Skip to content

Say hello

hello@radicandlab.co

An AI products company

We build AI products from the root.

Radicand Labs builds its own AI products for the people who build software. Our first, Afara, gives every coding agent the context of every team it builds on.

Meet Afara Join us
Scroll
Start at the rootPrivate by designMeasured, not vibedInside the workflowStable over cleverPlain wordsStart at the rootPrivate by designMeasured, not vibedInside the workflowStable over cleverPlain words
AfaraCoding agentsContextEvaluationPrivacyDeveloper toolsAfaraCoding agentsContextEvaluationPrivacyDeveloper tools

(01)Our name

rad·i·cand

/ˈradɪˌkand/ · noun

The number or expression under a radical sign. The part you work through before you get an answer.

Most AI products start with a model and go looking for a problem. We work the other way round: we start with a problem we've felt ourselves, work through what's underneath, and only then ask what a model can do about it. Then we measure it, and we use it ourselves before anyone else does.

(02)Products

Tools for the age of coding agents.

All products

BetaDeveloper tools

Afara

Every team's context, in every coding agent.

Afara turns each team's code and tickets into context, and injects it, with your company's enterprise and service patterns, into every coding agent that builds on it.

Visit afara.dev How it works
~/checkout-service
$ afara compare --ticket PAY-142
Reading commits on feature/refunds…
✓ Wireframe generated: 14 nodes, 21 edges
✓ Linked to PAY-142 "Partial refunds"
! Missing: refund reason is required by the ticket
! Extra: refunds over 30 days are blocked in code

$ afara push
✓ Context published to 3 teams

In the lab

Product two is in development.

We'll share it when it passes its own evaluation.

Build it with us

We're hiring engineers, researchers and designers.

(03)Principles

What every product we ship holds to.

Six rules we wrote down before we wrote a line of Afara, and still check every release against.

(04)How we build

From problem to product, in five steps.

  1. 01

    Problem

    Find the root

    We start with a problem we've felt ourselves, more than once, and write it down in one sentence before anyone writes a prompt.

  2. 02

    Spec

    Write the eval

    Real examples with the answers we would accept. The evaluation suite is the spec, and it exists before the feature does.

  3. 03

    Build

    Prototype

    The smallest thing that could solve it, tried against the eval and in front of the engineers who have the problem.

  4. 04

    Use

    Dogfood

    We use it on our own work, every day. Afara runs on Afara. If it annoys us, it will annoy you.

  5. 05

    Release

    Ship

    Signed releases, plain-word docs and a changelog. Then every model change has to pass the eval to ship again.

  6. Then

    Repeat, for the next problem.

    More about how we work

Radicand Labs in numbers

(05)Blog

Notes from the lab.

All posts

We're hiring · 5 open roles

Want to build AI products? Build ours.

We're hiringBuild with usJoin RadicandWe're hiringBuild with usJoin Radicand

Help us build what comes next.

We're a small team building AI products for the people who build software. If that sounds like your kind of problem, we'd like to meet you.