Choose Boring Technology (2015)
Points and comments are a snapshot, not live.
Companies have limited 'innovation tokens' to spend on risky new technology.
Dan McKinley argues that companies should favor 'boring' technology (e.g., MySQL, Postgres, PHP, Python, Memcached) over novel alternatives because the failure modes of mature tech are well understood. He introduces the concept of 'innovation tokens': each company has roughly three to spend, and choices like Node.js, MongoDB, or new service discovery tools consume one. Known unknowns are manageable, but unknown unknowns are larger for new tech. Polyglot programming imposes hidden operational toil, and the best tool for the job is the one that minimizes long-term operational costs. McKinley recommends adding new tech only after first trying to solve the problem with the existing stack, documenting what makes it too hard, and committing to migration timelines.
What commenters are saying
Commenters strongly endorse the 'innovation tokens' concept as a practical tool for engineering leaders, comparing it to 'weirdness points' or 'beans' from earlier posts. Several debate which technologies now count as 'boring,' noting that Node.js is boring in 2026 while Bun or Deno would cost a token. A common refinement: treat the idea as a risk budget rather than a fixed number of discrete tokens, and note that personal experience and team context determine what is boring. One commenter points out that while Node.js itself is stable, the npm ecosystem with its fast-moving tooling and supply-chain attacks remains non-boring.