The chat box was always the wrong unit.
Most people meet large language models inside a single textbox and a blinking caret. The textbox is generous — it accepts anything— but generosity isn't the same as clarity. The textbox treats every prompt as a flat string. The mind that wrote the prompt is anything but flat.
PromptCorrectlystarts from the opposite end: what does the person actually mean, and how can the surface make that meaning legible to both human and machine? That's the whole product. A canvas, a composer, a brain trainer, and an open library — each a different angle on the same problem.
Schemas are active organizations of past reactions or past experiences.
Both of you are running schemas. Yours just happens to be older.
Frederic Bartlett observed in 1932 that memory isn't a tape — it's a reconstruction, scaffolded by mental templates we never see. Rumelhart turned that into a formal theory of cognition fifty years later. The model in your chat window is doing something suspiciously similar: pulling from a vast schema space, betting on the most likely reconstruction of what you meant.
The job of a prompt isn't to issue commands. It's to align two schema spaces — yours and the model's — at the smallest possible distance. Every node on a Studio canvas is one alignment dimension.
Most of the time we are not thinking about thinking.
The prompt people regret is the one they typed without slowing down.
Kahneman's two systems aren't a metaphor — they're a load-bearing model of human cognition. System 1 is fast, associative, vibe-driven. System 2 is slow, structured, expensive. Most prompts come out of System 1. Most prompts that work required System 2 first.
Brain Trainer is built to make the System-2 detour cheaper — a five-minute drill, a structured field, a thinking textarea that fires before the prompt textarea. Sweller called this germane load: cognitive effort spent on building durable mental models, not wasted on noise.
Working memory is the bottleneck of all complex cognition.
Pitched right at the edge of your ability.
Csikszentmihalyi spent thirty years documenting the conditions of flow. Two of them are inescapable: clear goals, and a challenge level that meets — but doesn't exceed — present skill. Every Brain Trainer lesson, every Studio template, every Composer field is tuned to that gradient. Just hard enough.
The best moments usually occur when a person's body or mind is stretched to its limits.
You should leave each session feeling more capable, not more dependent.
Bandura argued that self-efficacy — the belief that you can act effectively in the world — is the foundation of every other skill compounding over time. The opposite is learned helplessness, the slow rot of treating the model as an oracle.
PromptCorrectly is built to grow your agency, not replace it. Auto-Generate exists, and it's good. But the canvas is the main event — because the canvas is where you stay a co-author, not a customer.
Built in public, priced for humans, powered by your own keys.
No data hoarding. No surveillance. Your API key lives encrypted in your browser; we literally cannot read it. The community library is licensed CC-BY for anything you contribute, which means it's yours forever. We make money only when Pro launches — roughly $16/mo, less than a streaming service, on top of your own model costs.
We're looking for the first 369 contributors to seed the library. Six months of full Studio access in exchange for thirty prompts and nine skills. That's the whole pitch.