Great article - thanks for sharing.

I was having similar thoughts. Devs are so used to (even love) adapting their tools (e.g. vim/IDEs, shells, themes, snippets, etc, etc) that many just double down on LLMs.

It's unavoidable but, in this case, is rather sad.

Very nice research direction!

But why is it even reasonable to assume that LLMs can "learn" an algorithm? Not just in the sense of "context" that results from an algorithm's description (or samples of its application), but as a methodology?

I don't know - it all feels scummy to me. A bit of searching around doesn't reveal an reputable source.

Cybernews, who supposedly found it, is a typical "ads driven portal", IMHO.

They talk about their research team, but otherwise have no factual blog posts, etc. that confirm its existence.

Sounds really interesting!

Dumb question, if I may.

Is the idea of "anonymity",when issuer and redeemer are the same entity, to enable "oracle" like verifications? I.e. one gets a credential, presents it to a 3rd party and that 3rd party then check the cred's validity with the issuer?

Rust does make NIF integration API usage both easier and robust. So it's a win.

Another aspect of "safeness", IMHO, is the amount of underlying C/C++ code.

In BEAM it's quite significant. It's well tested, but still, things can go wrong...

Nice talk. However, I really don't agree with their assessment of and positioning of Rust.

Yes, one can easily use unsafe scope, but it's really an exception and doesn't integrate easily with the rest of the code.

I'm actually developing something now using both Elixir and Rust.

Important topic!

That's neat! I didn't realize that flox was nix based.

So you are competing with e.g. devenv?

The analyst's note must have written with the help of an LLM :)

042?

any tips?

Perhaps new approach is needed, where config settings of terminals, displays managers, etc allow easy composition?
Not a flat file, but logical, reusable blocks? Maybe that can lead to standardization across components (apps) that do similar things?

The "Einstellungseffekt" makes it hard though to anticipate/recognize "easy" solutions to the orchestration challenges.
I'm wondering, whether it's possible to build a full featured solution on top, which'll be a lot simpler than K8S in the end.

Thanks for highlighting this talk! The idea is really powerful and, in hindsight, quite natural in the NixOs world.
It kind of goes in the direction of fly.io - deploying and orchestrating micro VMs vs containers.

OMG, finally!

Especially considering that I was explaining the inner workings of KERI/ACDC :)

If anyone is curious here's the repo github.com/vasiliys/ker...

I made an update recently that covers accounting for users with a self-hosted PDS. Still very high level and basic: chrismcleod.dev/blog/next-st...