Jeff Preshing

Software developer with 15+ years in AAA gamedev. Background in systems programming, concurrency, debugging, optimization and a little audio & 3D

[Jeff Preshing @preshing]

@Love2Code Same thing with GPT-6.1-Sol. It overengineers things all the time. Some people like to pretend it doesn't matter, but I think it depends on where the code lies on the tech stack. It's easier to tolerate brittle high-level code than a brittle foundation.

[Jeff Preshing @preshing]

@davidchalmers42 Do you really think that DeepSeek-V4-Flash, a computer subroutine, "might" literally experience the events described by the tokens it generates?

[Jeff Preshing @preshing]

@juliarturc He has been saying this for a long time. It doesn't hold as much water as it used to now that deep learning is a mature tech used by millions.

[Jeff Preshing @preshing]

@rbhar90 Have any of your friends read @michaelpollan's new book? Why don't they just look at AI models as a kind of digital exocortex instead?

[Jeff Preshing @preshing]

@matiasgoldberg Yeah, LLMs are very logically rigorous now. Personally, I'm OK with saying they "understand" quantum physics better than I do 🙂 It's just a casual way of saying they can generate better text on the subject. Doesn't mean I'm ascribing personhood to it (it's just a subroutine).

[Jeff Preshing @preshing]

@zack_overflow I love agents, but comparing a programmer directly to a model doesn't make sense. You should be comparing a programmer-with-agents to a programmer-without-agents. Instead you're saying that a model is more creative than a person? That's not how it works.

[Jeff Preshing @preshing]

@GeoffreyHuntley Hand-typing part of a commit can still be a good way to provide the agent with context on what to do next.

[Edoardo Vacchi @evacchi]

SOLVING is SOLVED.

ClaudeGPQwen 7 has just DROPPED and it's INSANE.

It's BROVER.

[Jeff Preshing @preshing]

@trq212 Good point

[Jeff Preshing @preshing]

@supahvee1234 I've been spoiling myself with a $20/mo Codex plan and a $20/mo Ollama Cloud plan with access to Kimi K3, GLM 5.3, DeepSeek-v4.1-Flash and other good models. Rate limits haven't been a big problem, even with the occasional batch job, but maybe that's a skill issue on my part 🫤

[gingerBill @TheGingerBill]

Yes. It takes a lot of skill to design a language that is fast to compile by design, and then skill to write and optimize the compiler for that language so that it is fast at compiling too.

But I know that's not what you meant 😉

[Vows ✪ @CloudisTwt]

"The most beautiful thing we can experience is the mysterious. It is the source of all true art and science. He to whom this emotion is a stranger, who can no longer pause to wonder and stand rapt in awe, is as good as dead: his eyes are closed."

— Einstein

[Jeff Preshing @preshing]

@esrtweet @filipcodes It doesn't assume a homunculus. It's just saying that in order for perceptions to appear, appearance must be happening somewhere, which is undeniable. Trying to explain it away as a by-product of neural activity is just distraction.

[Claire Lehmann @clairlemon]

Principal research scientist at Google's Deep Mind 👇
https://t.co/gJDZoYg9eR

[Jeff Preshing @preshing]

@dhh And micromanaging them.

[Jeff Preshing @preshing]

@SamuelePanzeri It's just the latest online distraction. Last week it was all about how RSI could take over the world. All arguments were backwards. It's too bad, because there are plenty of interesting discussions still to be had about the best way to apply LLMs.

[Jeff Preshing @preshing]

@JoshHochschild @paulg @Pontifex It makes more sense to view consciousness as a kind of carrier signal, and the brain as an instrument. AI fits in perfectly with this understanding.

[Jeff Preshing @preshing]

@valigo Very true. Here's a cross-platform C++ runtime in ~14,500 LOC, a Markdown parser in ~5,800 LOC and an agentic harness in another ~9,500 LOC. Agents are right at home in this codebase and can get a lot of work done using little context. https://t.co/1R7sjPkJBS

[Jeff Preshing @preshing]

@theojaffee You don't need one. What you need is an argument against fixating on this idea so much, you lose all objectivity and even cause distress. That's *definitely* not rational. https://t.co/7Pf41KUTRs

[Jeff Preshing @preshing]

I respect Hinton, but he seems too preoccupied with abstract theories about consciousness. There are no good theories of consciousness because consciousness transcends all theories. If you weren't conscious, there would be no theories.

Also, we recognize consciousness in others at a non-verbal level. Babies recognize their parents long before they can theorize about anything.

So when someone theorizes about whether an imaginary AI system could become conscious, and those thoughts create anxiety (or doomerism), that sounds like too much theory to me.