Practice
30+ curated tasks from easy to hard — loops, strings, arrays, hashing, recursion, sorting, two pointers — with statements in Uzbek, Russian and English.
uz · ru · en
Practice coding. Get hints, not answers.
Solve tasks in Python, JavaScript, C++ or Java, run them against hidden tests in seconds, and ask the AI tutor for a nudge when you're stuck.
solution.pypython3
1n = int(input())2a = list(map(int, input().split()))3best = cur = a[0]4for x in a[1:]:5 cur = max(x, cur + x)6 best = max(best, cur)7print(best)
Judging… Accepted7 / 7 tests38 ms
+25 XP · streak 🔥 4
Hint 2 / 3
110+
tasks with editorials
4
programming languages
10
System Design tasks
3
interface languages
Python 3JavaScriptC++JavaMonacoSandbox1v1ELOXPStreaksGeminiuz · ru · enPython 3JavaScriptC++JavaMonacoSandbox1v1ELOXPStreaksGeminiuz · ru · en
Everything a learner needs
From the first loop to a live 1v1 — one place, three languages.
Instant verdicts
Your code runs against hidden tests in an isolated sandbox: Accepted, Wrong answer, Time limit or Runtime error within seconds, with time and memory.
AI tutor
Three levels of hints that nudge you toward the idea: understand the task, find the approach, unblock your code. It never hands over the solution.
1v1 battles
Find an opponent, get the same task, race to the first Accepted. ELO-style ratings, rematches and spectators.
For schools
Teachers create classes, assign tasks, review submissions with comments and watch a class leaderboard.
Progress that shows
XP, levels, daily streaks and achievements. Editorials unlock after you solve a task — learn the intended idea.
How it works
Four steps, a few minutes each.
1
Pick a task
Filter by difficulty and topic. Every statement comes in your language.
2
Write code
Monaco editor in the browser — the same engine as VS Code, with templates for every language.
3
Get a verdict
Hidden tests run in a sandbox. See exactly which case failed and how fast you ran.
4
Learn the idea
Stuck? Ask the tutor. Solved? Read the editorial and compare approaches.
Race a rival to Accepted
Battles are 1v1: the same task, 15 minutes, first Accepted wins. Ratings move ELO-style, and the tutor sits this one out.
- First Accepted wins
- ELO-style rating
- Live battles
- Rematch in one click
Y +16
You
1000 → 1016
VS
R
Rival
1000 → 984
14:32
Simulationevening peak · 2 000 rps · 85 % reads
Accepted★★★
p95 142 mserrors 0.00 %$1,125/2,400
System Design: build the architecture, not just the code
A new kind of task for those who design systems: assemble Client, Load Balancer, App Servers, Cache, Database, Replicas, Queue and CDN, tune them and run a deterministic load simulation. You get p95/p99, error rate, monthly cost and the bottleneck on the diagram — the same engine judges your submission on the server.
- 1
Assemble blocks
Drag blocks onto the canvas, connect them, set instances, tiers and memory.
- 2
Run scenarios
Normal load, a ×10 spike, a database failure, a cold cache — with a replay second by second.
- 3
Meet the SLO and the budget
Verdicts like a judge: SLO violated, over budget, single point of failure — and stars for saving money.
Studio: schemas and architectures as text — and as a canvas
Describe a database in DBML or an architecture in YAML — Coddle draws the ER diagram or the system graph, lints the schema, keeps versions, computes the diff and the migration, and runs the architecture through the same simulation as the System Design tasks. Import from SQL and Prisma, export to SQL, Prisma and Mermaid.
- 1
Text is the source of truth
Edit DBML or YAML in the editor or move blocks on the canvas — changes flow both ways.
- 2
Versions, diff, migrations
Save versions and get a structural diff and an SQL migration between any two.
- 3
Your agent is already connected
Claude, Codex and other MCP clients read and edit projects from the chat — with your permission and only within the granted scopes.
From the spec the agent designs the database and writes it straight into Studio — the canvas updates live.
Your agentcoddle · MCP
The agent is connected to Coddle over MCP and sees your projects.
Message your agent…
coddle.uz/studio/you/barbershop/schema
savedliveempty document
Lint · 0Versions · 0
The checks appear with the schema.
CanvasSandbox
The canvas is empty — the agent will draw the schema.
0:00 / 0:17
A tutor that nudges, not solves
Hints unlock in order and are limited per day, so you keep thinking. Every hint is logged and reviewed for quality.
Never hands over the solution
1
Level 1 · Understand
Restates the task and points at what to look for in the input.
2
Level 2 · Approach
Names the idea — a hash map, two pointers, a sieve — without the code.
3
Level 3 · Unblock
Looks at your code and says where it goes wrong. Still no solution.
Schools and classes
Classes inside an organization; students join with a code or a link — no admin chores.
Homework after the lesson
Short drills on the topic: a variable already set, a function or input and output — doable even after the first lesson.
AI writes the drills
Describe the topic in words and get drafts with their checks. The reference solution runs before you ever see them.
Who is stuck, and on what
A students × drills table, the code of every try and an AI summary of what to go over next.
Tasks, deadlines, feedback
Tasks from the bank to a class or one student, with a due date; comments on solutions and a class leaderboard.
Templates and an AI agent
Take ready-made sets from colleagues, or run your lessons from Claude and other MCP clients.