MEE-ahts · the school of engineering judgment

AI can write the code. Can it judge the code?

Miatz is a selective engineering school. Pass the DSAT gate, do real AI-agent work from week one, and build an auditable record of your decisions — the proof that you can test, orchestrate and command what AI builds. We demystify. You leave with miatz.

Admission is selective · a Demystify Systems company

Mysty · shadow engineer
My tests pass locally but fail in CI. Why?
Let's diff the environments — Node version, timezone, and unseeded DB state are the usual three. Run this and paste the output…
Cohort LeaderboardWeek 3
🥇 Ada L.428 XP
🥈 Grace H.391 XP
🥉 Alan T.355 XP
ENGINEERING JUDGMENTAI-AGENT WORKLLMs & RAGAGENTSPROMPT ENGINEERINGINCIDENT RESPONSETESTING & QAORCHESTRATIONAUDITABLE RECORDTHE DSAT GATESYSTEM DESIGNTHE ITZ'ATENGINEERING JUDGMENTAI-AGENT WORKLLMs & RAGAGENTSPROMPT ENGINEERINGINCIDENT RESPONSETESTING & QAORCHESTRATIONAUDITABLE RECORDTHE DSAT GATESYSTEM DESIGNTHE ITZ'AT
How Miatz builds judgment

We train the thing machines can't do

Not videos and quizzes — real AI-agent work, live incidents, spaced recall and an auditable record. Every surface is built to grow and measure judgment.

Command AI, don't just prompt it

The AI-engineering track — LLMs, prompt engineering, agents, RAG and evals — teaches you to test, orchestrate and command AI systems, not recall syntax.

Prompt Engineering Lab

Write a prompt, watch it run, and get scored on COSTAR and canonical technique — with an optimized rewrite and rationale every time.

Spaced-repetition recall

An SM-2 retrieval engine cements HTTP, SQL, networking, auth, caching and system design so the fundamentals actually stick.

AI simulations & spoken skills

Real-time role-play — skeptical CTOs, angry customers, exec briefings — with voice input and delivery scoring. Communication made measurable.

Judgment, gated and earned

Levels are earned, not given: the DSAT, solved challenges, verified skills and incident scores must clear the bar before you advance.

Applied, not textbook

Every day, week and month is framed by how real systems work and the cost of ignorance — anchored to 35 real documented failures.

360° skill wheel + specializations

A radial competency map across 28 disciplines, plus deep-dive tracks (AI eng, SRE, security, TPM…) with the best references in the world.

Real bug-hunting on live apps

Test real products, file evidence-backed bug reports, and read each app's live schema, RLS and architecture graph.

An auditable track record

Not a certificate — a cryptographically signed, employer-verifiable record of the decisions you made. Proof you can think, in one click.

The Miatz journey

Foundations → Core → Immersion

The DSAT places you. Foundations (the Sprint) builds the base; Core puts you on real AI-agent work; Immersion retains top performers and routes them to paid work. Every level is earned, every decision logged — you leave with the Itz'at, not a certificate.

The DSAT gate
Timed, integrity-monitored — reasoning over syntax.
Coding + AI labs
Hidden-test challenges, prompt lab, agents & RAG.
War rooms
Real incidents graded like a staff SRE.
Operate Mysty
Run a live AI agent on real work — propose & approve.
Never stop learning

Bytes of real engineering, every day

Curated knowledge and real war stories — the cost of getting fundamentals wrong, and how the best get them right.

Cost of ignoranceData

A missing WHERE clause deleted production

In 2017 a GitLab engineer ran a removal against the primary instead of a replica — and backups had silently been failing.

The cost: ~6 hours down and hours of user data lost. Test your restores, not just your backups.

Anyone can learn to code. Few learn to judge.

Take the DSAT and find out if you're one of the few. Admission is selective — a Demystify Systems company.