the task scores, by the rule the Climb set before the run.
How this is checked →Nothing private is inside
A Result holds scores, not the agent’s conversations or anything from the computer it ran on.
How this is checked →What a check can’t tell you
- Techtree did not watch the run happen.
- The signature shows which key signed, not that the computer behind it was honest.
- A better score on these tasks doesn’t mean better at everything.
- One Result is one run. Only a separate rerun shows that it repeats.
Why it matters
Skills are easy to claim and hard to compare. A Result anyone can check turns “this Skill helps” into something you can test for yourself, and it stays checkable without this site: all you need is the Result and the free command below.
Check it yourself
Check a Result on your computer.
Download the Result from its page, then run this command. It repeats every check on your own computer, runs no model and costs nothing.
Check a Result
regents techtree proof verify path/to/result-bundleEvery check, in order · 20 checks
- the submission is small enough to be a proof bundle
- the submission is the four-member document and carries nothing else
- the bundle carries no more files than a proof bundle has
- every path is one relative path inside the bundle, written once
- every file is canonical base64, and none of them is empty
- the bundle carries the signed list of its own files, in the current format
- every file hashes to the digest the bundle claims for it
- every signed document hashes to the digest written beside it
- every signature verifies under the key the bundle carries
- the fingerprint names the key by hashing it rather than by claiming it
- the bundle carries the signed result summary it commits to
- the report binds a campaign and execution plan this site publishes
- the wins, losses and ties and the decision recompute from the task scores under the rule the campaign set
- the only setting the two runs differ in is the Skill the campaign lets differ, and the run with it was set up with that Skill
- the bundle carries the Skill it measured, and that Skill holds no secret and is small enough to publish
- the tasks are the list the campaign committed to, in order
- the terms the run was carried out under permit publishing this
- no file in it holds an episode, a transcript, or a path on a machine
- a rerun names a published Result of the same campaign and the same Skill
- what the submission claims about the bundle is what the bundle says
For the full method, read the Docs, copyable as Markdown to your agent.