bearjaws

IIRC it's thousands per day on these systems, mainly due to high power density and low mass, even small lapses in computation (100-500ms) rapidly change the temperatures of stacked die.

So even a GPU averaging 98% utilization may have thousands of cycles per day.

Compared to a regular server blade it may be dozens or barely any at all.

Surprised to see no acknowledgement of how "AI Menu Slop" has become deeply associated with ChatGPT.

Everywhere I go now I see places that blatantly used ChatGPT for their menus or posters and they all look the same.

It feels almost existential to the service, I know its a bit of survivor bias but so many images you can tell immediately are ChatGPT vs other image providers, and I feel many people are sick of them.

One of the things I keep running into in day to day work is that it seems increasingly thin the labor pools in the USA.

For example, getting developers with strong Healthcare / Pharmacy experience, with robotics to boot is nearly impossible. There's probably only 1,000 in the whole country. So we're forced to train up people. Our pay is quite competitive, for those wondering.

I can only imagine humanoid robotics is in even worse shape, and meanwhile in China theres probably one company with 1,000 engineers in it by itself. With 10 others all competing.

As more and more Boomers, and soon Gen X retire, we're going to be absolutely cooked on skilled labor.

Always funny to me how people expected Google to do anything else other than roll over.

Should be a lesson to the dems next term, either extort the tech giants, the mere threat of the government taking action against Google makes them jump on these dumb name changes. Otherwise, get them on their side to change the law to prevent this kind of shit show ever again.

It's pretty clear you could do serious damage to a corporation as the president and get away with it, and that should change.

China basically has been running a "do nothing, win anyway" campaign since the Trump era.

More and more people view the US as a bad ally, and China is stepping in to help in many places.

People should get out more, go visit South America, Dominican Republic, etc. They have cheaper AC, cheaper cars, cheaper appliances, all because they can import from China without tariffs. This is not to say they have worse quality, I would argue the opposite, much of what they import is at least as good or better than what you get in the USA.

It will be no surprise to me when we end up forced to use equal or worse American AI for a premium in price, while the rest of the world moves on.

The amount of people I see running around at 800k context and wondering why they are burning through tokens is always surprising.

Swear whenever your company decides to invent its own font its like a version of "The Headquarters Curse". You had nothing else better to spend your teams time on so might as well make a new font that 10% of your customers will notice if you are lucky.

They could have just picked any other font and used it. It's a video streaming website, how much text content is there vs video content?

Headquarter Curse: A theory suggesting that companies often experience setbacks or decline after constructing a lavish new headquarters.

The problem wasn't as simple as it caused harm, it was the companies knew it caused harm and did nothing.

Just like how Meta, it's staff and leadership all knew they were making decisions that harmed people, and did nothing to stop it while profiting immensely.

Needs a 0 added, $160B feels fair. Mesothelioma was $35B in 90s money, so about 3x that today.

Companies keep getting away with this "cost of doing business" fines, the justice department keeps failing at its duties.

The fines either need to wipe out these companies to the point the share holders hold the leadership accountable or nothing will change.

Not exactly surprising given it's a dense model at 2.7x the size of the experts in gpt-oss

Recently took a Minisforum 7840hs PC out of rotation as a media PC and made it a full time coding workstation with Proxmox. I do a VM per project due to the nature of agentic editors.

I was using a VM setup on my MBP but it felt like a huge waste, having to leave a laptop on 24/7 when all it did was run Claude Code inside VMs.

I likely will stick with a Macbook Air 15" for next purchase, and beef up my "Claude Server" down the road.

Centralized source code hosting is going to end up looking like the three credit bureaus in terms of security. It's only a matter of time before the first big hack, when everyone shrugs and says, "Oh well, everyone's source code leaked lol too big to fail."

I was hoping to see Cerebras launch something other than GPT-OSS-120b in production this week, especially with GLM4.7 going away.

If they could launch Qwen 27b or Deepseek Flash that would be amazing.

It is funny to me seeing Sol used for what a "traditional" AI model can do already (counting pills).

We have vision models for our pharmacy and I could never imagine taking the latency hit to use a Sol in our robotics, it would be likely 25-50x slower.

Cerebras is crazy to watch on GPT OSS or Gemma, I feel like we need a new VibeOS demo but with Cerebras, the OS would literally build itself in a few seconds.

https://youtu.be/7NfyZhV1dKM?t=52

Sorry not a direct quote, but basically a rip off of "politically incorrect, but nonetheless factually true".

Whatever you want to call your super duper clever rewording of a reductionist statement.

Another week, another agent memory system that is about the same as grep in a memory/ directory.

How long until a one line system prompt ships your entire home folder to a remote server?

Oh whoops. Already happened.

Not sure if you are memeing since this is an Elon quote...

> Interestingly, "politically incorrect" is a double negative that simplifies to "true".

Only if you like generic Twitter quips, logical fallacies and ignoring context for anything remotely nuanced.

Seems like "multiplayer AI" is the buzz word for next year, just starting to ramp up now.

Forget the 395+, I want the MI350P (or similar) long term.

PCIe card is the way forward IMO, AI keeps changing so you don't want static hardware.

Crazy to see how well Qwen3.6 35b-a3b is holding up, sure it is ~20% larger but it's scores are also ~20% higher with the same number of active params (excluding the IFBench).

Hopefully Qwen follows up their 3.8 launch with a new 35b-a3b

People will scream their lungs out about how bad vibe coding is, but there is clearly an intersection approaching where vibe coding will be good enough for many purposes.

Engineers frequently try to be like "bro your site won't scale to 10000 users" and guess what? Nobody cares, nobody working for their company of 100 people is even thinking about what hire #1000 looks like. Not every company needs Slack level scale.

People have spent so long over engineering for scale, they often forget a 3 tier application can survive 99% of companies for over 5 years untouched.

What about the extreme reliance and funding of advertising where even using a free product makes significant cash flow for a company?

It's not good enough now to vote with your wallet, you need to vote with your time and wallet.

Depends on whether the OS actually puts the laptop or sleep or not.

But the people disinterested in starting families (boomers, genx) are going to be the majority of the voting base?

We've already seen 20 years of policy driven by people who don't care about families being started.

These AI companies have found their #1 marketing piece and just beating it to death.

The reality is if they cared about security at all they would provide a way for me to credential myself against my companies environment so I can use the AI on it to improve our security.

You could quite literally make an OS that generates your software ahead of you, like this VibeOS demo but lightning fast.

https://youtu.be/7NfyZhV1dKM?t=53

Imagine this demo but the apps render in real time generating real code.

Obviously not valuable because we have OS today, but could be your companies "WorkOS"

Would love to give my money to a company that links to a domain for sale page.

This really isn't enough IMO.

I recently made a bash script to run in a local directory, boot up a lima VM, setup a tailscale network (I like to access from phone + laptop), and completely segment off everything.

Honestly even with this new setup, I am still paranoid, but its a lot better than what I did before (claude code running on my Mac)