sroussey

AI often goes the screen reader route though..

Or a drink with deep mind employee…

Yes, I agree, but I caught myself touching the screen twice this week forgetting it was a Mac...

I've been watching these robots carve stone on instagram for years now...

Deepmind creates these, Google Cloud productizes them.

If Google Cloud was in charge, Gemini 5 would be public, and they wouldn't be messing with Gemini 4 nor 1M context when they have 10M internal. C'est la vie.

Perhaps the cost is higher… for Micron.

Privilege escalation is nothing to worry about on root accounts

Constrained decoding came to JS land transformers last month for in browser structured output as well, also used for tool calling.

https://github.com/huggingface/transformers.js/blob/main/pac...

VCs will teach companies all about the wonders of regulatory capture!

Yes, I pay my HOA fees via bill pay so they write a check and send it next door via the mail. This is Wells Fargo.

Also, they remove the money when they mail the check… not when it gets cashed.

I never found a use for anything but single app mode on my iPad Pro.

That explains the 5x speed increases huggingface mentioned for Mac using metal.

Really just instrumentation schemes, I would think.

When you code, do you understand the asm code being built? What are the cache line access patterns? Most software developers have already given up the details of how things work, and lean on tools to handle them.

Which is fine, usually. Depends on what you are doing.

The things that always mattered were the core system, the trunk of a tree if you will. The branches less so, and the leaves not at all (though that is what people see and appreciate! The irony!).

Oh, no, a condo building.

And language does not map to every idea.

The idea of using language for reasoning about protein folding borders on ludicrous. Surely there is latent space more appropriate and (human) language has no expression for it.

My HOA pays the buildings water and sewer and trash. Also fixes the roof over the condo building, cleans the common interior, and does repairs (typically on our shared plumbing), and the elevators.

I run iCloud private relay and I get very few of their captchas. Actually, I think they are concentrated on just a couple of sites.

There is another project that also converted the go version to rust and got 2x performance like bun did, but will also generate .d.ts files where bun does not.

Edit: found it: https://github.com/tunnckoCoreHQ/tsgolint-rust

Why no Jinja-format chat template?

How does it compare to the same port that bun did and shipped last week?

Thanks! I’ll have a look and run some evals.

Note: this is checking only! Still need tsgo/tsc7...

Also: bun

Even that one repo is breaking GitHub this week…

I made a harness too, and recently added eval with opencode and pi and I sat between them. No gaming other than trying to figure out why I was really bad on some evals (I was not returning thinking tokens which broke cache, for example, and different model families do things way different here). I have yet to try my harness for coding but I do have a few terabytes for it to read. And I’d rather use a local model that’s fast and good at retention, but that run could take a year, lol.

Sometimes you want both. I wonder if there is a postgresql reranker…

Oh interesting… I have the entire SEC database to index (only indexing the text in a sliver of it at the moment) and use tsvector and I could use something faster.

That’s only going to work on an MRL embedding.

But there are some magical compression algos on the horizon, with pitches to cut RAM by 95%. And I’m all for it!

We are truly in the brute force phase of AI. Like a database without indexes, we are doing full table scans to pick 2 rows.