karmakaze

This is a bundle where you end up paying for parts that aren't as useful:

    - 252 GB HBM3e VRAM
    - 496 GB LPDDR5X RAM

I would rather have a system using 2x Instinct MI350P GPUs (288GB total) for much less.

Personally I'm using Qwen3.8-27B (MXFP4 quant W4A8) locally hosted on a pair of AMD GPUs (with DeepSeek Harness). It starts at 250 tokens/sec down to 120 past 128k context.

At work mostly Opus 4.8 (sometimes a GPT or Gemini 3.1 Pro). I find Opus 5 chatty/slower and Fable can venture into over-engineering itself into unnecessary complications.

Busabase[0]

> Languages: TypeScript 98.4%, Other 1.6%

Not for me.

[0] https://github.com/busabase/busabase

Seems like a random rant to me.

> What is the internet now to me? In many ways it’s something that I have to tolerate to do many things that functioned fine before. I have to use an app to pay for parking. I need to create an online account to pay for a family swim at the local leisure centre. I have to submit an online form to book a slot at the local recycling centre. I need an app to check my bank and credit card balances.

It's so much more annoying to do any of these things without/before internet. I don't do half of the other things in the rest of that paragraph. Like for phones I just get a used CAD$400 Android phone with good audio output.

I think it's largely due to psychology. If 210m is considered the best, then as they approach it they may start to tense up and choke. When the goal and possibility is known as 500m, then there's no point being concerned near 210m.

Similar here peak ~250 and down to ~120 as it gets close to 128k (which is where I set DSH compaction) though it can readily do 256k.

I just got DeepSeek Harness (DSH) set up with 2x R9700 and it's rather mind blowing that these can do actual work and quickly. Up until now I've always been evaluating and searching for better hardware/model/tweaks. This is much more than I even hoped for and considered getting extra 3090/4090. Now I can stop looking/tweaking and start using it for all the different things I've yet to discover it's good for. I do plan to also try/use Hermes and Pi. DSH is annoying that every plugin install/remove requires a restart--given that "everything's a plugin".

The way it does tensor splitting without all-reduce cost over PCIe bus wasn't something I thought was possible.

What kind of performance are you getting with 4x R9700s--what do you do with all the VRAM (batching, concurrent requests, etc)?

Thanks! Didn't expect to see this here. Exactly what I needed to run Qwen3.8-27B-Quark-AWQ-MXFP4-native.gguf as well as other experiments on one or 2x R9700's (I hope).

> Anything design-y was “too Apple” and unacceptably bourgeois.

That was definitely the case of the Unity desktop that really only worked well on netbooks. That's when I lost confidence in Ubuntu for design. Loss in Canonical on the whole came later.

The previous 'personal' AI Station I had pictured was the a16z one[0].

Each MI350P[1] in the TR Halo Station has 144GB VRAM and with 4.6 PFLOPs peak MXFP6 performance.

Four liquid cooled? Yes please.

[0] https://a16z.com/building-a16zs-personal-ai-workstation-with...

[1] https://www.amd.com/en/products/accelerators/instinct/mi350/...

NasaFTs are back in style!

It seems we could use a new kind of memory that streams the weight data in, like GDDR in reverse.

There's clearly a line running through the lake. It should only change name on the lower part.

AirBnb can only be used if you are willing to accept the risk of having an unacceptable or even no reservation upon arrival. The latter did happen to me and fortunately I was able to make alternate plans--though still not recoup full payment.

So it's good when it works, and can ruin your trip otherwise. I have no use for that range of outcome.

Actually no. I listed all the top stories of the day and this one made the cut. Quality is a low bar these days on HN.

Why are they tying the client with the AI model instead of using OpenAI or another popular http format? Running llama.cpp locally is about the right level of complexity for most doing local AI.

> FluidVoice turns rough, rambling speech into polished, ready-to-send text in any app. Free forever, open source, and 100% on-device.

Exactly, architecture for its own sake is debt on day 1.

I would prefer if it showed a regular 12 hours clock face and show the upcoming sunrise or sunset, or two faces one for each.

101 commits. First on August 7th. I don't consider this a serious project.

README doesn't mention anything about how transactions work across the shards.

I see this https://github.com/schapman1974/briskdb/pull/260 but it says shard-local.

Funny, I published information in the hopes that humans would benefit from it. If it happens to be through collective intelligence of LLMs I'm ok with that--even more so if through open models.

I see varying comments here like use AI, don't use AI, etc and I agree with all of it. Don't always use AI, and even when you do, learn the fundamentals and how to use AI to get the output/outcome you want--that means also learning what it is you want and ways that AIs fail.

On top of all that I'd add perhaps having a side hobby in learning about AI research/developments and harnesses (Claude Code/Pi/OpenClaw/NanoClaw/Hermes/etc) and auxiliary systems like vector databases and graph databases. Being an AI wrangler could be the equivalent skill that 'big data' processing was previously. Even just being able to speak competently on these could be an edge.

Yes programming can be an art (a la Knuth)--but artifact applications are not, unlike actual art.

Exceptions include video games where the author is instilling emotion into their creation for the player to experience--not utilities.

Exactly. If one panned out there would be no more.

If FIFA was opensource now would be time to fork it.

I'll wait and see what they do about factoring all the generations of settings so there's one way to do each thing in a single uniform interface.

Or Windows NT 3.51

> Note: This article has been created in natural human language to be read easily, devoid of complex technical jargon, so all readers can enjoy. View one of our other past 70 articles for deeper technical dives.

I would have preferred it was created by a natural human.

> That Node choice wasn’t a close call for us, and it’s worth being direct about why. Most of the graph tools in this space, including code-review-graph, are written in Python. We considered it, and we’re not doing it ever.

So AI generated marketing text for an AI generated product (probably) based on AI generated whitepapers?

Welcome to Web 3.0 I guess.

Running the Netscape or Mozilla browsers would use up 100% cpu, so it was fantastic to have a dual-cpu system so that your system was still 100% responsive.

The thing that I really liked was that you felt like you were programming for the hardware and not really for an OS. Most systems ran a single program at a time so you pretty much had full control except for interrupts. On DOS TSR (terminate, stay resident) interrupt handlers could also do amazing things.

Knowing a lot about my personal computer (Atari 400 and later Atari ST) made it so easy to do work on embedded systems as a student intern. Without internet, learning about your hardware/ROM was paramount from magazines and word-of-mouth.

Another thing that I learned from that time is how to make designs fast and small. This is a skill that's not really taught, shared, or practiced much and makes a huge positive impact on a codebase keeping it light on abstraction depth and high on flexible capability. Huge improvement to maintainability and extension into unforeseen directions too.