jasonjmcghee

Congrats Charlie!

This is very cool.

It needs spacers / extensions so it can be laid out as desired.

Many professors there didn't use homework as part of your grades when I attended. I think the reasoning at the time was students would just copy each other's work anyway or look up solutions online etc

But if you didn't do the homework / learn the material you'd be screwed on the mid-terms / final which was 100% of your grade.

I feel like the link to be used here should be https://claude.com/resources/articles/claude-duet instead.

My guess is this is meant to be a no friction VNC (though it doesn't seem to be using VNC). So you can be away from home and remote in.

There are VNC clients on iphone / ipad - but generally it's whole desktop.

RealVNC is the name brand but it's not exactly a good experience. "Screens 5" looks like a better alternative.

Fair - I should have

NIST makes sense, but co-opting Alpha is a bit odd.

Alpha historically had meant "Google DeepMind" not RL

It's the companion blog post to an Nvidia / UW paper of an approach to solving learning based high precision robotics tasks.

HN guidelines are to use the title of the blog post and not editorialize.

And there are even hibernateable websockets too - only paying for when data is actually transferring.

And the maximum worker time is based on processing time, not wall time.

The native / built-in API SpeechAnalyzer (introduced last generation on iOS and MacOS) works quite well.

Are folks finding it lacking / needing alternatives?

I sure hope that's not true. Maybe hit the comments section first?

I'm guessing you're getting downvoted due to dismissive tone, but I think there's truth to this.

Labeling something as "don't escalate" that should have been isn't great. Paying 5x and having to wait 10s instead of 100ms or whatever for a reasoning model (with tools?) is very likely worth it.

Different APIs for different things reminds me of the early auto-complete vs instruction apis.

Will this get folded into models / post training pipelines at some point and make them better at calibrated outputs?

Or even just- they've already procured OpenAI.

That's a big motivator.

Had the same thought - feels chart crime adjacent

> All three support Zero Data Retention for requests made through Cloudflare

And then on the providers page:

    Property              Value
    provider              exa
    Zero Data Retention   No

Last time I was using llama cpp you could just do:

    llama_state_save_file

or

    llama_state_seq_save_file

and the load equivalents.

That was a year or so ago though...

This and the Meta VR stuff - I would be all over it if it weren't Meta. It's objectively cool tech.

But you're handcuffed.

Exactly - this kind of borders on market influence.

Effectively: "Now's the time to invest..."

> YouTube spends way too much time chasing their competitors

Is A/B testing a thing on TikTok and/or Instagram (or similar platforms)?

Gemini 1.5 Pro claimed 10M input tokens before release.

And was 2M tokens IIRC after release.

There were also many rumors that Gemini 4 was going back to 2M. Just seems odd not to say what it is.

> 1M output token limit

what about input?

(Maybe I missed it)

Lots of focus on the negatives, but it's rather uplifting to see that fathers are spending 4x as much time with their kids.

In my experience it's not close and the benchmarks I've seen don't reflect my experience at all.

But I'm guessing people will find the right training regime and data mix soon to close the gap.

But big things I see are instability and inaccuracy - like pick a random problem.

They are not equal or better. Try them out on real things. I did - it's not close. The benchmarks are a misrepresentation.

That being said- hard to believe OpenAI etc couldn't build a frontier classifier too.

For the uninitiated, I can't recommend enough, The Illustrated Transformer:

https://jalammar.github.io/illustrated-transformer/

For what it's worth - over the last few years or whatever, it seems like Anthropic benchmaxxes the least.

That being said, I currently prefer Sol / Astra to Opus / Fable as I find both to be a better cost payoff to me.

I think the reason is in the general ballpark of people throwing LLMs at a huge variety of problems and being too slow disqualifies them from a bunch of things.

Now there's a new training-free thing that is fast enough to be useful on a new class of problem.

If you have a little data and can ask a frontier LLM to train a model, you can probably beat it on average for a specific task.

But... This is the case with LLMs too.

TypeSafe is still the provider. It's just proxied.

So it's a new subprocessor. Which can often be painful to onboard, especially if not compliant according to your needs.

The open source ones- I downloaded a number and tried them and compared to Jev.

Anything that required knowledge / familiarity mmBERT and ModernBERT post-trains performed much worse.

So it seems like they did some kind of useful expansive pre-training.

Things that were Qwen or Gemma Diffusion did better at those kinds of tasks but were generally pretty inconsistent in terms of whether they could succeed repeatedly (and be stable + reliable) on the many types of tasks that are in the cookbook part of the Jev docs.

If you ask Jev similar input + questions, it's pretty stable. And does a reasonable job on a lot of questions.

This one public benchmark (the only I've seen) seems to give the open versions way too much credit. It wasn't my experience at all.

It gave my a false wrong sense of what might be required to get it working for something at work to avoid needing a new subprocessor as - at least on Cloudflare / OpenRouter Jev is third-party not hosted.