Systems Ideas That Sound Good
Steven Sinofsky's catalogue of engineering patterns that sound appealing in theory and fail 9 out of 10 times in practice. Hard-won wisdom from decades at Microsoft, delivered with the authority of someone who watched (and sometimes shipped) every one of these mistakes.
Key Quotes
"The only way something is truly pluggable is if a second implementation is designed at the exact same time as the primary implementation. Then at least you have a proof it can work... one time."
"9 out of 10 times once you go outside a framework or the data layer and think you can manage asynchrony yourself, you'll do great except for the bug that will show up a year from now that you will never be able to reproduce."
"Fail as you diverge from the underlying platform or as you build capabilities that are expressed wildly differently on each target."
Key Themes
Eight traps, each seductive:
- Pluggability -- the API is the behavior, not the header file. You need two implementations at design time to prove it works.
- Platform via APIs -- adding APIs doesn't make you a platform. Partners have their own businesses.
- Over-abstracting -- premature abstractions rot unused; post-hoc abstractions strangle performance and security.
- DIY async -- frameworks handle it; you won't, and the bug shows up in a year.
- Deferred security -- addressing access control post-launch guarantees failure.
- Data sync -- remains genuinely hard; companies worth billions exist for this reason alone.
- Cross-platform -- you're building an OS. Microsoft forked Office in 1998 rather than maintain unity.
- Escape to native -- frameworks maintain state that native calls bypass.
The common thread: each pattern feels like it simplifies but actually shifts complexity to where it's less visible and harder to debug. This is Simplicity in the Age of AI-Assisted applied to architecture decisions -- the inherited complexity that LLMs won't question because they were trained on codebases that already made these mistakes.
The pluggability point connects to Spec-First Development at Benchling -- Benchling's unified contract is the rare case where pluggability works, because they designed the platform and the first consumer simultaneously.
Critical Analysis
Sinofsky writes from the position of someone who shipped Windows and Office at scale, which gives the prescriptions unusual weight. These aren't theoretical objections; they're post-mortems.
The weakness is that it's from a particular era (desktop + OS platform). Some of these patterns have become more tractable since (cross-platform via web, async via language-native constructs like Go goroutines or Elixir processes). But the core principle -- that familiar patterns are seductive precisely because they defer complexity -- is timeless. The fact that Loomkin chose Erlang/OTP specifically to handle concurrency is a direct nod to Sinofsky's async warning.
Sources: Systems Ideas That Sound Good Last updated: 2026-05-14