sharms

I have 10000+ inventory items to categorize but I need an intelligent model (not just if statements). Using LLMs has been slow and expensive and I needed to queue it to run for hours. Jev did it in minutes and for less than 1 cent

I am running GLM 5.3 across 2x DGX Sparks and was doing comparisons and it absolutely can beat Gemini. Yesterday it corrected a poor Fable 5 response even

Interested to see what Apple does in response - Siri AI has privacy protections while OpenAI would ingest and use this data (or at least that is what ChatGPT tells me)

I would love to see more of this mindset - bring something new and unique in to the world, instead of mass produced, least price per unit mentality, like how car colors are trending towards black and white.

We got a 17Kwh array this year (US) and a few powerwalls and it is one of my favorite purchases. I can run nearly any amount of GPUs for free, and never worry about power outages or what temperature to set the AC to. If anyone is on the fence about Solar, I highly recommend it

Good call out - but seeing 2.5W consumption at idle from people with it already on Linux so these numbers will hold (like Dell XPS 14)

This is great - a Macbook Pro for Linux users, made of CNC milled aluminum, haptic trackpad, and 20+ hours of 4k video playback under Linux

This is because the "thinking" you see is a summary by a highly quantized model - not the actual model, to mask these tokens

This is awesome! Does it work on most G14's or are there specific years / models?

AWS actually has two libraries they use instead: s2n and aws-lc https://github.com/aws/s2n-tls https://github.com/aws/aws-lc

The complaints about Apple are from decades of excellent design and about a pixel being off or other small items that people with well trained eyes spot. The problems with Windows are forcing you to run Onedrive and then deleting your files

Yes they have an Nvidia image and I just used it on a 5080 last weekend, worked perfect

Created https://www.spreadcheer.net - a Christmas list app that can store locally, isn't full of ads, no login required and should be pretty fast

FWIW I bought the M4 max with 128GB and it is useful for local LLMs for OCR, I don't find it as useful for coding (ala Codex / Claude Code) with local LLMs. I find that even with GPT 5 / Claude 4.5 Sonnet that trust is low, and local LLMs can lower that just enough to not be as useful. The heat is also a factor - Apple makes great hardware, but I don't believe it is designed for continuous usage the way a desktop is.

Thank you! Planetscale has been great so far especially the console where it shows the query insights and the p95 / p99 views. I have been able to make features without worrying about the database because it just works

I would imagine having more karma if I was AI

Missing from this announcement is language around Privacy. Cursor for example has a Privacy Mode that promises not to store code, and this seems like a critical feature for any AI enhanced dev tools.

Hey Lee - I tried setting up a self hosted POC today and it looks pretty tricky:

  * The shared cache loader is left to the user to implement (instead of say, just pass a Redis URL)
  * The revalidate XYZ doesn't seem to work with Cloudfront etc
  * The image lambda is left to the user
  * The Cache-Control header isn't correct

Ironically when you are not paying attention, it is the least safe to disengage FSD. Ideally all drivers not paying attention would have FSD safety features auto engage to stop them from causing accidents.

Autopilot is a standard Tesla feature which is different than full self driving or FSD.

I just drove the Tesla Model Y around where this past weekend it was -10F. Range was probably 15% lower overall (lots of heating / heated seats etc) and we were cautious to make sure we stayed charged up (huge traffic delays due to the storm meant you could be stuck in your car for 6-10 hours). Overall everything worked out, I did have battery anxiety but there were also gas cars which had anxiety.

It looks like these models won't be as useful for LLM inference which are heavily memory bandwidth constrained. The Macbook Pro page shows M3 at 100GB/s,150GB/s, and 300GB/s vs M2 at 200GB/s and 400GB/s. 400GB/s is available for M3 if you opt for the high gpu config, but interesting to see it go down across all of these models.

Not OP but I think it isn't always about storing, but having a log of events which get routed, processed, and aggregated in many cases.

I grew up North of Ohio and it was common to joke about driving down to Ohio as being more rural / farm stereotypes. Later I moved to Ohio and found out it was a great state with a larger economy (top 7) compared to most of the states which make fun of it.

This might not help and might sound new age / abstract: meditation is not continuous focus, but is about choosing to focus, letting thoughts pass, and when you recognize you are distracted to consciously let the distracting thought go, and return to focusing on what you intended to.

I have been using nat.dev to compare quantized models and it works great.