buildbot

After reading the paper and intro, I personally have more questions than when I started!

I mean, I can't remember what I had for dinner 6 months ago, but if I snuck into a presentation and anyone famous was announced as the surprise guest, I think I would remember that! I would assume most people do not recall everything with this much detail, just the (ha) memorable things.

It has! Hypernetwok/weight generation has been a thing for a long time - in fact my Master's thesis back in 2018 was based around PRNG weight generation for large percentages of the network. (https://open.library.ubc.ca/media/stream/pdf/24/1.0371928/3)

Why is this not overhyped in my opinion? Nothing has worked this well, ever.

That makes sense! As an IC all I saw was the public number.

Microsoft basically does the same with internal Azure services? When I was there our budget/usage was based on and consumed at the consumer price for GPUs and other compute. What happened behind the scenes really I have no knowledge.

Thanks for being a voice of reason. Every single comment above this is racist and really should be flagged/dead. Disgusting.

AI slop…

It might need to become a realistic view that >0% of people follow

Wow this is stunningly beautiful and extremely cool. And I learned something I had never heard about.

BRB going to go make hyperbolic outer billards art

I strongly believe all people are equal, and we should not apply different ethics based on national borders, race, sex, etc. They should be able to come work any jobs, and be treated the same as citizens.

Given $$$$ that might actually be reasonably possible to do if you can get a good enough model of the chassis.

Yeah the “brassing” is really cool

They can pay them less, treat them worse, then if there are any issues, deport them. It's very gross.

Mostly? https://www.reddit.com/r/thalassophobia/comments/18qr1pv/nor...

That’s more for comfort than anything I would think, a modern cruise ship can easily take pretty insane storms? They have active stabilization systems and huge amounts of mass while still powered by relatively strong engines?

Foveon while a good idea, is extremely complicated to get right from what I understand.

I imagine that different silicon fabs or even different runs of the same fab have slightly different absorption on each focal plan in the sensor and calibrating that at high quality is tricky? Can you even do BSI? I have no idea

> Does it still clog up?

Yes, even very high end printers like an Epson p5370 - https://luminous-landscape.com/the-epson-sc-p5370-printer/ "The printhead is the same relatively maintenance-light device that P900 owners (and predecessor P3880/3800 owners) really appreciate. These heads don’t clog much unless you leave the printer idle over long intervals and/or sit the printer in a very dry environment."

> Has printing quality improved?

Depends on the printer? A professional inkjet is extremely high quality these days.

>Or are they still a bag of hurt we should all just hope for better and cheaper colour laser / LED printer.

In a distant past I was tech support help desk person, which included poking the many color laser printers we had. They also actually broke all the time :(

New Foveon is a long ways away still, they basically scraped what they had -

https://petapixel.com/2025/10/27/sigma-says-full-frame-foveo...

“But there are — to be honest, the project has been a little bit delayed… because every time we run the wafer and get the prototype sensor, we found some technical issues because we have never designed a sensor by ourselves."

Ever looked into digital backs? Basically what you are describing. Sometimes minus battery - the (old) mega vision backs are just a sensor, FPGA, and firewire cable out to a host!

So basically anyone who used this CA is screwed if someone knows your email?

Tangential but people might think it's neat - there's reusable indium pads for replacing thermal paste - https://www.indium.com/products/thermal-interface-materials/...

They are used on the Supermicro Power9 platform weirdly.

We did a little demo/have a patent for a similar idea using idle GPUs on Mercedes vehicles once upon a time - https://patents.google.com/patent/US20230281046A1/en

> BitNet 2B model running on FPGA Very cool! What kind of FPGA & What kind of TPS are you hitting?

It was really something to be able to go in and interact with systems from decades ago. Huge loss that it was closed.

Ah, I always wondered how Vancouver Island came to be - a fight between Stephenson and Gibson makes perfect sense.

Sadly the $ per byte of xSPI and HyperRAM quite high

Plenty of experience ;)

I’ve observed exactly these patterns with Opus and Fable as well - for example, forgetting that they can edit files and instead use python scripts as a patching tool…

Downside of the MTIA chips is no open backend - so quite a lift to get them working if possible at all :(