ekr

Now up on the newsletter: Understanding Age Assurance Accuracy

https://educatedguesswork.org/posts/age-assurance-accuracy/

@esoterra@hachyderm.io Obviously, you're free to disagree with these evaluations, but it's worth working through what you think Mozilla ought to have done instead and whether that's really practical. In this vein, I think it's notable that Chromium has taken a pretty similar tactic of gradually incorporating Rust into Blink rather than doing a Rust only rewrite. 4/4.

@esoterra@hachyderm.io By the time this question was really ripe, it was possible to just write new Rust code in Gecko, and so Servo was no longer needed as a platform to develop that kind of functionality. 3/N

@esoterra@hachyderm.io Even the most mature pieces of Servo took quite a bit of work to move into Gecko, and much of that work was in refining functionality that was not really production ready. To get Servo to the point where it was a replacement for Gecko would have been extraordinarily expensive and time consuming and would have had a big negative impact on Mozilla's ability to maintain Gecko in the meantime. 2/N

@esoterra@hachyderm.io I'm not sure what other decisions Mozilla made has to do with this question. As far as Servo goes, I think there certainly were people who envisioned that it was going to just replace Gecko, but what I can tell you is that this question was looked at quite closely and we came to the conclusion that that wasn't practical. 1/N

Servo served its purpose of incubating components which eventually ended up in Firefox, but once that had happened, then there really wasn't a reason for Mozilla to maintain two different engines.

@soc@chaos.social @esoterra@hachyderm.io This is not really a good take. Servo was a research prototype and a number of pieces of Servo were eventually pulled into Firefox (specifically WebRender and Stylo). However, Gecko, like any modern Web engine, is enormous, and replacing all of it with new Rust code was never practical. Instead, Mozilla opted to maintain the existing C++ engine and gradually transition towards writing new stuff in Rust where plausible.

This week on the newsletter: "How not to mandate device-based age assurance"

https://educatedguesswork.org/posts/device-based-age-assurance/

In this post, we examine a number of enacted or proposed requirements for device-based age assurance and some of the ways they can go wrong.

Now up on the newsletter: Let's build a tool-using agent

In this post, I walk through in some detail how AI agent tool calling works, including digging into the inputs and outputs of the LLM before they get translated into a coherent-looking JS API.

https://educatedguesswork.org/posts/tool-calling/

Now up on the newsletter: Now up on the newsletter. "I automatically generated minutes for five years of IETF meetings (using AI)"

https://educatedguesswork.org/posts/ietf-minutes/

Now up on the newsletter, the final post in my series on memory management: Understanding Memory Management, Part 7: Advanced Garbage Collection

https://educatedguesswork.org/posts/memory-management-7/

Now up on the newsletter, part 6 in my series on memory management, this time covering basic garbage collection techniques.

https://t.co/NwzlH5oOC2