Labelbox

Frontier AI labs train on Labelbox environments and expert data.

Public description

Company overview

Frontier AI labs train on Labelbox environments and expert data. Engineering and research teams hand their recurring work to agents on Recursion. Labelbox builds the environments frontier labs train AI on, and the agents engineering and research teams hand their backlog to. Recursion isn’t another assistant in your editor. Its agents start on their own, on a red build, an alert, a new checkpoint or a schedule, and work in the background until the job is done: every flaky test fixed, every dependency kept current, every eval rerun, dozens of experiments compared at once. Here are a few of the jobs teams hand it. More Ticket moved to Ready Linear GitHub Vercel Slack Delivers: A tested pull request with a preview link, ready for review Software engineering Feature Builder Builds each ready ticket into a working feature, with tests, in its own sandbox. New issue in Sentry Sentry GitHub Linear Delivers: A pull request with the fix and a test that proves it Software engineering Bug Fixer Reproduces each new bug, finds the cause and writes the fix with a regression test. New package version released GitHub Jenkins Slack Delivers: One ready-to-merge pull request per package, tests green Software engineering Dependency Upgrader Upgrades dependencies as new versions ship, fixes what breaks and reruns every test. CI failure on main GitHub Jenkins Slack Delivers: A pull request with the fix and 50 green runs in a row Software engineering Flaky Test Fixer Reproduces each flaky test, finds Website: https://labelbox.com/ Source: Company site

Company

Company overview

Labelbox builds the RL environments the world’s leading AI labs train in, and Recursion, the platform enterprises use to run fleets of agents on their hardest work. We are a technology company, founded in 2018 and profitable, built on a conviction most of the industry has not caught up to yet: the models that matter will be the ones that learn by doing the work. Labelbox began in 2018 as tools for machine learning engineers, built by people who had hit the wall themselves. It became the platform teams used to build their training data, and it taught us early that the winners in AI would be the ones who took the data problem seriously. Horizon builds the environments and evals frontier labs post-train in: simulated software, expert-written graders, tasks with real stakes. Models stop being told the answer and start practicing the work. Recursion is our platform for enterprises: give a fleet of agents a goal and they do the work across your systems. Every run is scored, and the best runs train specialist models that beat general-purpose AI on your tasks. Backed by investors who fund category-defining infrastructure: Kleiner Perkins, Andr

  • Labelbox

    Company · Subject