Jon Claassens

iPad modeling tool that won't break the bank: konjur3d.com

Volumetric explosions coming along now. All this shader work needs to be ported to MacOS and PS5/6 eventually, so gathering a big pile of tech debt. This sort of ray marching is a pretty important tool in the bag of shader instruments

@ThePrimeagen Action without understanding is seldom a good idea

I am remembering right now for my own sanity. There is so much peace in deep thought instead of some insane frenetic agent orchestration fever dream

@runaway_vol Hrm, Im still crafting assembler bits around like 5 agents scurrying about doing other things. It doesn't actually know clever tricks to make an arm IR block just that little bit faster. It was trained for satisfaction, not optimality

and we felt oh so diddly important memorizing our silly little syntax shorthands libraries. it ruined a generation of men

@ChShersh @Waffl3x Not Linux. That's someone else's problem.

@JonathanClaass4 @Waffl3x You might be from a different universe. If you talk about Linux kernel, it’s written in C and C doesn’t have exceptions.

If you know what this is, we're automatically friends

@ChShersh @Waffl3x When you write kernel code, exceptions are the end. Panic. Im writing this sort of code now and finding new levels of language gymnastics

There’s a big conceptual debate about exceptions vs errors. Exceptions should be used only for truly exceptional situations. Like you expect a file to be present in a specific location but someone suddenly deleted it while your program was running. Errors as values shine when failure scenarios are part of your domain. Like a user entering a wrong password is not an exception. Surprisingly, it’s a valid use case. Unfortunately, many programmers are really bad in classifying failures in these two categories: errors and exceptions.

@ChShersh It's a neat parallel, but doesn't quite hit like anyhow cargo package... It is interesting how c++ doesnt favor a pattern per se. It opens them all up

@Grady_Booch Welded into training data, no one is free to vanish under the grave, they will proceed like digital HeLa cells immortality imprinted in weights, but trapped in computation

Old way: upon my death, delete my browser history. New way: upon my death delete my LLM conversations.

@yacineMTB Philosophical tautology

No, the difficulty was and still is in building the product. The root of you not having any users is because you cant build a good product

@ashleymayer Sure, our ilk are probably entirely immune to all this and we will benefit. But, this is politically difficult territory

@ashleymayer In fact, someone I know in graphic design was fired in favor of Gemini. He has since had a nervous breakdown. Friends are trying to help. It's going to become a battlefield... I think. But we don't really know.

I'm seeing people trying to wordsmith Eric Schmidt's commencement address (and sure, there was a LOT of room for improvement), but my hot take is that there is no way that he - or *any* leader in tech - could deliver a broadly inspiring pitch for AI in this moment. These speeches are not happening in a vacuum, and anyone who is seen as an architect of this technology revolution has too much to gain to be a credible messenger. I was trying to imagine who might be able to meet this moment. I think it's likely someone outside of tech, who is grappling with the change coming for their craft/industry. Someone who is both excited about what will be possible, and also wary of and clear eyed about the downsides and coming disruption. Someone who is long-term optimistic about human creativity, and can inspire that in young people facing what might be the most tumultuous labor market in human history. Maybe a scientist, a doctor, a creative of some kind. Weirdly, Ben Affleck came to mind given his earlier comments on AI.

@ashleymayer We are probably in labor vs capital territory and that's not good. It never is

@ashleymayer Many of the "fame" have S&P shares benefitting heavily from this rise. Empathy in words alone would convince very few. Measures are probably on the way in many places. Not sure about the US

@ashleymayer Remember also that the US is laissez-faire late stage capitalism. Most of the planet isn't. Racing to a money dream isnt widely viewed as noble. Balance in all things

@ashleymayer Yes, but in this case people tend to prefer rote tasks. They're being taken away in many domains... This isn't a On the Beaches speech moment. It's a time for empathy

@ID_AA_Carmack Running a similar experiment with my mega TCB

@ID_AA_Carmack Aspera is clever?

It is a shame that the simple act of transferring a large block of data as fast as possible over the internet is not handled effectively by the primitive operating system calls. You either multiplex over parallel persistent TCP connections to combat head-of-line blocking and slow starts, or reinvent reliable delivery and flow control over UDP. QUIC has a lot going for it, but it is a large library (six figure LoC!) and conflates security and performance in a way I don’t love. There is also fundamental information about competition with other processes and link layer congestion that should be useful, but is unavailable to user libraries. You should be able to just write(really_big_buffer) and it is all taken care of for you.

@DynamicWebPaige This bottleneck has a lot to do with tool costs. Cadence will burn your bank. Autodesk will blow it away. Alternative CAD systems for high tech manufacture is basically where all my money is going. Need to make that cheaper. Lithography isnt really that hard either

the biggest bottleneck in tech and science right now isn't a lack of tools, it's a massive capability overhang, and it's the most important thing anyone could be working on right now imho devs, scientists, roboticists, data analysts, etc. have access to wildly powerful AI toolchains, but most folks are still stuck in 1:1 chat windows instead of wiring up automations if you want to make an impact right now, bridge this gap: teach people how to actually drive these things, or build the ambient tooling that just does the work in the background without them even having to explicitly ask we've got to help bring people along and send a ladder back down, not leave them behind ♥️

An engineer is supposed to constrain uncertainty. That's the job!

My view on AI code isn't that it codes for you. Rather, use it to make better languages (and tailored) to capture what you are specifying concisely. Nicer algebras and aethetic, clever grammars trump english-to-code uncertainty with agents massively