Putting the majority (but not all) of future maintainers & consumers first.

  • build for “real” devs (99%: insurance, healthcare, retail, banking) rather than influencer-hyped-up edge cases (1%: big-tech unicorns)
  • large-scale solutions don’t work for medium-sized problems
  • best engineering practices constantly change; there’s no fixed gold standard. Instead, teams should constantly iterate on defining processes
  • for users, “progress/good-enough/fit-for-purpose” is the goal, not “pristine code/100% reliability/perfect security” beyond realistic requirements. You don’t need 100% code coverage - 90% is likely more than enough, make sure you also (cough) test in prod to “test systems rather than just code”
    • to be able to test-in-production, errors must be both non-silent actually useful/informative
  • true value of a product doesn’t fit in a 1-min demo/GIF. Needs to stand up on “day 2” too:
    • e2e end-to-end
             test-run things properly
    • integrate with existing workflows
    • reduce (collaboration) friction
  • migration (replacing systems/workflows) should be continuous/agile. If it’s discrete/waterfall instead, then it’ll be slow & users will have to rely on legacy solutions for longer