miniblog.

I'm regularly impressed by the engineering quality of OpenBSD. They're very much keen C developers, but many of their techniques apply to other PL communities. Software designed collectively, refactoring for clear design, even things like pledge.

Fun research demonstration of multi agent completion for machine learning training. The content is interesting, but the cute visualisations really add to the effect:

On the value of tracking upstream open source projects:

I'm amazed to learn that there are voice codecs that function at 700 bits/second! https://www.rowetel.com/?page_id=452 The target use case is amateur radio but I'm sure there's a range of useful applications.

On the dynamics of memory unsafe code, the economics of big finding, and future trends: https://www.cloudatomiclab.com/fuzz/

Excellent deep dive into the chain of events that led to an outage at Cloudflare in July:

Making your JS run faster, including some interesting comments on how expensive literal parsing can be:

Sometimes I wonder of the class browser is a local maximum for class/method definition in Smalltalk. It works well, but the pane design seems arbitrary. I know it's common to define methods from the debugger too, but perhaps other UIs could suit the ST model as well?

Remarkable discussion of the use of emoji as evidence in court: when is it evidence of harassment or a threat of violence? What if it renders differently on different devices? https://edition.cnn.com/2019/07/08/tech/emoji-law/index.html

Old meets new: van eck phreaking with neural networks! https://leveldown.de/blog/tensorflow-sidechannel-analysis/

What happens to developers specialise in a technology that has become obsolete? They tend to be very adaptable:

The most successful code review tools have a clear path to getting your patch included. It feels much more collaborative to ask for changes before a patch is accepted. Afterwards, a reviewer can be seen as difficult, even with the same feedback!

Quibi is exploring a different model to on-demand video content, even adding cute features like horror series which can only be viewed at night!

Usage of many raw materials has actually decreased due to smartphones replacing other consumer devices:

C : C++, Java : Scala/Kotlin, JS : TypeScript. There definitely seems to be space for languages that target the same platform. AFAICS the success criteria are: great interoperability, similar toolset, similar syntax, and a more elaborate type system. Are there counterexamples?