miniblog.

It still feels rather weird when my Twitter feed shows me things that others have liked. We've come a long way from chronological: likes and retweets are probably just inputs into some machine learning algorithm.

Great interview with the Twitter CEO on goals, tradeoffs, and some interesting discussion of emergent behaviour after allowing users to make long display names.

Pipes were simple enough that they won: https://apenwarr.ca/log/?m=200801#31 It's a pretty limited API though, and I wonder what an incremental improvement in power would look like.

The latest version of Rust can automatically fix lockfiles that have merge conflicts in them: https://github.com/rust-lang/cargo/pull/5831/ This is often possible with machine generated files, but it's rare to see in practice. Super convenient though.

Interesting article on attracting millennials to working with mainframes: https://info.model9.io/lack-of-mid-level-mfers (Has some links to OSS mainframe projects that I had no ideas existed!)

Much has been said about JS fatigue, but the upside is that there's tons of creative experimentation. For example, eslint has a whole range of different 3rd-party output formatters. You can try them all out, and you tend to converge on a good solution.

The complexity of USB C, with a diverse range of features supported (or not) by different cables that can't be visually distinguished: https://www.extremetech.com/computing/257684-usb-c-chasing-unachievable-goal

One comment I saw @jack make recently was that they didn't expect displaying follower counts to motivate users to increase the number. Do we need to show this number on Twitter at all?

The new iphone is shipping with x86 chips for the baseband processor!

Babel is dropping the 'stage-N' presets in favour of specific plugins for features: https://babeljs.io/blog/2018/07/27/removing-babels-stage-presets Iterating on PL features, especially with new syntax, is hard! Seems like they're striking a good balance.

Finding bugs in LLVM IR optimisers: programmatically generate small (loop free, no control flow) functions, run your favourite middle end passes, and verify the output is equivalent! https://blog.regehr.org/archives/1510 (Uses Alive to verify and 4000 CPUs of compute.)

Awesome, there's a whole miniKanren implementation in emacs lisp! https://github.com/nickdrozd/reazon This gives you logic programming somewhat like Prolog or Clojure's core.logic.

Fn lightning talk on using provenance data to improve Haskell type errors: https://youtu.be/rdVqQUOvxSU

A growing range of useful collection data types for Pharo! https://pharoweekly.wordpress.com/2018/05/04/containers-now-on-github/ Good examples of Smalltalk best code practices too.

Interesting @lambda_conf talk showing how Dialyzer ('discrepancy analyzer') offers gradual typing for Erlang and Elixir: https://youtu.be/FXCMiQWnWu0 It's a neat type system: it uses 'success typing', (it assumes you're correct) and exotic features like ranges and non-empty lists!