tmikaeld

My biggest issue with Devstral and even their biggest model is that they’re dangerous unless closely directed and reviewed and i mean CLOSELY. Unfortunately mistral models will believe and do anything.

See: https://petergpt.github.io/bullshit-benchmark/viewer/index.v...

See some of the test results, it’s horrifying

Compared to US datacenter buildout, it’s probably much more likely to actually go online..

I don't really see the novelty, you can achieve the same on any VM with SQLite + Websockets for realtime sync of state - and if start time is an issue or even deploy, Alpine linux is just 5MB..

The actual interesting part is the scaling, from 0 to 100 instantly.

depends on usage, Gemma 4 is better on visuals/html/css and language understanding (Which probably plays a role in prompting). But it's worse at code in general compared to Qwen 3.5 27B.

They claim to have qwen 3.5 27B on a card at end of year on the market. If they do, I’ll be buying one immediately.

Great writeup, I’ve been running my own mitm measuring token usage on antigravity since release. In the past week and with the introduction of ”credits”, the amount of usable opus on the 350$ ultra plan was >4000$ it is now <1200$ which means around 4x less usable tokens.

Isn’t it about time they get some competition?

I thought so too, until i saw what writing ~50GB to disk multiple times daily does to SSD lifetime for no good reason.

40-45C

Glue seal can easily loosen at 50 degrees C and a hot shower is 40-45C.. so it must have been very hot shower (or bad glue).

Before you do, note that battery time on Asahi is abysmal at best, so if you're on battery often I'd really reconsider.

I have the same experience. To make it worse, there’s a mile of difference between the all too many versions and efforts..

Warning, the OS doesn’t work with many European banking apps like BankID. If it did, I’d be all over it

Is it just me or is downloading 3MB for the DB runtime plus the database itself, kind of crazy?

At this point, this should be built into the browser which could fetch signed db data and be extremely performant.

So a potential 4-9% difference..

NOT worth it, especially if the whole infra is already using JSON.

It’s difficult to integrate as HBM and the density is very different. China wouldnt select 30x slower bandwith if they had a choice.

HBM stacks multiple DRAM dies vertically on top of each other (typically 4-12 layers). It’s very different from DDR, especially during manufacturing because it has to be integrated with custom chips. Afaik, this isn’t available to Chinese foundries.. yet.

Thus, they have DDR available. Specifically, theyre using LPDDR5 so far.

I expected that this was because China use DDR5 in their new AI chips (Due to not having access to (enough) HBM)

Just an FYI, the open source version FLUX.2-DEV cannot be used commercially.

https://huggingface.co/black-forest-labs/FLUX.2-dev/blob/mai...

I’m in the same position, but I use AI to get a second opinion. Try it by using the proper models, like Gemini 3 Pro that was just released and include grounding. Don’t use the free models, you’ll be surprised at how valuable it can be.

There is a setting to disable telemetry, unclear what this means though.

They expect me to walk around with my phone in a long sock? A fannypack even looks more appealing.. also, isn’t this a sign that phones are getting too big?

Swedish

Status: Endangered

"The child-bearing generation can use the language among themselves, but it is seldom being transmitted to children."

What!? A lot must have changed in one generation..

It doesn't, this is also the reason that streamers like Nvidia Shield or Apple TV are the only two choices if you want to view 4K content at all.

This seems to have been huge, not just one datacenter. All of my VPSs and company dedicated servers where unreachable, including load balancers..

It really is the criminals paradise, thus the record amount of shootings and rapes.

The Obi for your Wan

I also tried it for a full 100K credits (Wasted in 2 hours btw which is silly!).

Compared to both Udio and Suno, it's very very bad.. both at compositions, matching lyrics to music, keeping tempo and as soon as there's any distorted instruments like guitars or live, quality goes to radio-level.

The security model is very underestimated imo, it will be very evident when more bun projects reach production and not experimental.

But extremely expensive in comparison