High-leverage skills for hackers

gistfile1.md

• Terminal.
  • Master the Terminal.
  • Use zsh.
  • Show current git branch.
  • Reverse search your command history using Ctrl+R.
  • Quickly autocomplete commands by pressing up (again reverse search).
  • Delete+move quickly with cursor - use Alt+Left and Alt+Right to jump words. Alt+Del to delete a word.
• Coding.
  • Use CoPilot and/or other coding AI.
  • Google things, ask ChatGPT for sample code, ask questions on StackOverflow and Twitter.
  • Repeat yourself twice before refactoring.
  • There are two modes in software design - divergent (exploring the problem) and convergent (narrowing down the solution).
  • During divergence, write all your code in one massive file.
  • During convergence, refactor and split into different files.
  • Use comments a lot -
    • Document and structure your code using headings. Imagine a military recipe. Use the comment to repeat what the code is doing, but in terms which are faster to understand.
    • Provide sample data and variables inside comments. This is helpful when explaining code that parses things.
  • Source control.
    • Git is a microblogging network for coders.
    • When there is only one user in your social network, you are writing to your future self.
    • Commit when you have achieved something. Commits are designed to be checkpoints for progress.
    • During divergent design, I find source control to be mainly useful as a backup mechanism. I do not bother with commit messages, they are overhead. Simply git commit -m .
• Learning how to code.
  • Find a problem you have, and figure out how to solve it.
  • The most important skill in coding is the meta-skill - how fast can you learn?
  • Software engineering is generally a small set of ideas, expressed in 500 different ways. Greenspun's tenth rule - "Any sufficiently complicated C or Fortran program contains an ad hoc, informally-specified, bug-ridden, slow implementation of half of Common Lisp."
  • The best tools:
    • Python - for data-related apps (ML, data science, data analysis, crawling), and basic scripting (files). Use with Numpy and Jupyter notebooks.
    • Javascript - for API's and bots. NPM package ecosystem is one of the best. Use with Typescript's compiler, tsc, to reap the benefits.
    • Docker and Docker Compose - for devops. Containerise your build environment so that other people can use it.
    • Next.js/React.js - for building modern web apps.
    • SQLite - for simple databases.
    • Django - for simple backends.
    • Go - for simple systems programming that uses concurrency.
    • Rust - for complex systems programming (browser engines).
    • Nix - a package manager for CLI tools. Much better than brew.
    • Markdown - simplest language to express formatted text in.
• Software engineering.
  • Practice the MIT approach. "Worse is better".
    • Strive for completeness. Your software should address as much of the problem it solves as possible.
    • The interface should be simple, it is more important for the interface to be simple than the implementation.
    • Correctness - the design must be correct in all observable aspects. Incorrectness is simply not allowed.
    • Consistency - the design must not be inconsistent. A design is allowed to be slightly less simple and less complete to avoid inconsistency.
  • Learn functional programming.
    • Practice writing code which is functional - map, reduce, filter.
    • Remember that FP is a complement to imperative, procedural programming.
    • Writing using the FP paradigm will naturally produce reusable functions.
    • A parallel to this is the Linus Torvalds quote: "Bad programmers worry about the code. Good programmers worry about data structures and their relationships."
  • Document your process for later. Always. Whether it's in bash_history or in random markdown docs.
  • When working in teams, remember that software is principally a social process:
    • Microservices is an embodiment of Conway's Law