Gabe Mudel

[[components/metabind-buttons-embed]]

Quick Info

quick-info-display

= this.notes

Reached out regarding the work unblocking (evaluation distribution actions from 2026-03-27)
He said he forgot he was on support but would look into it
Told him no worries and to feel free to prioritize support, no rush on my end

Spent most of the day working together
Morning: investigated why the bump to Oval had degraded the CI pipeline - discovered an upgraded MLflow client connection had upgraded the PostgreSQL DB schema, making it incompatible with the previous version
Wrote up a post-mortem on the incident and had Gabe review it
Met from 2pm through 4:45pm discussing evaluation consolidation
Gabe had questions on why vendoring Miku client
Raised prioritization concerns for evaluations - during his 1:1 with Xiaofan on 2026-03-26, he heard they wouldn’t be prioritizing evaluations, which differed from what I had heard from Pranav
I asked if I could speak with Xiaofan directly to get clarification and Gabe agreed
Discussed Agent Harness, Evaluation Harness, Evaluation Framework, and CI to map out how the consolidated system would look
Showed Gabe how to make the requested changes in Oval
Follow-up: Get Gabe a set of actions he should work on regarding the distribution of evaluations

Gabe reached out to consolidate his framework and mine, and set up time to go over the technicals of Oval
Agreed and set up a meeting for next Friday (2026-03-27)
Sent him a Google Doc with an agenda
Gabe made his first PR to Oval and I accepted it

Spoke at the LE team meeting regarding collecting evaluations when Gabe & Pranav are in London
Gabe mentioned he was unblocked regarding the distribution of evaluation cases
I said this is blocked on my review because it impacts how I operate in Oval
Spoke again during the research sync, where I went over my research proposal
Discussed the distribution of evaluations - stated there shouldn’t be two different products, it should be the one in the multirepo (mine)
Made it clear this is an engineering problem and that I’m not looking to budge

Reached out regarding LLM story time next week where he’ll present AgentEval
Had previously asked me to come onboard to co-present but now felt like he was kindly asking me not to
Told him that’s fine and we can present separately since there’s enough content for both of us