Skip to main content
TryCrucible
AI Challenge Platform · Free for candidates

Resumes tell you what someone claims. TryCrucible shows you what they've built.

Pick a real AI challenge, build it, get AI-scored across 6 dimensions, and earn a verified artifact on your public profile.

Try free — no account needed →Browse challenges
RAGAgentsMCPCoding AgentsEvalsAI Tools
// Meeting Action Items · ~10 minNo account needed

Test input — your prompt receives this

MEETING TRANSCRIPT — Product standup, 4 Aug 2026

[Priya]: OK so let's kick off. Rohan, any update on the API rate limiting?

[Rohan]: Yeah so I was looking at it yesterday, ran into some issues with the Redis config actually —
Anil can you check that later today? The config file seems off.

[Priya]: Sure, Anil, can you do that?

[Anil]: Yeah I'll check it. Oh by the way did an
✓ No GitHub repo✓ Scored by GPT-4o

Your system prompt

You are an expert meeting facilitator. Extract every action item from the transcript. For each item output: owner, task, and deadline if mentioned. Only include genuine commitments — not discussion topics. Format as a numbered list.

Sample result82/100

Captured 4 of 5 action items with correct owners and deadlines. Format is clear and scannable. Missed Priya's follow-up with Kavya on the staging deploy.

Try this challenge →

Also try: RAG Relevance Judge →

// pick your starting point
~10 minNo account needed

Prompt Challenge

Write a system prompt in the browser. One test input, scored by GPT-4o in seconds. Perfect if you want to see how TryCrucible works before signing up.

Meeting Action Items →RAG Relevance Judge →
~1 hourIn-browser

Scenario Challenge

One system prompt, three different test inputs. Scored across all three. A real step up — harder than a single prompt but no local setup needed.

Support Email Router →Context-Grounded Q&A →
2–12 hoursBuild locally

Full Challenge

Build a real RAG pipeline, AI agent, or MCP server locally in any language. Submit a GitHub repo. Get a scored artifact permanently on your public profile.

Browse all challenges →
// engineers who've already proved it
R

Rohan Verma

@rohan-verma

91

Production RAG with hybrid search and re-ranking

View profile →
P

Priya Patel

@priya-patel

90

Build an AI agent that triages GitHub issues

View profile →
nagendrawebpipl

nagendra

@nagendrawebpipl

87

Build an AI agent that triages GitHub issues

View profile →

View the full leaderboard →

// How it works

From challenge to verified portfolio

Three steps. Fully evaluated. Permanently yours.

01

Pick a challenge

Browse RAG pipelines, agents, MCP servers, evals, coding agents, and AI tools. Each challenge ships with a real dataset and a scoped LLM key. Or start with a 10-minute prompt challenge in the browser — no setup at all.

02

Build locally & submit

Work in your own environment, in any language. Submit a public GitHub repo and a brief decisions doc. We clone and run your code against real test inputs.

03

Get scored & discovered

AI evaluates across 6 dimensions — correctness, architecture, decision quality, LLM usage, robustness, clarity. Scores above 88 get human expert review. Your verified artifact lives on your public profile permanently, searchable by companies hiring right now.

Scored across 6 real dimensions

Correctness · Architecture · Decision quality · LLM usage · Robustness · Clarity

Scores above 88 get human expert review

// Ready to prove your skills?

Stop claiming.
Start proving.

Free forever for candidates. No resume needed — build something real and let the evaluation speak for itself.

Try a challenge free →Create free account