↓ Ir para o conteúdo principal

← todas as notas

📎 Webclip

Rules of thumb for a 1x developer

The post frames software work from the perspective of a 1x developer who has limited time, values conventional wisdom, and wants to improve by choosing rules that reduce waste and compound over time. It argues that many skills are context-specific, so learning should favor general-purpose ideas, durable tools, and habits that create a virtuous cycle.

Reading notes
#

  • Rules of thumb provide guardrails, structure, and a more explicit way to decide when goals require rationality rather than simple pattern-following.
  • Most learning is highly contextual, so much of what is learned in one job does not transfer well to the next.
  • Learning time should go toward things that compound, such as typing faster, building relationships, and reflecting on what is consumed.
  • Java is presented as a good fit for large enterprise systems, while C# is the Microsoft equivalent.
  • Python and Ruby are described as scripting languages used when speed matters more than legibility or debugging, with Python also used for ML and AI.
  • For harder languages, the post favors Go or Rust for fresh services with latency or performance needs, and Kotlin over Clojure or Scala for JVM work.
  • JavaScript advice is to use TypeScript, keep logic on the server when possible, use Vue or React for interactivity, and keep unit tests.
  • C is treated as something that is chosen by the problem space, while C++ is tied to performance-sensitive domains such as robotics, games, and high-frequency trading.
  • Serverless functions are recommended for small, simple code that runs only occasionally.
  • SQL is recommended for ad hoc queries and ACID transactions; noSQL is the fallback when those needs are absent.
  • Unit tests should be written when defect cost is non-trivial, integration tests when external code is not trusted, and end-to-end tests for user-session coverage, regression, or hard-to-predict results.
  • A dedicated SRE should be advocated when after-hours pages are a real risk, especially for inherited systems.
  • Security should not be delegated entirely to InfoSec, because that tends to produce overly restrictive rules.
  • Design sessions should be for input and questions, not for approval or signoff.
  • Estimates are described as creating pressure more than enabling planning.
  • The post distinguishes hard deadlines, soft deadlines, internal deadlines, and expected completion dates, and warns against conflating them.
  • When a team says Agile, the author prefers Kanban over Scrum because Scrum can turn estimation into pressure to work extra hours.