Chapters
15 chapters across 4 topics
Parsers & Languages
- 01 The Parsing Pipeline: Tokenizers and ASTs How tokenizers break text into tokens and parsers assemble them into abstract syntax trees — the foundation of every programming language.
- 02 Parsing Numbers and Strings Extending a recursive descent parser to handle numeric and string literals — tokenization patterns and AST construction.
- 03 State Machines and Regular Expressions How finite state machines power tokenizers, and how regular expressions provide an elegant specification-based approach to token recognition.
- 04 Statements and Statement Lists Evolving a parser from single expressions to multiple statements — grammar definitions, delimiters, and test-driven development.
- 05 Blocks, Scopes, and AST Formats Implementing block statements, nested scopes, and exploring different AST representation approaches.
Interface Design
- 01 How a Schedule Interface Works The architecture behind a scheduling system that handles thousands of shifts across hundreds of facilities — daily views, monthly calendars, and the data flow that powers them.
- 02 Designing Bottom Sheets That Nest How to build mobile bottom sheets with drag-to-dismiss, velocity detection, iOS spring curves, and transparent nesting that scales sheets automatically.
- 03 Making Updates Feel Instant The multi-stage optimistic update pattern — how to make the UI respond in 5ms while the API takes 200ms, with rollback on failure and cache key precision.
- 04 How Timezone-Safe UIs Work Why storing dates as ISO strings prevents an entire class of bugs, and how to build interfaces that work correctly across every US timezone.
AI, ML & Dev Productivity
- 01 Distilling Large Models into Small Ones When frontier models are too expensive at scale, teacher-student distillation can cut inference costs by 85% while preserving quality for your specific domain.
- 02 Context Compaction for AI Agents Why production AI agents need a fast, cheap subagent to compress conversation history - and how diffusion LLMs make this practical at 1,000 tokens per second.
- 03 The Harness Is the Product Agents write 75% of our merged PRs. The bottleneck is no longer code generation - it is proving the code works. Here is how we are building the verification harness that makes autonomous delivery possible.
- 04 The Future of Software Engineering Agents write better code than most humans. The job is shifting from writing code to deciding what to build, making solid plans, and proving the output works. Here is how I think about it.
- 05 Go Fast: Write Minimal Software The highest quality, fastest software is no software at all. Complexity is the speed killer - the parachute behind the drag racer. Here is how I think about writing less code to ship faster.