GERRA

Gerra / Case studies

How the data
gets built.

Engagements written up with the mechanism, the measurements, and the parts that went wrong. Buyers are not named; the numbers and the methods are real.

  1. 01
    A frontier AI labCoding-agent training and evaluationAug 2026

    Contamination-proof coding environments at 300-600 accepted tasks per week

    A sustained supply line of graded coding environments for RL training, where each task ships with executable proof that a model cannot pass it by reproducing a library it saw in pretraining.

    Accepted Tasks
    300-600 / wk
    Held-Out Assertions
    1,308
    Plagiarist Gate
    8 of 8
  2. 02
    A quantitative trading deskSystematic equity researchMay 2026

    Testing a compute panel for equity alpha, and publishing the null

    A four-year cross-provider GPU rental panel built to test whether compute input costs predict the equity returns of the companies buying them. The directional result was null, and it was published.

    Study Panel
    6,216 obs
    Providers Covered
    38
    Directional Alpha
    None
  3. 03
    A systematic equity fundAlternative data for L/S equityMay 2026

    Four integrity tests a sentiment panel has to pass before a backtest means anything

    A fund that had already rejected two sentiment vendors ran a four-point integrity audit against eighteen years of human-labeled retail sentiment before licensing any of it.

    Messages
    426M+
    History
    18 years
    Label Source
    Human