Application
Tell us who you are. Name, track, a short note, and two written answers. Then, if you want, pick a task for one question.
Full name Email
Track SoftwareOct–Nov 2026 October–November 2026. The first cohort — a realistic software team. DataEarly application Early application for January 2027. Investigate sources and deliver findings a team can use. Machine LearningEarly application Early application for January 2027. Train, evaluate, and ship an ML workflow. Project ManagerEarly application Early application for January 2027. Scope, coordinate, and keep a team shipping. Marketing & SalesApplications open Applications open. Find the audience, write the line, pick who to call, and get a real product in front of people.
City / timezone GitHub or LinkedIn Where you are nowSelect… Student Early career Employed, looking to switch Looking for work Other Hours this monthSelect… 5–10 hours / week 10–15 hours / week 15+ hours / week
What work do you want to do, and what have you already shipped? Optional
Pick a task for one question, or skip.
You inherited a small web-GPT. It never shows a real page. It turns the window into a pile of HTML tags, asks a person what they want, and suggests the next click. A headed Chromium starts before anyone has said the objective. Clicks fire at the last seen box. Every suggestion waits for a human approve, with no replay and no score.
Coding Engineer Data ML Marketing ML The model suggests the next click. Holdout looks fine. Production does not.
Did the next-click model get better?
Split Click hit Nav success p95 Notes Holdout 0.74 0.61 1.8s Last week's model Prod week 0.61 0.48 4.1s Pages shifted after crawl
Would you ship this week’s model, wait, or roll back — and which number decides it?