Every miatz playground has one job: make hidden machinery visible. Each one carries a miatz signature — a reveal no other tool shows — and ties into a module of the curriculum.
Straight from the catalog — no account, no email.
Watch your words become the tokens a model actually sees.
Play nowSlice a document the way RAG systems do — and see why they miss.
Play nowDissect a real prompt with COSTAR, piece by piece.
Play nowWhen you say 90% sure — are you? Your confidence-vs-accuracy curve, in five minutes.
Play nowWalk a scripted agent run and make the Delegate / Accept / Deploy calls — then see the consequences replay.
Play now16 playgrounds
Watch the planner-executor loop light up, node by node.
Watch the coding agent's plan, diffs, and blast radius live.
Bury a fact in context and watch the model forget it.
Watch a demo's pennies turn into a production cost cliff.
Where the code grader and the AI judge disagree, look closer.
Slide through cached training runs to find the tune-vs-prompt break-even.
Wiretap the protocol traffic between your agent and its tools.
Blind-guess which model answered, then see the real leaderboard.
Watch VRAM fill up as you pick a model size.
See the exact prompt the model actually receives, every time.
Lint any prompt against the model's official rules — with receipts.
See why that chunk got retrieved, in 2D space.
Edit one reasoning step mid-thought and watch the answer fork.
Break the model, then watch your attack become a regression test.
Build an AI pipeline, then watch traffic find its breaking point.
Watch a pipeline halt exactly where the model's output lied.
20 playgrounds
Plug in your numbers, watch the break-even cross.
Drag one slider, watch search quality rise then crash.
Paste any AI hype, get the substance score.
Swipe fast: real answer or confident AI lie?
King minus man plus woman, mapped and clickable.
Five models race one task, cost ticking live.
Scrub the slider: pure noise becomes a picture.
Pack the AI's memory before it forgets your question.
Three search strategies race for the same buried answer.
Turn off JSON enforcement, watch a mock app crash.
Same prompt, five dice rolls, one slider controls chaos.
Watch the AI model map reshuffle month by month.
Watch AI decide which words matter to which.
Steal the AI's secret, level by escalating level.
Type a word, watch AI shred it into tokens.
Rewrite one sentence, watch the AI agent stop guessing.
Find the exact step where the AI agent broke.
You train the judge, then we expose its bias.
Same confident tone, one answer is grounded truth.
Will that model actually fit your GPU's memory?
16 playgrounds
Pick a model and your GPU. We do the honest math.
Every playable ends in one card worth posting.
Don't answer the AI. Interrogate it first.
One of these AI answers is confidently making it up. Find it.
One prompt, one timer, one shippable thing. Go.
Build your own AI future. Compare it to the forecast.
Your AI skills, rendered as a character you actually earned.
Every playable feeds one public board. Rank travels with you.
One AI mystery a day. Free forever. Streak or it didn't happen.
Drop a real miatz playable into any blog post. Free.
Write the check that would have caught it. Then test it.
Live a real AI failure. Make the calls that matter.
Paste any AI text. Get plain English and the receipts.
See exactly what's holding your stack hostage, and the exit.
Every model, ranked, priced, and told straight.
Watch exactly where the agent went wrong. Every step, exposed.
7 playgrounds
Every roadmap bet hides an iceberg under its estimate.
Peek early, win often, be wrong nearly half the time.
One prompt, two builds, one hidden eval decides.
The customer's real truth stays hidden until you submit.
Sort real quotes, spot the feature request in disguise.
Nudge one weight and watch your 'objective' backlog reshuffle.
One metric climbs while the one that matters falls.
6 playgrounds
Guess pass or fail before the ratio math proves you wrong.
Pull one token thread, see what secretly doesn't move.
Same screen, two passes, two very different issue counts.
Tag the flow's ten sins before the timer runs out.
Your assumed journey against what the data actually shows.
Every silent decision the agent made, pinned and counted.
7 playgrounds
Every channel bends — find the knee before your CFO does.
Your sharp positioning line, scored against everyone else's.
What you swore mattered against what actually closed deals.
Loops compound, funnels flatline — watch the crossover arrive.
Tag every leak before the timer beats your best guess.
The spam filter grades your email live, line by line.
Write two versions, see which one the AI actually quotes.
5 playgrounds
Delegate under a clock, then watch the real blast radius.
Build a RACI matrix, then watch the project stall live.
Talk a skeptical exec into a guardrail before trust hits zero.
Watch a live dollar counter price out your own meeting.
Write an AI policy, fast-forward a year, see what it missed.
4 playgrounds
Drag the option pool and watch your ownership quietly shrink.
Paste your pitch, watch every claim get graded live.
Pull three levers, watch your real runway cliff move.
Negotiate a term sheet, then replay who actually gets paid.
4 playgrounds
See how often 'significant' happens with zero real effect.
Write SQL, then see why 'fast enough' was a full scan.
Forecast the future, then see the uncertainty everyone hides.
Stop an A/B test early, then watch it turn out wrong.
3 playgrounds
Claim automation savings, then audit your own number.
Grant an agent permissions, then see the blast radius light up.
Negotiate an AI vendor's pitch, then check what's actually proven.
3 playgrounds
Negotiate an offer against a market band you can't see yet.
Type your bullet points, see what an AI screener actually reads.
Answer interview questions, then see what was really scored.
Playgrounds are how miatz teaches — from the inside. Sit Chuun and train on them for real.