happyPersonR

My we should crowd fund distillation with 2 bit quants on a deepseek 4.1 flash and see how it does ? Bet it does well ..

Requires the vpn provider to snitch and possible tag the ip frames or http frames lol

Folks would just host their own vpns various places and this would be pointless ….

Penpot has an mcp… might be time to take a look.

some but not all:

they did " Mass-Scale Regression Testing: Leveraging our internal data processing infrastructure, we validated the rewrite against a dataset of over 30 million GIFs. This ensured that the new implementation produced results identical to the original version across a large set of real-world inputs."

didn't say they did anything with a VLM to check the output vs the original library

"Differential Fuzzing: We implemented a differential fuzzer that exercised the original C and new Rust implementations side-by-side. In over six days of continuous execution, the fuzzer performed over 200 million iterations without identifying any logic deviations."

this just checks parity/equivalence.. i'm talking about new fuzzing just for the new rust library as a harness. This isn't that.

'Adversarial AI Analysis: We utilized specialized LLM prompts to perform "adversarial reviews," asking models to identify subtle behavioral differences between the two codebases that might escape traditional testing.'

this is just looking at the code...

i'm sure they probably did all the other things google's internal standards required.. but tbh, i would hope google are updating and enhancing those for the agentic harness era. Now that we can we should :)

i'm glad they had a skeptic

might be overkill/flawed , but maybe something that checks to see using a VLM that the GIF's that are generated/modified etc using the old lib and the new one look similar ish at least?

if you have google's tpu's ... could probably be done

also an agentic flow that generate's gif's that might break the C version or actually cause pointer type exploits and see if similar approaches work or don't in the rust version?

maybe even an agentic red team that tries to do bad things to this through gif's to the library and see what it does?

just thoughts... need coffee haha :)

lol time to buy some fpga’s … even if they’re slow

I never thought we would talk about VCD's again LOLLL

nice to see something about this technology again. It was pervasive across SE asia for a little :).

The best is “it’s going to create jobs”….

What kind of jobs? Best case: “temporary construction jobs” lol

This is exactly what’s missing from enterprise deployments of Linux.

Teardowns of dji stuff is important to see how far behind we’ve fell

From an electrical engineering perspective, the one thing that strikes me is just how cheaply some of the engineering that goes into FPV drones actually is.

Like everything is thick wires hand soldered

If you look at DJI and their PCBs they use flexible PCBs for a ton and sensible sane connectors. Getting the board design, the ESD, back EMF right probably took them a ton of time in design and validation but once they did … they stamped it out.

It can be done again and might actually make sense to do as an opensource effort. Even with the fully assembled PCB … it wouldn’t be terrible.

Could totally be done in the US, but I doubt for things other than military applications a US company is going to put the time in to do something like that. In DJIs case, they were at the very least probably state subsidized early on.

We don’t have a similar push specifically in the US other than for defense which tells me our country still isn’t serious about any of this.

man... that sucks :(

There are still ISP's that REALLLLLLY don't wanna think about ipv6 LOL

and i'm not totally sure why?

sorry you have to put up with that :(

what's wild about this(not coming at you :), just in general wanted to post ) .... folks pick various details but i think sometimes miss the big picture. If they can make current/next gen HBM with similarish PUE... idk that DUV vs EUV vs 5nm vs 3nm matters?

I think you’re wrong ….. I gave output for a conclusion that I checked from a tool that I had used in order to clearly show I had used it. I also checked these conclusions on my own.

This doesn’t actually meet any definition of AI slop? Other than I put the word “deepseek” in there.

You not understanding how a tool works or what AI slop is and just instantly getting triggered is the bigger issue.

You’re over applying the term and probably need to go outside for a walk.

lol there’s supposed to be …. But then greed gets in the way lol.

Usually it’s a combo of a tier 1 that decides that a resell agreement is “good enough” and “well just eat the loss” and then by the time stuff like this rolls around “we’ll get back to you” + a bunch of silly explanations that boil down to “you’re not gonna sue us though” start coming out lol

I’m going to say something … and it might trigger folks

Learning and getting good at shell scripting is something that takes a while usually. I would recommend that people new at shell scripting actually try to use Python instead writing shell scripts where possible.

I’ve seen people not use traps, not know when to use set properly, xargs if you don’t know what you’re doing with it is just a giant foot cannon, and folks new at shell scripting don’t know how to debug running shell processes.

They may have learned some or all of these things for python?(please? )

lol the answer is actually correct

Stop getting big mad dude

“Senators stripped the bill of two key provisions”

Which senators ?

I asked deepseek

“ Short answer: if you need a name for who sponsored the amendment that stripped the DUNS identifier and extended the timeline to five years, it's Sen. Bill Cassidy (R‑La.), via the committee substitute. Sen. Jon Husted (R‑Ohio) sponsored the other filed amendment, which addressed address disclosure rather than the identifier, and which did not make it into the reported bill.”

it's interesting to watch all of this having dealt recently with behvaior in an environment that expected

1) SLAAC/static ip 2) multiple default gateways, and then you import routes(including multiple default gateway's) over bgp and start BFD to figure out which one is active 3) and then you start broadcasting your own address over bgp

i wish there were a daemon in systemd, that i could configure to do this. It seems absurd to people that aren't aware of how it works, but it's quite the interesting flow in a datacenter.

wish there were an equivalent in the kubernetes/normal linux world, but i can understand the scalability concerns :)

I know BIRD exists, but i wish this was kind of built in ... lemme see if systemd supports extensions... it would be cool to add :)

i'm sure more people might want it.

lol $12k is basically the main problem

Tbh I actually like this even more than writing a driver …

Makes it so if there’s an issue I don’t get a kernel panic

Maybe it could write something CUPS compliant though ?

If people read nothing else, they need to read this. Now that they have EUV etc CXMT and others in China will scale much much faster. They always had the EE’s and hardware people. I’m excited to see what happens.

Or they’ll just take the easy money and join the price fixing cartel

Or they’ll run the others out of business by doing a Walmart and then just jack up prices.

And watch … $5 says they have nothing to do with the ram because even if there was enough power there’s no use case for this much spend

If I had to guess they’re just buying up ram to keep others from having it same way meta, Google, etc hired up engineers to keep others from having them

Hospitals, now that private equity is involved, do this weird cost shifting accounting BS with shell companies etc, as such their books aren’t straight forward and the 2-5% thing is likely greatly underestimated given the amount of understaffing PE has driven in that space and how much gouging there is from PBM etc.

The other thing about rFC is that you can use it on things that have issues being tested in LAL :)

Def quite nice

Some of the manufacturers of endotoxin testing kits charge less for rFC as well :)

LOL

Although they may internally use LLM, I wanna say said they were not going to allow their core offering to be integrated with LLM

Although maybe they reversed course?

you're right ... but china hahaha

Yo this would be cool to have embedded as a PR trigger on GitHub for 3d files

And maybe as a preview option between branches ?

$5 says it's because of "blah blah blah sabre is based in a single datacenter" .... and it went down

it was an amazing system for its time, just grew too complicated over the years

>” To be honest, I think I'd remain happy if they just stayed at the level they are now, maybe bit faster and cheaper (and ideally more energy efficient), best would be to run it locally too, but I don't really need them to get "better" to make good use of them at this point.”

I have bad news … you’re gonna be mad… they’re gonna make it expensive otherwise they can’t get the hockey stick ROI.

Are there any benchmarks for this? Maybe a good next step :)