bionhoward

Oh im not a c++ guy but that actually looks pretty awesome because it’s like Hoare logic

Google privacy policy is an absolute joke so this is hardly surprising. Pretty much everything gets used for model training. It’s hard to find options that respect your privacy without studying 50 support docs pages. Plus they conflate anonymization with privacy, as if ripping off someone’s data becomes OK as long as you … take their name off em

I feel like if Google simplified their privacy/model training policy then their stock would go up.

It feels so complicated and hard to know what we can use vs what hands them our codebases

That level of idealism only works if the capitalist system heavily changes so the poor can afford to purchase “replicators” which do not exist today. As it is right now if AI/robots take your job you can easily wind up on the street unable to afford food, shelter, medical care

opencode DeepSeek v4.1 Flash isn’t us/eu hosted as of recently, so not sure how this impacts privacy / model training

Mechatronics is cool

“Distilling” is a funny way to say “learning from”

the minimum description length principle encourages us to describe living entities at higher levels of abstraction (as motivated rational agents rather than collections of atoms jiggling)

The universe is absolutely hierarchical and neglecting it just because we don’t strictly need hierarchy to define things seems silly

Looks epic, id like to connect a Dropbox folder

Opus 5.5 does seem competitive with/better than Astra and is more affordable so usage doesn’t run out so fast

That’s not even an attempt to be charitable to jev. It can make decisions from unstructured data. Nobody uses anything just because it is cheap…

whoa

> plan mode has always been a prompt — it has never changed the toolset because doing so would break the prompt cache, and so would be expensive

seems like plan mode could turn off some tools, even if it doesn't change the set offered to the model, the ones that they have which would mutate your codebase could just not work with an error message, and plan mode could change permissions in the security approval prompt for "auto"

anyway, isnt the right way to know if plan mode helps or not, to run an experiment? we're all guessing unless we have data

read only agent mode sounds straightforward and useful to me

huh, i never heard of cosmogony, that's cool, thank you

Brute force is guaranteed to eventually find the most efficient possible solution (in an extremely inefficient manner, assuming you run it long enough)

I feel cosmology ought to grapple more with, “Why is there something rather than nothing?” That question and this question are certainly linked. If expansion is a mirror of collapse, how do we get from some initial nothingness to a point where there is something capable of collapsing? Starting with an infinite amount of stuff everywhere feels like cheating.

Even with efficiency breakthroughs, it would just afford packing more agents per unit of memory. Scaling compute and data keeps paying off, leading to smarter models, and smarter models have more demand even at higher prices because they can accomplish more work at higher quality. Sovereign AI hasn’t even really taken off yet to anywhere near the level it could. That’s going to dramatically increase the number of massive-scale users of AI agents. So no, IMHO they will “never” have “enough.”

Me, Is self-awareness thermodynamically favorable? (2015) https://docs.google.com/document/d/1Ed9ikW47Key-jYZy0CNZH75q... - “the probabilistic nature of reality is what drives Life and forces the evolution of self-aware consciousness”

I feel like Nike’s biggest fumble was not advertising the Invincibles more. It seemed they had some evidence the PEBAX foam could reduce injury in base mileage, which is the number one thing many/most runners spend the most time on. So why not make sure everybody knows this? It seems weird. If you have a lot of athletes who need conditioning and don’t want to get injured, why not advertise the crap out of the injury prevention shoes?

I think the API runs on Azure

I personally think this acquisition is genius anybody who could afford it would be wise to own HuggingFace because they are incredible

I was wrong here, it’s more about the RNG they use for the word choices [1]. Although I wouldn’t be surprised to see UTF-8 substitutions also.

[1] https://www.anthropic.com/news/claude-text-watermark

I think it’s a bunch of hidden UTF-8 character substitutions so the text looks the same but the underlying bytes are different

It’s amazing how much Python punishes you for modularizing your code

I’m heavy into rust and never really use golang, but one big benefit of go over rust is compile times are significantly quicker, which could be more fun if you’re running CI checks 50 billion times

Yeah it’s a coding beast if you get it dialed in right

gotta admit i groaned a bit at this because it would make rust more complicated, but on my 2nd read i realized:

- some things (compile time bounds checking tensor shapes) are hard / impossible to implement now; "pattern types" could be great for that

- however "no panic" is already handled by clippy, might not be much uplift for doing that at a type level.

my 2c: it's great to be excited and brainstorm, some of these ideas might be gold. conveying the benefit is key. it would be good to focus on stuff for which rust doesn't already have a workable solution. i like the pattern types, the rest would take convincing

Anthropic is not trustworthy for this because they force every Claude Code user to agree to a noncompete while also opting them in to model training.

That means, by default, every Claude Code user is actively getting royally screwed

surprisingly accurate! Is Gary the AI equivalent of the “nothing ever happens” guy?

genius, great idea and follow through, please keep it up, this could improve the ML industry tremendously, maybe some einops inspired interface for this would be good?