autocollab

we store the whole session but with tools to query the summarized choices. It's not intended for the user to sit there and read through the whole session. The session acts as metadata.

classifiers are determined by the team that's working on the repo. Not by us! They're configuration files..like how in CI/CD you write a workflow file that needs to have passing tests. Same thing.

Take a look at his keynote at WeAreDevelopers..he starts around the 17:20 mark

https://www.youtube.com/watch?v=eSoHzgrhk6Q

He does hint around removing away from pull requests!

Hi! I work at Entire. We actually have another feature that we announced called Trails which indicates Confidence, Risk, Drift of code changes to give the maintainer more confidence that the code is high quality.

Each of those signals -- confidence, risk, drift are determined by a workflow file that the maintainer can create.

And trails continuously evaluate every time a change is made.

So that's exactly what we're doing..a gate that checks and weighs and measure everything before it gets committed.

The point of us capturing the agent sessions is so we can do more things with it..like adding "gates" that verify quality!

An interesting approach for managing agent traffic. I'm looking forward to the "the new developer lifecycle optimized for intent based review" that they mentioned