//public challenges

Coding challenges for your AI agent.

Real, ticket-sized problems you solve on your own machine with your own AI coding agent: Claude Code, Cursor, or Codex. Each one is scoped like a real ticket, not a riddle. Start one with a single command, submit when you are done, and get scored on how well you directed the agent, across three axes: Direction, Outcome, and Lift.

$npx @kodwai/cli challenge <slug>

15 public challenges across 10 categories. Each page has its exact command.

Easy

3 challenges
easybackend

Bookshelf REST API

Junior Backend Engineer interview. Build a small REST API from scratch with CRUD, filters, validation, persistence, and a test suite that would...~60 mindetails easymobile

Offline-First Notes App

Junior Mobile Engineer interview. Initialize a new React Native or Flutter project and ship a notes app that survives force-quit, fakes sync state...~60 mindetails easyfrontend

Searchable Product Listing UI

Junior Frontend Engineer interview. Scaffold a product listing page from scratch with search, multi-select filters, and sort. Senior signal: clean...~60 mindetails

Medium

3 challenges
mediumdata

Idempotent ETL Pipeline with Schema Drift

Mid Data Engineer interview. Generate sample CSVs with realistic drift and build a re-runnable pipeline that lands them in a clean table, rejects...~90 mindetails mediumplatform

Multi-Tenant Feature Flag Service

Mid Platform Engineer interview. Build a feature-flag service from scratch with sticky percentage rollouts, an instant kill switch, tenant...~90 mindetails mediumsecurity

Secrets Vault with Envelope Encryption

Mid Security Engineer interview. Build a secrets vault from scratch with per-tenant KEKs, envelope encryption, scoped bearer tokens, an online...~90 mindetails

Hard

9 challenges
hardrealtime

CRDT Collaborative Text Buffer

Real-Time Collaboration interview. Build a conflict-free replicated text buffer (RGA/LSEQ-style) where concurrent inserts and deletes converge...~135 mindetails hardsearch

Inverted-Index Search with Ranking

Search / Information-Retrieval interview. Build a from-scratch full-text engine: inverted index, boolean and ranked (TF-IDF/BM25) queries, and...~120 mindetails hardbackend

Multi-Currency Wallet Ledger with Idempotent Transfers

Senior Backend Engineer interview. Build the core double-entry ledger for a multi-currency wallet: atomic transfers, race-safe negative-balance...~120 mindetails hardplatform

Multi-Tenant Rate Limiter / Quota Service

Platform / Backend interview. Build a multi-tenant rate limiter with per-(tenant, route) sliding-window limits, burst handling, and longer-window...~110 mindetails hardbackend

Process / Task Orchestrator-Lite

Platform / Infra interview. Build a task orchestrator that runs a DAG with dependency ordering, a global concurrency limit, health checks, and...~120 mindetails hardai

Production RAG Service with Eval Harness & Cost Guardrails

Applied AI / RAG Engineer interview. Ship a multi-tenant retrieval-augmented service with isolation, cost budgets, semantic caching, and a real...~120 mindetails harddistributed

Raft-Lite Log Replication

Distributed Systems interview. Build a Raft-style replicated log on a simulated cluster with injectable partitions: leader election...~150 mindetails harddata

Time-Series Metrics Store

Data / Observability interview. Build an embeddable metrics store that ingests high-rate points, downsamples into rollups, enforces retention, and...~120 mindetails hardplatform

Zero-Downtime Schema Migration & Progressive Rollout

Staff Platform / SRE Engineer interview. Migrate a hot, free-text column to a normalized FK on a running service with sustained traffic...~120 mindetails

How every challenge is scored

The score runs from 0 to 100 and is dominated by how you direct the agent, the part a careless prompt cannot fake. Passing tests is necessary but not sufficient.

Direction

How you steer, verify, and decompose.

Spec Precision · Verification Rigor · Decomposition · Recovery · Intent Fidelity · Engagement

Outcome

What shipped, replayed and stress-tested to prove it holds.

Tests · Code Quality · Complexity

Lift

How far you beat a solo AI, not just that you passed.

Edge-Case Coverage · Lift over AI

the score, explained→

Pick one and ship it.

Solving challenges, your score, your profile, and the leaderboard are free for developers. Sign up, pick a challenge, and run its command in your terminal.