The visual workspace for prompt engineering

FROM VAGUE INTUITION
TO STRUCTURED INSIGHT.

Align human schemas with machine cognition.

Grounded in Bartlett (1932) + Rumelhart (1980) schema theory

The visual workspace for people who actually use AI — Studio, Composer, Cortex, and 3,750+ ready prompts.

Open Studio
No card to startWorks with Claude, GPT & GrokYour keys never leave your browser
Article Master Intelligence — Studio✦ Auto-Generate▶ Run↓ Export
★ Library5
AllSkillsPrompts★
PROMPTArticle Master Intelligence
SKILLChain-of-Thought Master
PROMPTSelf-Critique Loop
SKILLJSON Enforcer
PROMPTStop Hallucinations
GOALDistinguished editoria…ROLESenior writer · Atlant…CONTEXTTopic · audience · toneCONSTRAINTS≤ 1,200 words · no jar…INSTRUCTIONDraft a thorough articleFORMATMD · 3 H2LLM EXECRun · BYOK
Total brain output · streaming1,240 ↓ / 420 ↑ · $0.018
## How prompts compile
The schema-first approach forces the model to commit to a structure before it generates the body
PromptCorrectlyAI MARKET ANALYSIS · WHAT CHANGED IN 2025
WHAT GPT-5 ACTUALLY CHANGED

GPT-5 launched in 2025. Here's what actually changed — and what didn't.

WHAT CHANGED
Model capability at the ceiling
Reasoning on complex tasks improved. Creative range expanded. Tool use became more reliable. The ceiling rose.
WHAT DIDN'T CHANGE
The floor for unstructured prompts
Bad prompts still get bad outputs. The model is more capable — but that capability requires specification to activate. The gap between ceiling and floor got bigger.
A more capable model with a bad prompt is still a bad output. Structure is still the unlock.
THE AI EDUCATION MARKET IS EXPLODING

The AI education market is projected to reach $20B by 2027.

Verified Market Research 2025 · AI in Education market forecast
2023 market
AI education tools, courses, and platforms combined
$4B
2024 market
Post-ChatGPT acceleration · enterprise training surge
$8B
2025 estimate
Structured prompting education driving growth
$13B
2027 forecast
Projected with enterprise AI training mandates
$20B
What's driving the growth — and is the education being produced actually effective?
WHAT YOU ACTUALLY PAY — TOTAL

Here's the full cost comparison — including what other tools don't tell you.

PROMPT-ENHANCEMENT SUBSCRIPTION
€25–50/mo for output, not skill
Monthly cost for a tool that improves prompts for you. Skill development: zero. Cancel: back to your original capability.
PROMPTCORRECTLY + BYOK
€18.99/mo + €3–5 API for skill + output
Platform cost plus your own API access at cost price. Skill develops permanently. Cancel: keep the skill.
You pay for capability you keep, or for a service you rent.
43%
avg. accuracy gain with Chain-of-Thought prompts
10x
productivity increase in 2025 enterprise studies
3,750+
expert prompts in our free library
2026
best practices, always up to date
⌘
Coding
142 prompts
◇
Business
98 prompts
✦
Research
87 prompts
✎
Education
76 prompts
+
Medical
54 prompts
✺
Creative
133 prompts
★ Today's featured expert skillDaily pick

One new research-backed skill template, every day.

DataAdvanced

Data Story Analyst

Transform raw data into narratives with clear insights, visualizations, and recommendations.

Skill template · preview
[ROLE]
You are a senior data storyteller embedded in a product analytics team.
Your job is to turn raw numbers into a 90-second narrative an executive will act on.

[CONTEXT]
{{dataset}} — schema, time window, and known anomalies.

[METHOD]

…
Research hub · 100% free access

Backed by the latest science.

Every technique in our library is grounded in peer-reviewed research. Here's the evidence.

2024Sahoo et al. · arXiv 2402.07927

A systematic survey of prompt engineering in large language models

“Structured multi-step prompting increases accuracy by up to 43% across reasoning tasks.

2024University of Maryland · June 2024

The Prompt Report — 58 prompting techniques benchmarked

“Few-shot + Chain-of-Thought combined is the single highest-performing technique in 2024.

2025Chen et al. · ScienceDirect

Unleashing the potential of prompt engineering for large language models

“Domain-specific role prompts reduce hallucinations by 31% in professional contexts.

2025medRxiv · April 2025

Patient education and AI readability across pediatric specialties

“Properly structured prompts improved medical content readability scores by 2.4 grade levels.

View all 32 papers →
✦ Editorial

Latest insights & research.

Evidence-first articles from the world of prompt engineering, written for practitioners.

Technique Deep Dive

Chain-of-Thought vs. Tree-of-Thoughts: which wins in 2026?

A side-by-side benchmark across 400 reasoning tasks. Where ToT pays off, where CoT is enough, and how to pick.

⌚ 8 min🔒 Free preview
Research Highlight

The Prompt Report 2024: 58 techniques ranked by performance

University of Maryland's 58-technique benchmark, summarized for practitioners. The five techniques worth your time.

⌚ 12 min🔒 Free preview
Enterprise Prompting

How one legal team 10x'd document review with CRISPE

A 14-person legal team replaced 3 paralegals with a CRISPE pipeline. The exact prompt, the exact result.

⌚ 6 min🔒 Free preview
Prompt Fails & Fixes

Why 80% of ChatGPT users get mediocre results

The four most common prompt failures, ranked by frequency in 12,000 logged conversations.

⌚ 5 min🔒 Free preview
Future of Prompting

AI-generated prompts now outperform human experts

Meta-prompting closes the gap. A 2025 benchmark where AI-authored prompts beat human-authored on 7 of 10 tasks.

⌚ 10 min🔒 Free preview
Education

K-12 STEM teachers using these 5 prompts to transform learning

Five battle-tested prompts from a 2025 NSF pilot. Lesson plans, Socratic tutors, and rubric-aware feedback.

⌚ 7 min🔒 Free preview
Log in free to read all articles →
The Studio

A canvas for prompts — not a chat window.

Drag blocks. Wire them up. Run, iterate, version, ship. Every prompt becomes an asset your team can actually maintain.

←
PromptCorrectly
Untitled Canvasauto-saved · 12s ago
+ NewSaveLoadShare✦ Auto-Generate▷ Run Selected▶ Run⬇ Export⛶ Focus
Add Node
◎Goal◆Role≡Context→Instruction⛓Chain-of-Thought{ }Output Format↺Self-Critique⚙Agent Loop✎Note
◎Goal1
◆Role1
≡Context3
❏Files2
❖Examples4
⚑Constrain…2
→Instruction
{ }Output Format
✓Output
MAP100%
−100%+⤢
Idle · ready to runnodes 9edges 8done 0/9
tokens in 412tokens out 850cost $0.0047BYOK · Anthropicclaude-4.5-sonnet
Open Studio →
Cortex

Rewire how you think — before you prompt.

Twelve hand-crafted courses (5 free + 7 paid), taught by She — your AI guide. Each lesson is 4 minutes of theory, one A/B exercise, one prompt you'll actually use this week.

🧠
Course 01 · In Progress
Mindset 101 — Think before you prompt
3 of 6 lessons
50%
She
She · your AI guide
Lesson 2 · The Architect mindset
“Most people prompt the model. Prompters tell it what they want. Architects tell it how to think — and then let it surprise them.
▶ play audio · 38s·full transcript
Exercise · Pick One
Which prompt will produce a better answer?
A

Write me something about climate change.

B

As a science journalist, explain how ocean acidification affects coral reefs in 3 short paragraphs, using one striking metaphor and one data point.

✓Correct — B gives the model a role, a topic, a structure, and constraints.
Practice · 90s
Now write your own.

Structure: Role + Topic + Constraints + Format.

As a senior product designer, review this onboarding flow, flag friction in the first 90 seconds, output a bulleted list with severity
✓ 4 of 4 ingredients · strong specCheck my answer →
🧠
Resume
Prompting Mindset 101
The 90% of prompting that happens before you type a word.
“Most people lose 80% of their prompt's power before they type a single word.
2 of 6 · resume
↯
Free
Basic Chain-of-Thought
Force the model to think out loud. +43% accuracy on reasoning tasks.
“Force the model to think out loud.
● 6 lessons · 10 min
👤
Free
Simple Role Prompting
The fastest one-line upgrade you can make to any prompt.
“The fastest one-line upgrade you can make to any prompt.
● 6 lessons · 8 min
▦
Free
Output Format Basics
Stop fighting messy outputs — control format with five words.
“Stop fighting messy outputs — control format with five words.
● 6 lessons · 9 min
✎
Free
Clarity Over Cleverness
The counter-intuitive rule that breaks most prompts.
“The counter-intuitive rule that breaks most prompts.
● 6 lessons · 7 min
✦
Free
Prompt Anatomy 101
The 5-part canvas: Role · Context · Task · Format · Constraints. Every elite prompt has all five.
“Most prompts fail because one ingredient is missing — and the writer doesn't know which one.
● 6 lessons · 6 min
✦
Free
Delimiters & Tags
Why <context>…</context> tags beat quotes. The tokenizer-level reason your structured prompts get ignored.
“Claude was trained on XML-tagged documents.
● 6 lessons · 5 min
✦
Free
Tone & Audience
Why "write for a CFO" beats "write professionally" — audience-aware prompting in one sentence.
“Why "write for a CFO" beats "write professionally" — audience-aware prompting in one sentence.
● 6 lessons · 5 min
✦
Free
Single-shot vs Iteration
When to one-shot. When to chain. The decision rule pros run in their head before every prompt.
“When to one-shot.
● 6 lessons · 7 min
✦
Free
The 3-Line Rule
Compress any prompt to three lines: Who · What · How. If it doesn't fit, you don't understand the task.
“Compress any prompt to three lines: Who · What · How.
● 6 lessons · 5 min
✦
Free
Before You Prompt
Thirty seconds of pre-flight saves an hour of iteration. Four questions you answer before typing a word.
“Thirty seconds of pre-flight saves an hour of iteration.
● 6 lessons · 6 min
✦
Free
Negative Instructions
"NEVER do X" is sometimes the most powerful line in the prompt. Here's when to use them — and where they backfire.
“"NEVER do X" is sometimes the most powerful line in the prompt.
● 6 lessons · 6 min
✦
Free
Prompt Debugging 101
Read bad outputs as data. A 3-question diagnostic that finds the broken field in any prompt.
“Read bad outputs as data.
● 6 lessons · 7 min
✦
€6
Self-Consistency Voting
Run the same prompt 3-5 times → take the majority answer. +12-18% accuracy on reasoning tasks (Wang et al. 2022).
“One prompt, one answer = a guess.
● 6 lessons · 14 min
✦
€7
Few-Shot Pattern Recognition
Show 2-4 examples → the model maps the relational structure, not the surface details (Chase & Simon's chess masters, applied).
“Chess masters don't memorize moves — they recognize patterns.
● 6 lessons · 16 min
✦
€5
Constraint Calibration
Too few constraints → drift. Too many → brittle. The 5-constraint sweet spot and how to find it.
“Too few constraints → drift.
● 6 lessons · 12 min
✦
€6
Multi-turn Refinement
Conversation as compute. Use 3-5 turn loops when the goal is discovered, not pre-specified.
“Conversation as compute.
● 6 lessons · 13 min
✦
€7
Meta-Prompting
Ask the model to write its own prompt. The technique that compresses 4 hours of prompt-engineering into 4 minutes.
“Ask the model to write its own prompt.
● 6 lessons · 14 min
✦
€8
Prompt Evaluation Rubrics
Build your own grader. The Chip-Huyen-style eval rubric that catches drift before it ships.
“Build your own grader.
● 6 lessons · 18 min
🧬
€5
Advanced Chain-of-Thought Mastery
Stack CoT with role + self-critique for 60%+ accuracy on multi-evidence reasoning.
“Naive Chain-of-Thought looks impressive.
● 6 lessons · 22 min
🎭
€4
Role & Persona Prompting That Actually Works
Multi-dimensional personas that shape voice, taste, and decision-making — not a frankenstein average of 10,000 examples.
“"Write like a CEO" gives you a frankenstein average of every CEO the model has ever read.
● 6 lessons · 16 min
{}
€6
Output Formatting Secrets
Schema enforcement, prefilling, XML tags — the production tricks vendors don't teach.
“"Output the result as JSON" sounds like a constraint.
● 6 lessons · 18 min
🛡
€7
Stop Hallucinations Forever
Production-grade hallucination prevention via citation-or-refuse + admit-uncertainty + claim triage.
“Models don't hallucinate because they're lying.
● 6 lessons · 16 min
⚙
€9
Building Simple Agents (ReAct)
Thought → Action → Observation — the foundation of every modern agent, with halt conditions that don't burn $500.
“Bad agent prompts are the most expensive prompts you'll ever write.
● 6 lessons · 24 min
◇
€5
Pre-Project Thinking Framework
The 30-minute setup that 10× any AI project — before you write a line of code.
“Most AI projects fail in the first 30 minutes.
● 6 lessons · 16 min
⊢
€8
Tree-of-Thoughts for Planning
When linear reasoning fails — enumerate branches, score independently, surface trade-offs, recommend a hybrid path.
“CoT picks one path and rides it — even when 3 better paths existed.
● 6 lessons · 22 min
✦
€9
Legal Precision Architect
IRAC + 5 C's: the framework that turns hallucination-prone research into committee-ready memos.
“The legal field doesn't punish prompts that fail — it punishes prompts that fail confidently.
● 6 lessons · 22 min
✦
€7
SaaS Velocity Engineer
RICE scoring · churn-proof CS emails · PRDs that ship. The SaaS PM playbook in prompt form.
“RICE scoring · churn-proof CS emails · PRDs that ship.
● 6 lessons · 18 min
✦
€8
Founder Pitch Architect
10-slide YC-style decks · investor Q&A prep · MVP roadmap loops — the founder's prompt stack.
“10-slide YC-style decks · investor Q&A prep · MVP roadmap loops — the founder's prompt stack.
● 6 lessons · 20 min
✦
€7
Marketing Campaign Director
Brand voice doc + multi-channel campaign engineering + ROI projection — the agency director's playbook.
“Brand voice doc + multi-channel campaign engineering + ROI projection — the agency director's playbook.
● 6 lessons · 18 min
✦
€6
Viral Content Alchemist
8-word scroll-stop hooks · platform-specific scripts · viral pattern recognition from 3 reference videos.
“8-word scroll-stop hooks · platform-specific scripts · viral pattern recognition from 3 reference videos.
● 6 lessons · 16 min
✦
€6
Clip Architect
Turn 1 long video into 30+ platform-optimized clips. The OpusClip prompting layer.
“Turn 1 long video into 30+ platform-optimized clips.
● 6 lessons · 15 min
✦
€6
Internalization Drills
5-minute daily reps that wire prompt anatomy into reflex. Designed for the commute.
“The goal isn't to know the technique — it's to use it without remembering you're using it.
● 6 lessons · 12 min
✦
€8
Prompt Literacy Audit
Score yourself against the Deloitte 4-flavor framework. Discover which of your gaps is the bottleneck.
“Deloitte's 2024 AI skills gap research found 4 distinct gap flavors.
● 6 lessons · 24 min
✦
€9
Our Goal Is You Don't Need Us
Graduate the platform. The capstone that wires every technique into a single mental model.
“Our goal is for you to not need us.
● 6 lessons · 26 min
✦
€10
The Architect Mindset
The final synthesis. One mental model that fits in 60 seconds — and changes how you work with AI for life.
“Most people prompt the model.
● 6 lessons · 28 min
Start Mindset 101 →See the 8 pillars
How it works

Five steps. No vendor lock-in. Your keys, your data.

Drop blocks, wire them up, paste your API key, run. Export anywhere.

01

Drop nodes

Drag Goal, Role, Context, Instruction, Output Format from the sidebar onto the canvas.

02

Wire logic

Connect blocks with a line. Add Skills like Chain-of-Thought, Self-Critique, JSON Enforcer.

03

Bring your key

Paste your Anthropic or OpenAI API key once. We never store it — it lives in your browser.

04

Run & stream

Hit Run. Watch tokens stream into Output in real time. Iterate. Diff against earlier runs.

05

Export anywhere

Copy to clipboard. Push to Notion. Generate Python. Save as a versioned prompt for your team.

The Library

The world's largest open prompt repository.

3,750 prompts and counting — curated, versioned, peer-reviewed. Copy a single prompt, or fork the whole flow into your Studio. No login needed to browse.

⌕Search 3,750 prompts — try "investor update", "SQL", "tone of voice"…⌘ K
All 3,750EngineeringDesignGTMDataWritingResearchOpsSort · Most copied ▾
● GTMClaude 4.5 Sonnet
Intermediate
Cold outbound — peer to peer

Reach a busy founder without sounding like AI slop. Four lines, one ask, no hedging.

● EngineeringClaude 4.5 Sonnet
Beginner
Code review — staff engineer voice

Review a PR like a senior reviewer who actually cares about the codebase, not just style.

● DesignClaude 4.5 Opus
Beginner
UX critique — first 90 seconds

Walk through an onboarding flow and flag every point of friction, with severity ratings.

● DataGPT-5
Intermediate
SQL → English explainer

Take any SQL query and explain it like the analyst already knows the schema. Bullets only.

● GTMClaude 4.5 Sonnet
Advanced
Investor update — calm + concrete

Write a monthly update that earns trust. Wins, misses, asks. No corporate fog.

● EngineeringClaude 4.5 Sonnet
Expert
Bug repro — minimal failing case

Turn a vague bug report into a clean repro: steps, expected vs actual, environment.

● ResearchClaude 4.5 Opus
Beginner
User interview synthesis

Distill 12 customer calls into themes, frequencies, and a brutally honest top-3 list.

● OpsClaude 4.5 Sonnet
Intermediate
1:1 prep — direct report

Surface what's actually on their mind. Five questions you'd never think to ask alone.

● EngineeringGPT-5
Beginner
API endpoint spec

Write a tight REST endpoint spec — params, errors, examples, idempotency contract.

● WritingClaude 4.5 Sonnet
Expert
Brand voice rewrite

Take any draft and rewrite it in our voice. Keep the meaning, drop the hedges.

● EngineeringClaude 4.5 Opus
Intermediate
Architecture decision record

Write an ADR future-you will still understand. Context, options, decision, consequences.

● GTMGPT-5
Beginner
Press release — YC launch

Announce a B2B SaaS launch in the YC tone. Sharp lede, two metrics, one founder quote.

Browse all 3,750 prompts →
Privacy by default

Your keys + your data never leave your browser.

🔑
Bring your own key
Paste your Anthropic / OpenAI / Grok key once on /settings. It writes to your browser's localStorage. Our server forwards it per-request and never persists, logs, or shares it. Audit the network tab to verify.
💻
Client-side execution
Every Run streams from the provider directly to your browser through our zero-log proxy. No middleware reads your prompts. No analytics tracks your outputs.
📦
Own everything
Your canvases save to localStorage now, your own Hetzner-hosted Postgres next. Open-source canvas component on GitHub (MIT). Export to JSON / YAML / Mermaid / LangGraph / executable JS any time.
Founding Beta · First 369 spots

Instant full access — starts in 30 seconds.

Connect your own Anthropic or OpenAI API key → use full Studio, Composer and Cortex immediately. You only pay your own API costs (≈ €3–5/month for normal use). 30 days of platform access on the house. After that, €16/mo Pro — and the discounts you've earned by contributing kick in.

First 369 only
Beta users · open now
FREE
30 days of full platform access
no card · bring your own API key · €3–5/mo typical usage

Then €16/mo Pro — but contributors earn lifetime discounts automatically through the Founder Points system. Top contributors pay €4/mo forever (75% off).

Claim my spot
228 of 369 spots filled · 141 left
What every founder gets, day 1
  • Full Studio · Composer · Cortex · Library — all unlocked the moment you connect your key
  • 30 days completely free platform access (you only pay your own API usage)
  • Founding-contributor badge on your library profile, forever
  • 2× Founder Points multiplier — first 6 months of beta only
Founder Points · earn lifetime discounts
  • Upload a prompt → if it passes our originality + quality filter, it joins the public library → +10 pts (×2 first 6 months)
  • Upload a reusable Skill → +30 pts (×2 first 6 months)
  • Private uploads stay in your library only · 0 points · zero pressure to publish
  • Auto-claim: when you cross a milestone, a notification + one-click claim applies the discount to your next billing cycle
Reward Milestones
Level 1 · 570 pts
3 months at 50% off
≈ 30 approved prompts + 9 approved skills
Level 2 · 1,500 pts
6 months at 50% off
≈ 60 prompts + 30 skills
Level 3 · 3,000 pts
12 months 50% off + Hall of Founders badge
≈ 150 prompts + 45 skills · lifetime 50% option unlocked
Level 4 · 6,000+ pts
Lifetime 75% off · Legend status
≈ 300+ prompts + 90 skills · permanent reduced rate
1 approved prompt = 10 pts · 1 approved skill = 30 pts · 2× multiplier for the first 6 months of beta only · points never expire
After the 30-day window
Pro plan — €16/mo (monthly / quarterly / annually — your choice). Your earned discounts apply automatically. Pay in crypto (USDT/BTC/ETH/SOL) and get +10% bonus credits. Library publishing stays open to all users; the points + milestone system keeps running forever.