Brain Trainer
Pro · €5Intermediate · 16 min

Pre-Project Thinking Framework

The 30-minute setup that 10× any AI project — before you write a line of code.

€5 · Free during beta — paid in v1.1: extra challenges, bonus canvas templates, private community channel.

She · your AI guide
Most AI projects fail in the first 30 minutes. Not because the model is wrong. Because the question was. The 4-question pre-flight is the cheapest 10× multiplier you'll ever apply.

Sweller's 1988 Cognitive Load Theory identified working memory as the bottleneck of all complex cognition — including the model's. Every ambiguity in your prompt becomes working-memory load the model has to spend deciding what you meant instead of solving your problem. The same is true downstream: every ambiguous artifact (spec, ticket, design doc) the model produces inherits the original ambiguity, and every human who touches it has to re-solve the underlying question.

The production fix is the 4-question pre-flight. Before you write a prompt — before you write a brief, a Jira ticket, or a design doc for an AI feature — answer 4 questions: WHO is this for (specific user persona, not 'users'), WHAT decision must they make (the one thing they're stuck on), WHAT's the smallest output that helps them decide (format, length, fidelity), WHAT will we measure to know it worked (one number, not a vibe). These 4 answers compress the working-memory load by an order of magnitude.

The deeper insight: the 4 questions are about removing AMBIGUITY at the source, not adding STRUCTURE at the surface. Ambiguity removed at the source compounds — every artifact downstream is sharper. Structure added at the surface evaporates as soon as someone re-asks 'wait, who is this for?' If you can answer the 4 questions in 30 minutes, the next 6 weeks of work are 10× more efficient. If you can't answer them, you're building the wrong thing — and the 30-minute pre-flight just saved you 6 weeks.

Working memory holds ~4±1 chunks simultaneously — the bottleneck of all complex reasoning
Cowan (2010), reaffirmed by Doolittle et al. 2019
Projects with explicit success metrics ship 2.3× faster than those without
Microsoft Research, AI Project Lifecycle Study, 2024
67% of failed AI projects had no clear answer to 'what decision does this help the user make?'
MIT Sloan AI Initiative survey, 2024
~60s read