ac29

6 ms·

91.
by ac29 5mo ago
> Has canned fruit actually lost popularity? Compared to Del Monte's heyday in the previous century? Absolutely. A remarkable amount of fruit is available all year, or most of the year now. I cant imagine eating canned fruit by choi
92.
by ac29 5mo ago
Unsloth Dynamic, just some branding from Unsloth for their quants (other people use similar techniques)
93.
by ac29 5mo ago
Memory and compute/energy overhead
94.
by ac29 5mo ago
Not a great example since using Anthropic subscriptions with third party applications was never allowed, they just didnt take steps to prevent it until recently.
95.
by ac29 5mo ago
> My set up is not the best but it is working for me and is saving me money. I've got a local setup too but unless you consider hardware zero cost, there is really no way to save money. The class of model you can run on <$5k of h
96.
by ac29 5mo ago
Modern GPUs aren't optimized for MoEs though? The advantage to a dense model like this Mistral one is that it is as smart as a much larger MoE model so it can fit on less GPUs. The tradeoff is that it is much slower since it has to rea
97.
by ac29 6mo ago
Payment processing likely eats up at least 2-3% of that
98.
by ac29 6mo ago
Thunderbolt 4 and 5 are just USB (40, 80 Gbps) with mandatory support for otherwise optional USB-C features like video and high power.
99.
by ac29 6mo ago
According to wikipedia the current marketing names for USB are just their speed: USB 5/10/20/40/80 Gbps. No version numbers or anything else.
100.
by ac29 6mo ago
Your table doesn't indicate reasoning vs non-reasoning, or reasoning level
101.
by ac29 6mo ago
Your benchmark has Opus 4.7 performing significantly worse than Sonnet 4.6. Even if true on your benchmark, that is not representative of the overall performance of the models.
102.
by ac29 6mo ago
There are Yoga rooms in terminals 1, 2, and 3
103.
by ac29 6mo ago
I'm using it via OpenCode Go, which claims to only use Zero Data Retention providers. How much you trust any particular provider's claim to not retain data is subjective though.
104.
by ac29 6mo ago
The memory requirements aren't that intense. You can run useful (not frontier) models on a $2-5K machine at reasonable speeds. The capabilities of Qwen3.6 27B or 35B-A3B are dramatically better than what was available even a few month
105.
by ac29 6mo ago
It was a weird point to make in the post given that exe.dev charges $0.07/GB for transfer. That's arguably worse than the major clouds, who charge about the same for egress but give you free ingress.
106.
by ac29 6mo ago
Google's naming might be misleading, currently 3.1 flash image outperforms the available pro version (3.0 pro) on most benchmarks: https://deepmind.google/models/model-cards/gemini-3-1-flash-...
107.
by ac29 6mo ago
From the upstream project: > Can I change the display of all ESLs in a store at once ? No. For two reasons: Unlike radio waves, optical communication must be line-of-sight. Even from wall and ceiling reflections, an unique transmitter ha
108.
by ac29 6mo ago
GLM 5.1 is pretty good, probably the best non-US agentic coding model currently available. But both GLM 5.0 and 5.1 have had issues with availability and performance that makes them frustrating to use. Recently GLM 5.1 was also outputting g
109.
by ac29 6mo ago
> So say someone built an under $10k system, with perhaps dual RTX 5090. That same system will be able to easily run 20 parallel requests. The only cost is electricity. You can run it 24/7. For 1 year, that's ~$6million I dont
110.
by ac29 6mo ago
I agree, but do the potential customers of my business? We need to meet the customer where they are and that means making our site more accessible to search engines, mobile devices, LLMs, or whatever comes next.
111.
by ac29 6mo ago
> What Strix Halo system has unified memory? All of them. The static VRAM allocation is tiny (512MB), most of the memory is unified
112.
by ac29 6mo ago
That was the carrot, but it was followed immediately by the stick (5 hour session limits were halved during peak hours)
113.
by ac29 6mo ago
This 35B-A3B model is 4-5x cheaper than Haiku though, suggesting it would still be cheaper to outsource inference to the cloud vs running locally in your example
114.
by ac29 6mo ago
The very best open models are maybe 3-12 months behind the frontier and are large enough that you need $10k+ of hardware to run them, and a lot more to run them performantly. ROI here is going to be deeply negative vs just using the same mo
115.
by ac29 6mo ago
1M context window is still a separate, non-default model in Claude Code and not included with subscriptions (billed at API rates only)
116.
by ac29 6mo ago
> we can’t have unlimited liabilities stacking up forever The liabilities are completely offset by prepayments from your customers though. Even better, you can earn interest on the deposits without paying any out. If you just dont want t
117.
by ac29 6mo ago
> By comparison gemma-4-E4B-it-GGUF:Q4_K_M scores 15/25 (that is a 4B parameter model!) Gemma 4 E4B is slightly confusingly named, its a 8B param model
118.
by ac29 6mo ago
> Suits in agriculture don't drive the combine either, a farmer does. Advanced RTK based positioning systems have been in Ag for a long time now, so increasingly the farmer doesnt drive either
119.
by ac29 6mo ago
pnpm installs to ~/.local as well
120.
by ac29 6mo ago
This article is about a MoE model with only 4B active parameters, it shouldn't take 10 minutes to answer a question about a small project. I measured a 4bit quant of this model at 1300t/s prefill and ~60t/s decode on Ryzen 39

More ›