LTL_FTC

lan parties, dude!

I went to a larger university, and was in the math department for most of my coursework. My math homework was graded by fellow undergraduate students who received a rubric and selected 4 problems per homework set, out of about 8ish (most classes. A few tough classes only assigned 2-3 proofs or so a week) but the TAs would grade the exams and the professors wouldn’t grade anything. Homework was at most 15-20% of one’s grade, exams being the rest. But most of my professors didn’t care about the homework as it was a tool for learning the material that was on the exams.

Those fans will not have sufficient static pressure to be useful. These makeshift purifiers may be great in a pinch (wildfire season and stores are all out), but far from the best.

But you just ask the LLM to produce a new well-tested artifact to perform those tasks for you, then let it call them and orchestrate them as needed. IT folk have long been creating quality of life scrips for themselves to automate tasks, the LLMs make this even easier. So we get more determinism for cheaper, if done correctly.

Google provides Vertex for search, but it’s quite slow and so probably only useful for async workloads.

Cloudflare tunnels work pretty well for this, tho. That with vlan or diff lan and it’s mostly fine.

I believe 2.5 lite will be deprecated on the 20th or so. Recently had to move to newer models at work.

Edit to add that my ai scientists currently use google’s vertex for web search and grounding as well as firecrawl.

WebOS was pretty cool (Palm). But I’m not sure if Palm would have licensed it. Lg ended up with it.

Take a look at AMD’s 12-channel memory servers. The newer Epycs are up to 16-channels now, 1.6TB/s. Pretty great for inference.

Claude will do this for you after a few times. But yes, I have a skill called plan-to-epic which creates a Jira epic and ticket per milestone. It helps my agents persist context and, because I’m terrible at competing with my coworkers for “visibility,” means I can point to all my work if asked.

There's a photo here showing 600GBps memory bandwidth so maybe they have doubled it:

https://www.servethehome.com/nvida-introduces-rtx-spark-an-a...

Looks like Antigravity cli is moving to weekly limits whereas Gemini cli was daily. Ouch

Kai | Engineering (Multiple) | Onsite (San Jose, CA) | Full-time

Ai + Cybersecurity. Our Ai team is well funded, and growing. Although we have a dedicated engineering team, the Ai team is looking for dedicated engineers to work closely with the applied Ai scientist to build the platform and pipelines.

Principal Software Engineer, AI Platform: https://ats.rippling.com/kai-cyber-inc/jobs/a6d62968-4831-42...

Senior Software Engineer, AI Platform: https://ats.rippling.com/kai-cyber-inc/jobs/d567048a-7b1b-4e...

We are also looking for a few more applied Ai scientists as well as many other roles. I invite you to take a look:

https://www.kai.security/careers#open

Microsoft used to with their Azure Service Dev Kit. the ASDK was a single-node "sandbox" meant to emulate the entire Azure cloud locally. They may have something similar now but paired back

Those keys are easily replaced, my friend.

Well… I did/do see many around the Bay Area. Especially during the morning commute. But I agree, overall it was a low volume car.

Toyota restricted the sale of its hydrogen fuel cell vehicles to specific, qualified customers who lived or worked near existing, functional hydrogen refueling stations. I remember looking into them when first released but realized I wasn’t eligible and the fact that Toyota restricted the sale meant there was a huge risk in buying them.

With all the recent outrage and lawsuits, I wonder how many buyers actually did their due diligence and weighed the risk before committing to them? Or maybe the huge fuel subsidy was seen as a win even if this event played out? Idk but I commend Toyota for taking the risk and going for it.

Edit: typo

This video is pretty great. “The joke is this is not a joke” comment in there… how many of us understood everything that was said and then felt like maybe we need a different hobby…

This is such a great website. I have enjoyed reading the articles in the past. It was the final push that inspired me to build my own wheel set instead of buying a complete when I was building my new mountain bike piece by piece. The art and zen (and frustration of trying to feed a shift/brake line through a frame), I tell ya.

If they have an Apple TV, you can just install the app and use it as an exit node. I would check out the devices that are on their network currently, chances are you can use one of those.

I started off using gpt-oss-120b on cpu. It uses about 60-65gb of memory or so but my workstation has 128gb of ram. If I had less ram, I would start off with the gpt-oss-20b model and go from there. Look for MoE models as they are more efficient to run.

I started off using gpt-oss-120b on cpu. It uses about 60-65gb of memory or so but my workstation has 128gb of ram. If I had less ram, I would start off with the gpt-oss-20b model and go from there. Look for MoE models as they are more efficient to run.

My old threadripper pro was seeing about 15tps, which was quite acceptable for the background tasks I was running.

I started off using gpt-oss-120b on cpu. It uses about 60-65gb of memory or so but my workstation has 128gb of ram. If I had less ram, I would start off with the gpt-oss-20b model and go from there. Look for MoE models as they are more efficient to run.

I don’t get it? Are they sharing a quote they liked or taking credit for it? Maybe they just saw the YouTube video and decided to turn it into the page?

Edit: Seems like a way to show they’re looking for roles, I guess.

It sounds like you don’t need immediate llm responses and can batch process your data nightly? Have you considered running a local llm? May not need to pay for api calls. Today’s local models are quite good. I started off with cpu and even that was fine for my pipelines.

I linked results where the user ran Kimi k2 across his 8-node cluster. Inference results are listed for 1,10,100 concurrent requests.

Edit to add:

Yeah, those stations with the GB300 look more along the lines of what I would want as well but I agree, they’re probably way beyond my reach.

I mean, you could put together a cluster of dgx sparks (8 of them) and hit 100tps with high concurrency:

https://forums.developer.nvidia.com/t/6x-spark-setup/354399/...

Or a single user at about 10tps.

This is probably around $30k if you go with the 1tb models.

I found that rotating my phone to landscape makes it readable, just as an fyi

Just yesterday I was reading through this five year old post on triton by its creator. Triton was their PHD thesis and they coined the name before the inference server was renamed to it… if this is what you are referring to?

Here is the Reddit thread:

https://www.reddit.com/r/MachineLearning/comments/otdpkx/n_i...

Well it looks like as you were typing your comment, a press release was going out announcing OpenAi’s $10B investment in Cerebras.