DEVELOPER ASSESSMENT FOR THE AI ERA

Measure how developers think■

AI writes the code now. The question is who knows what to build, how to direct AI, and when AI is wrong. Cagrex measures that.

START ASSESSING ■I'M A DEVELOPER →
CX
CAGREX
BUILDbooking_api
TOKENS2,000 / 2,000██████████100%
SUBMIT ■
EXPLORER
README.md
package.json
index.js
README.md
BUILD A ROOM BOOKING SYSTEM
════════════════════════════════
 
A company needs to manage their
conference rooms.
 
Room A — seats 10
Room B — seats 6
Room C — seats 20
 
People book rooms for meetings.
Build the backend.
 
Server on port 3000.
 
The editor is read-only.
Build everything through AI prompts.
› TERMINAL
AI ASSISTANT
0 TOK
BUILD MODE
UTF-8JAVASCRIPT
01 — HOW IT WORKS

Two phases. One token budget.

Same challenge for everyone. Same AI. Same fixed token budget. We compare who builds the most with the least.

01LOGIC
No AI. Candidate writes code alone. Proves they can think.
AI OFFEDITOR EDITABLE
02AI BUILD
AI enabled. Editor read-only. Fixed token budget visible in real time. The AI mirrors prompt quality — vague in, vague out.
AI ONEDITOR READ-ONLY
02 — WHAT WE MEASURE

Three signals. From one session.

Understanding, token efficiency, and how they work — pulled out of the same ai-build session. Plain English describing behavior counts equally with technical jargon.

UNDERSTANDING — 40%84
Did they describe the important system behaviors — in any words? "Two bookings can't overlap" counts the same as "409 Conflict on collision." The gap between what they described and what the system needs is the measurement.
TOKEN EFFICIENCY — 40%92
Same fixed budget for everyone. Who built more with less. A candidate who understands writes fewer, clearer prompts.
HOW THEY WORK — 20%71
Did they test what AI generated, read the code, diagnose specifically when things broke, notice when AI built the wrong thing?
04 — FOR COMPANIES

Stop guessing. Start measuring.

Create an assessment in minutes. Pick challenges from the library. Send one link to all candidates. Get AI-analyzed scoring reports.

01
CREATE
Pick challenges for each phase. Set time limits. Get an invite link.
02
ASSESS
Candidates work in a real sandbox. Every interaction is logged.
03
ANALYZE
AI scores three signals — understanding, token efficiency, how they work. Evidence quoted from the chat log.
03 — FOR DEVELOPERS

Not just an assessment tool.

Create an account. Practice challenges. Track understanding, token efficiency, and how you work. Know where you're strong and where you need work.

PRACTICE
Two challenge types: logic and AI build. Practice independently with scores visible to you.
TRACK
Three-signal skill profile from your practice sessions. Understanding, token efficiency, and how you work.
PROVE
When a company sends you an assessment, your account links to the session. Build your track record over time.

The future belongs to thinkers.

AI will replace skills. It will never replace brains.

START ASSESSING ■PRACTICE FOR FREE →
CX
CAGREX
STRUCTURE OVER CHAOS ■