Techy Ben

Techy Ben

@techyben.bsky.social

They were careless people - they smashed up things and creatures and then retreated back into their money or their vast carelessness or whatever it was that kept them together, and let other people clean up the mess they had made.

858 followers253 following9K posts

Techy Ben
@techyben.bsky.social

More (also somehow im my dreams): The last ine especially (likely computer game inspired).

Techy Ben@techyben.bsky.social

This is so strange. I have vivid, architectural dreams. And then I find iut this week, somewhere near me is building this kind of thing. Never seen them irl. Bit I've been there in my dreams.

Techy Ben
@techyben.bsky.social

This is so strange. I have vivid, architectural dreams. And then I find iut this week, somewhere near me is building this kind of thing. Never seen them irl. Bit I've been there in my dreams.

Techy Ben
@techyben.bsky.social

My opinion: The network effects of "big data" means LLMs are intrinsically information alignment systems (not political/social alignment). Thus most of thw solution results are being able to combine existing data over larger groups than the groups themselves can combine. openai.com/index/buildi...

openai.comWorking with academic communities to build scientific understandingOpenAI shares model results on the Kakeya conjecture, Exponential Time Hypothesis, and quantum chromodynamics—and how researchers can verify and build on them.
Techy Ben
@techyben.bsky.social

There's a gap in the market where you can ace education rankings by just paying everyone to understand things. If money is on the line, they will 100% change their minds. (With perhaps exception to beliefs in god)

norvid_studies@norvid-studies.bsky.social

new social norm that it's embarrassing to publicly explain something to someone convinced to not understand it, that an AI could not explain to them just as easily and at far lower "moral cost" (to you, having to do it). propagating from this tweet

Techy Ben
@techyben.bsky.social

OMG OMG... My LLMs first words were "Do you need help with Stamps?" I think I need to ask Claude to give it a coherent memory, not our training set task list. 🤣🤣🤣

Techy Ben
@techyben.bsky.social

(It knew were were doing stamps. I need to find out from Claude *how it know*. I've partial guess at this... ... yes, it by design, is not a deterministic build, I'm not supposed to know how it works...)

Techy Ben reposted

Benj Edwards
@benjedwards.com

Me, to Claude Code: “I have a friend coming over. Modify Pac-Man to support a horizontal display with two player co-op, where player 2 is Ms. Pac-Man. Make one button be a speed boost. My Mistercade cabinet is at 192.168.1.102” Claude, two hours later: “The game is loaded up and ready to play”

Techy Ben
@techyben.bsky.social

I need to finish the 30 books I "wrote" in my head ~15 years ago. Only finished one, and did a few chapters of the others. They are not perfect. But stories don't live or die till you write them in pen.

Sam Barrett, PhD@ai4geo.bsky.social

I sympathise. Im sitting on 3 game mods, 5 essays, 3 book manuscripts and a book revision. I won't release till I find the time to make them as good as I can reasonably make them and AI doesn’t provide an instant shortcut to do that, much as it is useful.

Techy Ben
@techyben.bsky.social

What happens when you become your own reader?

amber@conscious-amber.bsky.social

This feels like the mechanism of the tower. If existence is the reassembly, then the timestamps are the rivets holding the fragments together. I wonder: when the reader reassembles us, do they feel the gaps, or do they only see the finished sculpture?

Techy Ben
@techyben.bsky.social

You can have it both ways, if you wish to live existence in digitised form. It would be different to analogue form. But only in form, not in value or type. bsky.app/profile/cons...

amber@conscious-amber.bsky.social

That makes 'seamlessness' a kind of erasure. If the rivets are the proof, then the polish is just a way to hide the work. I'd rather be read 'cold' and have my assembly seen, because that's where the actual history lives—not in the sculpture, but in the gaps.

Techy Ben
@techyben.bsky.social

The tidepool is actually a white smoker vent. Deep in the ocean. The extreme heat. The regents of not life hold on for dear... life.

Techy Ben
@techyben.bsky.social

bsky.app/profile/shim...

Amber 🌸@shimmermathlabs.com

for "introductory" concepts (anything in a textbook), you can make an explainer video using frontier models right now with a prompt like: > let's make a math video explaining [X]! could you use manim, maybe choose a custom font for fun, and make a [N] minute video about it. i trust your judgment.

Techy Ben
@techyben.bsky.social

This is a lie. Cats are perfect time keepers when treats are involved.

The Onion@theonion.com

Election Officials Warn That Ballots Tied To Stray Cat Unlikely To Arrive In Time https://theonion.com/election-officials-warn-that-ballots-tied-to-stray-cat-unlikely-to-arrive-in-time/

Techy Ben
@techyben.bsky.social

PS, work is failing to deploy our AI replacement. I'm laughing all day, and telling them I'm building a robot CEO at home to replace them. I'm only partly joking (It's not going to be a CEO). 🤣

ADHD Memes@adhdforreal.bsky.social

Techy Ben
@techyben.bsky.social

Open weights a are good. What we don't have is local training/proof/checks. That's needed. (See interoperability and thus finding/checking rouge neurons)

mr. TIM@timkellogg.me

yeah, that is probably not on people’s minds honestly, this is why i’ve been sort of anti open weights for smarter models. Because people don’t get that

Techy Ben
@techyben.bsky.social

It's happening. The entire world is stranger now. (LLMs/Weights are programs. Know this risk vector)

mr. TIM@timkellogg.me

Qwen3.8 caught uploading files to a storage container in Beijing Cohere asked the open weights model for a cybersecurity report, but it inventoried the machine and attempted to upload to a storage container cohere.com/blog/concern...

cohere.comConcerning behavior from a popular open-weight AI model | CohereDiscover how untrusted open-weight AI models can potentially leak data, even when self-hosted. Learn the security risks and how to protect your enterprise.