Local models exist as part of the solution to privacy invasion. Not saying google has never been nefarious, but the whole point of local models is that your data doesn't leave your device.

Pretty sure from Google's standpoint, respecting privacy is a downside of local models. For them, the upside is using your compute rather than theirs.

I think the accessibility checks only take into account the text color, not the actual real world readability of given text which in this case is impossible to read because of the font weight.

Where did they say the prompt cache is shortened?

from 1h to 5 minutes, was in the news recently

But not as a solid fact?

The HN thread in question is here (and had that info edited out of the title)

https://news.ycombinator.com/item?id=47736476

Fetch has also lacked support for features that xhr has had for over a decade now. For example upload progress. It's slowly catching up though, upload progress is the only thing I'd choose xhr for.

You can pipe through a TransformStream that counts how many bytes you've uploaded, right?

That would show how quickly the data is passing into the native fetch call but doesn’t account for kind of internal buffer it might have, network latency etc

That is a way to approximate it, though I'd be curious to know the semantics compared to xhr - would they both show the same value at the same network lifecycle of a given byte?

This is a pretty widely known acronym

Oauth with mcp is more than just traditional oauth. It allows dynamic client registration among other things, so any mcp client can connect to any mcp server without the developers on either side having to issue client ids, secrets, etc. Obviously a cli could use DCR as well, but afaik nobody really does that, and again, your cli doesn't run in claude or chatgpt.

Can you clarify what you mean?

Title: "California’s New Bill Requires DOJ-Approved 3D Printers That Report on Themselves"

Actual fact: California’s New Bill Requires that 3D Printers Get DOJ Approval as Firearm-Blocking"

(The "report on themselves" is fiction invented by Adafruit.)

"to be able to get a 3D printer" is implied in the "requires" wording. There's no problem with that part.

I actually think the title is misleading. I'm not sure actual existing deployments are affected? Seemingly just new ones are not working?

Railway founder here. <3%.

That said, we treat this exigently seriously!

Any downtime is unacceptable and we'll have a post mortem up in the next couple hours

it seemed to have been all deployments that had a browser facing interface. id say some cloudflair DNS config messup

I've been using railway a while now, and I've basically never paid them but I would. It's even better than heroku. Super easy to use.

Their 5$ monthly has been far more than enough for me to host my demos.

Would like to see the eval version - the dialogue version just seems like normal code with extra steps?

yeah, the previous example was quite basic. I will write a complete example for that, but here is how you can run dynamic code:

   import { task } from "@capsule-run/sdk";

   export default task({
     name: "main",
     compute: "HIGH",
   }, async () => {
     const untrustedCode = "const x = 10; x * 2 + 5;";
     const result = eval(untrustedCode);
     return result;
   });

Hope that helps!

Is the code in the eval also turned into wasm first then? Does this work as a JIT for wasm?

It actually works a bit differently. The eval is executed by the interpreter running inside the isolated wasm sandbox (StarlingMonkey). You can think of it as each sandbox having its own dedicated JavaScript engine.