Isn't the post more about the interesting capabilities of the language rather than an engineering context?

It's entirely possible that I'm being unnecessarily pedantic, but this doesn't feel like a post about "capabilities" so much as "hacky workarounds".

Maybe my years of contributing to undisciplined codebases has made me bitter but TypeScript has ways to solve these problems and it hurts to see more attention brought to hacks than the interesting (and underdocumented) solutions TypeScript already has. Given that roughly half of the documentation is in the damn changelog, TypeScript is especially in need of independent content that highlight features (not hacks) of the language

What methods would you want to bring more attention to?

I remember the days of countless insane, boutique javascript libraries. React's ubiquity is a victory - let's not encourage throwing something away in the name of some vague notion of "innovation".

Does performing gradient descent on token input embeddings lead to interpretable results? And if not, why?

After working pretty closely with vLLM and SGLang over the past few months, this is EXACTLY what I had envisioned what a successor project would look like - analyzing an operation dependency graph and then fusing (or, at a minimum, scheduling tasks smarter). Congrats to the team.

Thanks a lot for your positive feedback! We believe that MPK can enhance existing LLM serving systems, especially for low-latency LLM serving. We are very excited about the opportunity to collaborate with others on direction.

It is unclear to me is whether Kent recovered his cognitive ability or not. He says he has, "thanks to the genie" (which seems to be his name for coding LLMs). Does this mean that with a genie, he is now able to produce like he was before? Or that he used a genie to help "re-train" himself?

Robots that can do anything.

Is it though? He's perfectly positioned to do so.

As I've posted elsewhere in the comments here, I think there are good solutions, but it will take some cooperation to get them off the ground (adoption of standards, licenses, etc.). Here is what I've proposed, which comes pretty close to what the CAI (Content Authenticity Initiative) is doing: https://dev.to/kylepena/addressing-the-threat-of-deep-fakes-...

I've got some good solutions on this front, I think: https://dev.to/kylepena/addressing-the-threat-of-deep-fakes-...

Watermarking is not the way to go. It relies on the honesty of the producers, and watermarks can be easily stripped. With images, the way to go is detect authentic images, not fake ones. I've written about this extensively: https://dev.to/kylepena/addressing-the-threat-of-deep-fakes-...