Deliberate practice: getting better on purpose

Ten years of experience is often one year repeated ten times. The difference between the two is deliberate practice — effortful, targeted work at the edge of your ability with fast feedback — and there is a real research literature behind it.

The foundations

• Peak: Secrets from the New Science of Expertise by Anders Ericsson — the primary source, from the researcher whose work everyone else cites (and whose “10,000 hours” was famously oversimplified by others). The core: naive repetition plateaus; improvement needs specific goals, full attention, immediate feedback, and discomfort. His Harvard Business Review summary is the 30-minute version.
• Deep Work by Cal Newport — deliberate practice requires unbroken attention, and this is the book on engineering your schedule to get it. Pairs with So Good They Can’t Ignore You: skill precedes passion, not the reverse.
• Ultralearning by Scott Young — the practitioner’s manual: directness (practice the actual task, not adjacent ones), drills for isolated weaknesses, retrieval over review, and overlearning.

The supporting science

• Make It Stick (Brown, Roediger, McDaniel) — what cognitive science actually says about learning: retrieval practice, spacing, and interleaving beat re-reading and highlighting, every time it’s measured.
• A Mind for Numbers by Barbara Oakley — focused vs. diffuse thinking and why stepping away is part of the work. The book behind the Learning How to Learn course.

Mapping it to engineering

Engineering mostly lacks the tight feedback loops of music or chess, so you have to construct them:

• Directness: to get better at systems design, build and measure systems — reading about them is warm-up, not practice. Every project on this site exists partly for this reason: the benchmark harness is the feedback loop.
• Drills: competitive programming problems for algorithms; re-implementing a paper’s core result for ML; postmortem-writing for judgment. Isolate the weak sub-skill; don’t just “do more projects.”
• Feedback: code review from someone better than you, benchmarks with committed baselines, and writing in public — each converts vague effort into corrections.
• The edge: if your success rate is near 100%, the work is too easy to cause adaptation. Pick problems where you expect to struggle and occasionally fail.

The uncomfortable summary from all five books: comfort and improvement are mutually exclusive, and the schedule — not motivation — is what decides whether practice happens.