k__

While it's true that unsupervised AIs build a heap of trash, I have to admit, it doesn't look much worse than some of the code I saw powering products that sold for millions.

The only thing that changed is the scale.

k__··on Satellite Tracker – Live Map of Starlink and 30k Satellites

Strange, there aren't even 20 at a time above Germany.

How many uplinks can one satellite handle?

k__··on Muse Spark 1.1

Yeah, DeepSeek V4 (Flash and Pro) is below $0.004 for 1M cache hits.

Even with usage based billing I'm below $1 writing code all day.

k__··on Europe's new climate in seven charts

"Europe's rapid warming is partly the result of [...] a drop in the number of tiny polluting particles in the air. This means that less of the Sun's energy is reflected back into space, leaving more energy to heat the Earth's surface."

Hmmm...

k__··on Kimi K2.7 Code is generally available in GitHub Copilot

I tried DS4 Pro yesterday for one particular task that Flash struggled with.

That one Pro task cost $1.50 and the whole day of using Flash cost me $5.

k__··on Kimi K2.7 Code is generally available in GitHub Copilot

Thanks!

I'll try that.

k__··on Kimi K2.7 Code is generally available in GitHub Copilot

Yeah, v4 flash is dirt cheap, but it's running in circles quite often.

Might very well be that a better model is cheaper if it gets things right the first try.

Maybe I should route to a better model when v4flash hasn't solved after a specific number of tokens.

k__··on Kimi K2.7 Code is generally available in GitHub Copilot

Nice.

I paid $6 yesterday for DeepSeek V4 Flash on OpenRouter. That's like $120 dollar for a month, and it's not even a good model.

k__··on Qwen 3.6 27B is the sweet spot for local development

Llama3.1 instruct seems to be doing okay on that page, mostly because it's dirt cheap.

k__··on Tidal AI Policy

So, their incentive is to promote AI music, since they don't have to pay royalties for them.

k__··on Nim Conf 2026 (Online, Sat June 20)

Did Nim 3 go all-in on compile time memory management?

k__··on AMD silently removes memory encryption from consumer Ryzen CPUs

I'm curious about Denuvo's opinion on that.

k__··on DeepSeek Introduces Vision

Maybe, you could pipe it through T5 or something.

k__··on Running local models is good now

Training DeepSeek was magnitudes cheaper than training the SOTA models it relied upon.

In theory, other countries should be able to replicate that effort and improve it.

k__··on Has AI already killed self-help nonfiction books?

As I understood it, they were only a US thing anyway. Because of healthcare issues etc.

k__··on Running local models is good now

I tried some smaller Gemma4 and Qwen3.6 quants on my MBA with M5/16GB and had like 20-60 tokens per second. At 60 it felt pretty okay and that hardware is on the lower end.

I'd assume a Mac with 32-64GB memory would get some reasonable results.

k__··on Can Europe train a frontier AI model on the compute it owns?

I thought the joke was that people aren't paying enough money.

k__··on Don't trust large context windows

100K seems quite much.

I had the impression, models would get inconsistent after just 3000 words.

k__··on Notes from the Mistral AI Now Summit

Why doesn't Mistral distill?

k__··on Expertise in the age of AI

I tried it with embedded programming, and failed miserably.

k__··on Dehydration's role in learning and memory

200ml every waking hour?