// AFK

Away From Keyboard building for owned projects with AI.

AFK (Away From Keyboard) is AlgoArena's early agentic builder for turning ideas into repos, exports, and code you actually own.

Join the AFK waitlist Read our vision

The near-term contract is ownership: real files, clear status, export paths, and eventually GitHub/deploy handoff.

https://algoarena.net/afk/workspace

AFK conceptconcept preview

Rena AFK

chat to app workspace

You

Build a tiny web app where students track daily algorithm practice and see streaks.

Generating first pass

Scaffold Next.js dashboard

Add local progress storage

Generate streak card and goal list

Run smoke check

arena-practice-app0 of 4

app/page.tsx+84

app/api/progress/route.ts+31

components/GoalCard.tsx+42

lib/storage.ts+18

npm run check

lint passed

preview ready

app/page.tsx

1export default function PracticeTracker() {

2 const goals = ["arrays", "graphs", "dp"];

4 return <main className="grid gap-4">

5 <StreakCard days={12} />

6 <GoalList goals={goals} />

7 <RunSummary accepted={8} />

8 </main>;

Browser previewbuilding

AFK workspace

A real workspace for agentic project building.

Build from a browser workspace with real project files, exportable output, and evidence from AlgoArena's assessment layer.

Workspace

Build in a real browser workspace.

Real project files, real environments, explicit status, and a clean path to owned code.

Join the AFK waitlist

Owned artifacts

The output should be yours.

Export repos you own, with clear status on builds, tests, and deploy handoff.

Join the AFK waitlist

Assessments first

Reuse the evidence layer.

Controlled files, AI modes, tests, prompt traces, and verification from Assessments, extended for project building.

Preview Assessments

AFK conceptconcept preview

Rena AFK

chat to app workspace

You

Build a tiny web app where students track daily algorithm practice and see streaks.

Generating first pass

Scaffold Next.js dashboard

Add local progress storage

Generate streak card and goal list

Run smoke check

arena-practice-app0 of 4

app/page.tsx+84

app/api/progress/route.ts+31

components/GoalCard.tsx+42

lib/storage.ts+18

npm run check

lint passed

preview ready

app/page.tsx

1export default function PracticeTracker() {

2 const goals = ["arrays", "graphs", "dp"];

4 return <main className="grid gap-4">

5 <StreakCard days={12} />

6 <GoalList goals={goals} />

7 <RunSummary accepted={8} />

8 </main>;

Browser previewbuilding

Assessment bridge

Built on the same evidence layer as Assessments.

Candidate workspaces already need controlled files, AI-mode policy, tests, replayable evidence, and recruiter review. AFK extends that foundation for owned projects.

Preview Assessments

Candidate workflow

Product Card - Design to Spec

Rena mode

Candidates work in one controlled browser tab with tests, AI mode controls, and replayable evidence.

No AI

Rena

Agentic

$ npm test

product-card.spec.ts

3 passing

1 visual check queued

Evidence loop

Question locked

Rena hint requested

Candidate edits code

Tests rerun

AI policy

Same model access, visible prompts, and clear human ownership.

Learning bridge

Plain-English builders need a place to ship.

Vibecoding Courses can graduate learners from commands into real project creation once the AFK workspace is ready to own files, tests, and deploy handoff.

Preview Vibecoding Courses

Code bridgeoptional view

Plain English

for every green ball, move it to the green box

for ball in world.balls:
if ball.color == "green":
rena.pick_up(ball)
rena.move_to(green_box)
rena.drop()

Recent from the field notes

All field notes

Vibecoding2026-07-17

An Agent Editor Built to Be Read Twice

Inside the agent editor that powers AlgoArena vibecoding assessments: why a turn renders as a story, why every diff is a decision, and why each convenience also has to leave evidence a reviewer can trust.

Read

Classroom2026-07-17

An Educator Assistant That Shows Its Work

The classroom dashboard now has an assistant that drafts questions and sets up assignments, grading suggestions arrive backed by checks the platform re-runs itself, and every judgment call stays with the educator.

Read

Assessments2026-07-16

How to Assess Developers Who Code With AI

The old assessment measured whether you could write code unaided. That signal is collapsing. How to measure judgment when a candidate builds with AI in the loop.

Read

Graphs2026-07-14

DFS vs BFS: Which Graph Traversal Should You Use?

A direct comparison of depth-first and breadth-first search: traversal order, memory, shortest paths on unweighted graphs, and cycle detection, with a table and code.

Read

Early access

Turn an idea into a real, runnable project.

A zero-config AI workspace that outputs owned projects, not screenshots or locked-in artifacts.

Join the AFK waitlist