Pure and impure software engineering

Open original

Sean Goedecke argues that many engineering clashes, like open-source purists versus big-company engineers, come from people doing fundamentally different kinds of work rather than one side being less skilled. Pure engineering is open-ended craft aimed at the best technical solution, while impure engineering is about shipping good-enough solutions to someone else's deadline.

  • Pure engineering resembles art or research and is driven by aesthetics, while impure engineering resembles plumbing or construction and must compromise to meet business needs and schedules.
  • Pure engineering had more room at large tech companies in the 2010s, when hype-driven hiring funded it as a kind of developer marketing; that funding has shrunk as companies now need to make money.
  • Skilled pure engineers often struggle with impure work and vice versa, and the Casey Muratori versus Windows Terminal dispute and George Hotz's Twitter stint are used to show that impure work is harder than it looks.

Sean Goedecke's core claim is that a lot of friction between engineers comes from doing different kinds of work, not from one group being smarter. Pure engineering treats a technical problem as something to solve as perfectly as possible, closer to art or research. Open-source libraries and game engines are the typical examples. Impure engineering treats a real-world problem as something to solve well enough, on a deadline, for someone else's needs. Goedecke, who says he mostly does impure work and came from academic philosophy, argues that each side tends to assume the other is incompetent when they're really just optimizing for different things. Pure engineers often struggle under deadlines and compromise, while impure engineers can be too quick to accept a hacky fix and may lack the depth to find the cleaner path. He also argues that the balance has shifted. In the 2010s, big tech companies funded a lot of pure engineering because it produced impressive open-source work that doubled as developer marketing, and because hiring outpaced what companies had to do with the engineers. Goedecke points to large migrations like monolith-to-microservices and HTTP calls to event sourcing as examples of that excess. Now that hiring has slowed and companies have to make money, that kind of work is getting less funding, and many pure engineers are finding work more political than before. He also notes that companies still need both kinds: they lean on open-source for generic components like Kafka or Redis, but build some things in-house when their requirements are unusual, such as GitHub's own HTML parsing for Markdown rendering. To make the point about impure work being hard, he uses two cases. In 2021, Casey Muratori clashed with the Windows Terminal team over rendering performance. Goedecke thinks Muratori was technically right, but the team still shipped the feature, which landed in settings in February 2022, a reasonable timeline given competing work and edge cases like multi-glyph code points. His takeaway is not that Muratori is a better engineer than the whole team, but that a skilled pure engineer working alone on one problem can beat a team inside a company, because impure work fights legacy decisions, politics, and accumulated complexity on every feature. George Hotz's failed attempt to "fix search" at Twitter, followed by his move to the purer tinygrad project, serves as a second example of the same gap between the two skill sets.

Ask about this article