muxify

Technical assessment, rebuilt

Hire engineers who ship with AI.

Replace coding puzzles with real build sessions. Watch how candidates think, decide, and direct the agent.

v24.14.1 ~/sessions/take-home-chat git:(main)

session active

 ▐▛███▜▌
▝▜█████▛▘
  ▘▘ ▝▝

Claude Code v2.1.132

Opus 4.7 (1M context) with xhigh effort · Claude Max

~/sessions/take-home-chat

❯

~/sessions/take-home-chat (main)[Opus 4.7 (1M context)]8% used / 92% left

Found 2 settings issues · /doctor for details

✻+± 0▭ /remote-control▤ File explorer↳ Rich Input ^G

▢ ~/sessions/take-home-chat⎇ main

Why now

The job changed. The interview didn't.

What every legacy platform still tests for, against what the work actually demands today.

hiring/skills.md

$git difflegacy..muxify

+4/−4

--- a/skills.md

+++ b/skills.md

@@ −1,4 +1,4 @@

Solve puzzles alone. No internet, no tools, no AI.

45 minutes of algorithmic recall under time pressure.

Measured on syntax fluency and trivia.

Predicts on-job performance? No.

Ship a feature on a real 40k-LOC repo.

Full session captured for review.

Measured on decisions, problem decomposition, agent control.

Predicts on-job performance? Yes.

1 file changed, 4 insertions(+), 4deletions(−)press q to quit

For engineering leaders

How hiring works on Muxify.

From setup to a hiring decision in three steps. Your repo, your stack, the AI tools your team uses.

configure

Pick a repo. Pick a problem.

Connect your GitHub repo, set the time limit, and you're live in two minutes.

repositoryacme-co/checkout-api ●

stackTypeScriptPostgres+2

levelL3L4L5L6

time box60m

promptAdd idempotency keys to /charge

invite

Send a single magic link.

One link, zero setup. Candidates work in a cloud terminal with their AI tool of choice, exactly how they would on the job.

muxify.tech/s/a8f-92k1-7dcopy

expires in 7 days·single use

RKRiya K.opened · 2m ago

TMTheo M.submitted · yesterday

JLJamie L.invited · 3 days ago

review

Watch the session, not just the score.

See every prompt, edit, and decision in a full session replay. Read the final PR and hire with confidence.

replay · candidate-782100:42:18 / 01:00:00

14:02opened README.md

22:41prompt → Claude

33:08edit client.ts +24 −2

56:11ran tests · 9 / 9 passing

For candidates

What the candidate actually sees.

No webcam. No invigilator. No memorized tricks. Just a repo, a terminal, and the AI tools you already use.

before

Algorithm grinder

— Algorithm puzzles under a stopwatch
— No internet, no AI, none of your tools
— A pass/fail score that misses what matters
— The same question bank every platform asks

on muxify

Real build session

+ Open the repo, ship like you do at work
+ Use Claude Code, Codex, or OpenCode. Your choice
+ 60 minutes async, on your own schedule
+ Reviewed on the work, not a number

Session replay

Watch the work, not the résumé.

Every terminal command, file edit, and AI prompt is recorded. Scrub the timeline, comment inline, and share with your hiring panel.

Muxifysession activecandidate-7821 · take-home-chat

00:42:18Share replay

Files

src
App.tsx
components
MessageList.tsx
Composer.tsx
lib
package.json
README.md

src/components/MessageList.tsx

1import { useEffect, useState } from "react";
2import { fetchMessages } from "../lib/client";
3
4export function MessageList() {
5  const [items, setItems] = useState([]);
6  useEffect(() => {
7    fetchMessages().then(setItems);
8  }, []);
9  return <ul>{items.map(...)}</ul>;

~/sessions/take-home-chat git:(main)

claude-code

 ▐▛███▜▌
▝▜█████▛▘
  ▘▘ ▝▝

Claude Code v2.1.132

Opus 4.7 · ~/sessions/take-home-chat

❯Wire MessageList to /api/messages with optimistic updates.

·Reading src/components/MessageList.tsx

⏺Adding fetch + optimistic queue. Two edits.

·Edit src/components/MessageList.tsx +24 −2

✻Brewed for 4s

❯—

~/sessions/take-home-chat (main)[Opus 4.7]32% / 68%

2 edits · /history

✻+± 2▭ /remote-control▤ Files

⎇ main

Timeline00:00:00 — 01:00:00

Replay length00:42:18·Files changed7·Diff+148 −22·Agentclaude-code·Tests9 / 9 passing

The assessment library

Real problems, sourced from real products.

Every problem is sized so a senior engineer hand-coding for the full hour would not finish. The point is to see how someone directs an agent under realistic pressure.

build60m

Multi-tenant analytics dashboard

Wire a tenant-scoped /analytics page with four chart types, a date-range filter, the API routes, the SQL, and tenant-isolation tests into a 40k-line SaaS starter.

integrate60m

Tool-calling for an existing chatbot

Extend a production chat app with three internal tools (search_users, create_ticket, fetch_user). Schema, agent loop, retries, safety guards, tests.

debug45m

Regression across the checkout flow

A refactor broke checkout in a 50k-line monorepo. Trace the bug across the frontend hook, the API client, the server endpoint, and the DB query. Ship the fix plus regression tests.

vs. legacy assessment platforms

The same problem, solved differently.

LeetCode tests for a world without AI. Muxify tests for the one your team actually works in.

Muxify

Legacy platforms

What gets measured

Shipping working code on a real repo

Algorithm puzzle pass/fail

Tools the candidate uses

Real terminal with their AI tool of choice

Pen, paper, no internet

Session length

30–90 minutes

45 minutes of stress

What you review

Full session replay + final PR

A score and a leaderboard

AI usage

Encouraged. The whole point.

Strictly forbidden

Signal for senior ICs

Direct. They did the job.

Indirect. A proxy for tenacity.

A note from the founder

We hire engineers who use AI every day. The interview never caught up.

Every engineer on a serious team now ships with Claude Code or Codex daily. The "no internet, no AI, solve this puzzle alone" interview measures a job no one actually does anymore.

When the test does not match the job, the signal you hire on is the wrong signal. Strong builders look mediocre on algorithm puzzles. Mediocre puzzle-solvers look strong. Hires miss either way, and the team pays for it for two years.

Muxify is the assessment for the job your team actually hires for. Real repo, real AI, real session, real signal. We built it because we wanted it for our own loops first.

Tod·founder

Hiring stack

Built for the tools your team already uses.

GitHub

Slack

Linear

Greenhouse

Notion

Loom

Pricing

Flat pricing. AI compute included.

One tier covers most hiring teams. Custom plans for higher volume or specific procurement requirements.

Starter

$1,000/ month

Up to 20 candidate assessments / month
White glove onboarding
Pre-built challenge library + custom challenges we help design
AI-powered session replay, evaluation, scoring
All AI usage included

Request Demo

Enterprise

Custom

Custom candidate assessments per month
Bring your own AI provider keys
SSO and audit logs
ATS and third-party integrations
Dedicated point of contact

Contact Sales

Questions

The questions engineering leaders ask first.

Is using AI during the session allowed?

Yes, that's the whole point. Muxify shows you how candidates work with AI, not whether they can avoid it.

What happens to the candidate's code?

It stays in the candidate's session repo. You get the final state, the diff, and the full replay. Export to a private GitHub repo your team owns, or discard after a retention window you set.

What signal do you actually get from a session?

You see how a candidate reasons through ambiguity, where they get stuck, how they direct the agent, and what they choose not to do. The journey is much harder to fake than a finished commit.

What languages and stacks are supported?

TypeScript, Python, Go, Rust, Ruby, Java, and their surrounding tooling. The cloud terminal is a full Linux environment, so anything that runs on Muxify.

How long does a typical assessment take to set up?

Two minutes. Connect a GitHub repo, write a one-paragraph prompt, and set a time limit. Or start from our library of pre-built problems.

Will Muxify fit into our hiring stack?

We're constantly rolling out integrations. If there's an integration you want, lets talk.

request

Get started

Watch a real session before you commit.

Tell us about your team. We will send a short walkthrough and book time if you want it.

30-minute live demo with the founders
Two sample sessions tailored to your stack
No pitch deck, no sales floor