kristianp

Tangentially for Go programming, the last time I looked at optimising some Go code with SIMD there were a few different options available, but they were either not maintained any more or had incomplete support and required first writing your function in C++ with intrinsics and generating assembly, then converting it to go assembly with a tool [1]. I never got my function to work in go despite the C++ code working fine. In short, not really a production ready option for Go. This was a year or two ago, though.

Edit, there's now an experimental official library at https://go.dev/pkg/simd/archsimd/ see https://go.dev/doc/go1.26#simd and at https://github.com/golang/go/issues/78902 so things have moved since I tried it last.

[1] https://github.com/minio/c2goasm

So you have to create a group for your testers, but there's a good chance that it will be suspended by Google's automated systems. And your testers may also have to wait hours after joining the group before they can download the app. What an abysmal process.

Why is stated:

a lot of apps run 100% of traffic through openrouter while for us it's a small subset

but companies we were trying to work with used openrouter as a source of truth on popularity so they thought products that were much smaller were bigger than us

The remake isn't going to be exactly the same as the original. It is sensible to want more photos of the new units.

https://xcancel.com/kaleybrauer/status/2078185882926846044

Points to study at https://arxiv.org/abs/2607.03502

The image of the cover says 1977, I'd say the date is an error.

The youngest boomers are 62 this year. Not many would have living parents. Are you referring to inheritance they may have gotten?

identity is a local BIP39 seed phrase that signs requests to verify user sessions without backend, no email, no account. continuity

All those little phrases pieced together. Seems very LLM written to me, perhaps Claude Sonnet? It's also quite hard to understand what that means until you realise it's a local-first phone app.

It says double-sided tape.

Not really, the example shows when the output is different/truncated.

1976 at the West Coast Computer Faire.

According to wikipedia, the first "Faire" was in 1977 (where the Apple II was demonstrated). https://en.wikipedia.org/wiki/West_Coast_Computer_Faire

Apparently Apple is "in talks" with the PrismML: https://www.cnbc.com/2026/07/14/apple-prismml-ai-compression...

Interesting, surf is based off WebKit2.

https://surf.suckless.org/

At least the repos are still there, for now. Time to get forking.

Who was it that said: "worse is better"?

This guy: https://dreamsongs.com/WorseIsBetter.html

Numbers like that buy a model a real migration effort.

Such a silly choice of words. I wish the human directing the LLM writing the article put some effort into rewriting the worst examples of LLM style.

But it did extremely well, and the promise was immediate and specific: builds finishing in less than half the wall-clock time, at 27% lower cost, scoring at or above our incumbent on completed work.

The way the LLMs write (Claude perhaps?) With short phrases separated by colons, commas or full stops, is so poor and frustrating.

There some good insights behind this article, so it's worth reading, for example below, but it isn't easy to read.

Earlier GPT models cached implicitly on partial prefix matches, which gave decent hit rates for free. GPT-5.6 dropped partial-prefix matching:

Ok. but why are you posting a spoiler?

The increased nuclear mass causes orbiting electrons to speed up to a significant fraction of the speed of light, where the rules of Einstein’s theory of relativity are important.

In the relativistic regime, an electron’s spin — the magnetic moment that points either up or down — and the electron’s orbit are no longer independent of each other, a state known as spin-orbit coupling.

Interesting stuff. I've never heard of sigma or pi bonds.

https://www.science.org/doi/10.1126/science.aei1285

I you look at https://arxiv.org/pdf/2401.04088 table 5 on page 8, you'll see that expert(s) used can change from token to token. The experts aren't divided along predictable lines.

This is a little site I've had a for a while now. I've added a new page that compares the prices and memory bandwidth of machines with RAM shared between CPU and GPU. I've only listed machines based on DGX Spark, AMD Max+ 395 and Mac Mini and Studios, as they all have sufficient memory bandwidth and size to run LLMs.

There was a recent comparison at https://aimultiple.com/dgx-spark-alternatives discussed at https://news.ycombinator.com/item?id=48732679 . My site aims to give up to date prices.

Just about every release of go has incremental performance improvements [1]. It's a very stable language, so don't expect anything mind-blowing. For example the latest 1.26 moved a new garbage collector from experimental to production [2]. In the latest compiler they state that: "The compiler can now allocate the backing store for slices on the stack in more situations, which improves performance".

[1] https://go.dev/doc/devel/release

[2] https://go.dev/doc/go1.26

https://news.ycombinator.com/item?id=48840113 seems to answer that question. I.e. an api for the compiler is being worked on.

Yes, the way the URLs are interspersed between short passages makes reading it difficult. It's like reading an article on a news site without an adblocker.

I'm not familiar with .net on Mac, what binding did you use to make this? .NET MAUI? https://learn.microsoft.com/en-us/dotnet/maui/developer-tool...

where it's easy to install a current OS and browser in 15+ year old computers

That's true unless you're using Windows or Mac!

https://en.wikipedia.org/wiki/Id_Tech

If you look at https://openrouter.ai/z-ai/glm-5.2#providers there's about 28 providers, including z.ai and Alibaba. Most outside of China. I've never seen so many providers for a model on there before, glm 5.2 is popular.

It loses to Intel in CPU

Not really, their CPU market share has dropped in the last decade.

The Mac Studio M3 Ultra with 96GB of RAM with 1Tb SSD costs $6799, but has triple the memory bandwidth [1]. It's almost twice the price of Strix Halo and when the M5 Ultras come out it will be interesting to see how much they go for. I expect a lot of demand for them!

There was a post recently, that showed the DGX spark is twice as fast as the M3 Ultra at prompt processing, but half as fast at output tokens [2]. They used gps-oss-120 for that test with a small context.

[1] https://gpuquicklist.com/apus?models=GB10%20Grace%20Blackwel...

[2] https://aimultiple.com/dgx-spark-alternatives , https://news.ycombinator.com/item?id=48732679

Interesting insight into Notch's creativity before Minecraft was released, although a public alpha was released beforehand in 2009.

Kind of a shame that Java applet support was removed from browsers. Obviously the security problems made it too much work to maintain.