miniblog.

The recommended solution for many NP-hard problems seems to be "just give it to a SAT solver". Are there cases where this isn't a good fit?

Elisp has no namespaces. Some tools actually take advantage of this: completion with all known functions can work really well for discovery!

A wonderful phrase for when Twitter is inhibiting your ability to communicate clearly!

Thread discussing a remarkable way of being memory safety if you have FS access (you can write to /proc/PID/mem!)

Are there PL features that you think are really powerful and elegant, but not worth it overall? If so, which?

AlphaGo Zero is simpler, does not require a training set, computationally cheaper and yet stronger!

Impressively, you can even tell typed racket that your lists have certain lengths, and it can skip bounds checks! https://docs.racket-lang.org/ts-guide/optimization.html#(part._.Lists)

Blunt words from Servo's HTML parser.

Brilliant, practical talk on implementing a JS type checker: https://youtu.be/MuC8I1JBKv0 (including soundness tradeoffs and flow-based analysis)

Introducing racket-mode for Emacs: https://youtu.be/QWiteH8PARQ (has a really cute REPL option that ensures scratch variables are removed)

Racket has 2 object systems, both single dispatch and CLOS style multimethods! Single dispatch is the default, which seems to help cohesion.

Great talk introducing @racketlang and typed/racket by @samth, but I particularly enjoyed this reminder!

Making my first Rust 1.0 project run on modern Rust. So many libraries needed nightly then, so I can't just compile without changes

A brilliant example of Twitter's character count methodology being rather unbalanced! (So many sites *advantage* English speakers.)

Defining an existing method on an existing type in Julia can break code. It has a really cute name: 'type piracy'