We need to talk about the gap between the promise of AI and the reality of using it in real software work.
The promise is simple: AI will make you ten times faster. Generate boilerplate in seconds. Refactor entire layers in minutes. Debug complex logic while you sip coffee. The metrics in marketing decks point one way: speed. Tokens per second. Latency. Autocomplete hit rates. Code acceptance percentages.
But if you sit next to an actual developer doing real work with these tools, you see something completely different.
You don't see smooth code flowing from brain to editor. You see a ping-pong match. The developer writes code, hits a blocker, alt-tabs to a browser, copies a snippet, pastes it into a chat window. They spend two minutes explaining context that isn't visible in those lines: framework versions, database schema, implicit constraints, edge cases living in comments and tribal memory.
The AI sends back a confident answer. They copy it to the editor. It fails—missing import. They fix it, run again. It fails—the AI used a method that doesn't exist in this version. Back to the browser. Clarify. The model apologizes. Try again.
This isn't ten-times speed. This is friction.
The bottleneck in AI coding today isn't model intelligence. Models are already good enough to help with many parts of the job. The real bottleneck is context recovery. Every time you step outside your editor to consult an AI, you pay a tax.
You pay it in focus. In time. In the fragility of the mental map you're holding while you work.
The core thesis: AI-assisted coding doesn't fail on model quality. It fails on context friction.
This book is about that tax:
- How developer context actually works
- How current tools burn it without noticing
- How to measure that loss instead of guessing
- How to design (or choose) AI tools that respect flow instead of breaking it
Who this is for:
- Engineering leaders evaluating AI tools for your teams
- Senior/Staff engineers who feel something's wrong with current workflows
- Tool builders designing the next generation of developer AI
We'll stay close to the work. No hype. No generic productivity slogans. Just the mechanics of attention, context, and tool design—from the perspective of people who actually ship code.