When CUDA started taking off we had ATI hardware, to support their open source pledge, and looked into ROCm.

Just getting the drivers to build on EL-anything was an extreme effort, and it wasn't my first rodeo.

Without betraying confidences, I was told second-hand that there were only ten people on the GPU driver team across all platforms and that they were doing their best and not sleeping enough as it was, with Compute way behind gaming bugs on the priority list.

I couldn't independently verify of course but the theory fit the data.

I immediately empathized with the suffering of the devs and went out and bought nVidia cards, annoying binary drivers and all.

Since then I've felt like that some bean counter at AMD wrote nVidia a trillion dollar check.

If you're not a tiny company *overstaff* your engineering departments so you don't miss new opportunities as they arise. The opportunity costs exceed the opex costs.

Same here but this lack of support will matter much less than dropping i486.

There are still embedded systems sold today that only meet i486 specs. I don't use them but some industries do.

Sure a $12 ESP32 can handle those tasks but it's a revalidation thing.

Not that anybody from those vendors stepped forward to maintain a tree.

OTR might be small enough.

The Axolotl Ratchet is much better but perhaps too big for SMS.

TextSecure probably would have offered it were that feasible.

Of course if you can arrange one-time pads you're 1:1 at 140 characters.

Really this is far more probable than a dozen people went crazy in the same way at the same time which seems to be a plurality of comments.

If there's an American great ape they live in Washington state, not Ohio, based on report frequency.

Star Wars nerds? Yes, so many in Ohio.

His best buddy specialized in making exploding pagers.

Good luck out there.

In my younger and more foolish days I had a Pontiac and I opted out with wire cutters to the Surveillance module's power cables.

At the time I was actually more concerned with remote unlock hijacking than tracking but still I didn't trust GM.

All together now: WE TOLD YOU SO.

If I had to guess 20 years later doing that would disable the ECU.

Yeah, this sounds like some kind of jaded Transhumanist humiliation ritual.

If true as written, those 'monks' are without honor or reverence.

Are there any tools to watch my lsmod, walk /lib/modules, and configure modprobe.d to deny loading anything I'm not using?

All of these exploits are in distro modules I'd never use.

Local access, unless you haven't patched apache this week, then it's remote access.

> What are normal people supposed to do?

It looks like a omen to take up farming.

Maybe I should pull some of my 4TB nvme sticks and buy my kid a used car...

It's not md5 collisions that are the problem here, it's poor passwords.

Which is something that we can't actually fix, and maybe have tried for the past 22 years.

As mentioned above, memory-hard algos are better than hashes now.

I did some math the other day on running local AI models and the net result is most homes can't afford to run the current median models.

They don't just need 80GB of VRAM, they need newer architectures - to be supported by CUDA, to be supported by pytorch, etc.

These problems may well be solvable with more clever use of hardware, MoE, acceptable quantization, etc., but today you're in for several grand and something north of 100W idle to use what is effectively a $20/mo plan.

A small enterprise can afford local, so that's good. We paid more than that for one SGI machine back in the day.

The point of the exercise was to plot the position on the curve. We're at something like 2006 YouTube where nobody could afford the drives or bandwidth that YouTube/Google was giving away for free (aka with VC money). Eventually hard drives got cheaper, people got gigabit at home, FlashServer was replaced with h.264/HTML5, phones could stabilize video locally, etc.

So it looks like these AI companies need to stay alive for about seven more years giving away product at a loss, or at least highly oversubscribed, to turn a profit. Hence the low token allowance, the banning of OpenClaw, etc.

On the other hand, I read the blog of a security researcher yesterday who found an exploit with (IIRC) Claude, tried to refine the PoC, but got dinged on "out of tokens" before he could finalize it. So he just deleted the work and moved on.

It sounds like they're trying to not lose money at such a velocity and are trying to find a sweet spot where people don't just declare it too underpowered to use.

A global energy depression may well take out the supermajority of the companies that believe they can burn investment money for seven more years. There is circular financing money, then there is real return on capital money. One is to fool the markets, the other is grounded in current physics.

Dawkins: Claude, say, "I'm alive!"
Claude: I'm alive!
Dawkins: Oh my GOD!

Would it have cost more than $250M to get a Siri LLM working?

Maybe Tim was afraid of another Apple Maps and bailed?

Who wrote it?

Sounds like some janky offshored slop.

Maybe it's janky American slop, but it would be good to know.