functional_dev

BTW magpies have passed the mirror test

valid point, it reminds me of video games. GPUs got faster, devs pushed higher resolutions, more complex lighting instead of saving power :)

spot on... that difference between evaluation and storage is exactly why C++ is so hard to keep in my head

I thought constexpr was a hard physical constant, but in reality it's a weird hybrid

this visualisation helped me to wrap my head around it - https://vectree.io/c/c-constness-and-evaluation-qualifiers

100% agree

Have you tried to feed "tsc" error messages to Opus? That should reset model's logic

imo one huge benefit is that ipv6 is much easer for hardware to process.. ipv4 headers are messy and change size.

ipv6 headers are 40 bytes! routers have less thinking.

this visualisation might help to dive deeper - https://vectree.io/c/ipv6

maybe they are considered direct link between heaven and earth :)

consumer car needs hundreds of pounds of reinforcement... just to pass side impact test

yeah, hexane is the industry standard for extraction... even bio oils are tied to petroleum.

here is comparison on extraction physics vs chemistry for turning it into biodiesel - https://vectree.io/compare/biodiesel-chemical-engineering-vs...

even if traning is hitting a wall I think they are shifting more to reasoning phase to get better results... and that is inference compute scaling

What about rust const fn()? I think it guarantees there are no side effects

are you using vector store for long term persistence or standard fs and model interaces with it via tools?

i wonder what is the limit of quantization when it starts to destroy the logic of weights?

does this handle the timing differences between browser execution and real time hardware? or is it focused on logic verification?

exactly, that map looks like a mess of random blocks...

big blocks are registration groups (countries) and squares inside are registrants (publishers). like a hierarchy. this visualisation helped me to put pieces together - https://vectree.io/c/isbn

wasnt PEP 740 an attempt to solve this?

The shape argument works well in small packages but it starts to fail once you have multiple domain models starting with the same letter

trick is maintaining enough domain expertise... so we can actually audit those shallow outputs.

If the baseline knowledge drops too low we cannot tell when the AI is being lazy or wrong

exactly, and they are cheap, right?

Why 600km exactly?

here is Wayland vs X11 visualised.. it might help - https://vectree.io/c/compositing-window-management-architect...

maybe humanoids because tools already designed for human likes? if you send specialised robot it cannot use wrench or climb a ladder... you have to redesign everything

I am not expert, but I heard brains learn way better when we actually use our hands to write stuff out.

like the logic sticks deeper in your head that way... using computer is fast, but sometimes it just goes in one ear and out the other

Incredible work! Fitting transformer into 32KB RAM is crazy

For those who read this project and do not know PDP-11 it could be hard to understand that working with these memory limits is difficult. Here is visual guide for PDP11 architecture - https://vectree.io/c/pdp-11-hardware-architecture

Thanks for this amazing project!

I learned recently that white sand is mostly parrotfish poop... they eat coral and then it becomes sand

TLDR https://vectree.io/c/anatomy-of-the-claude-folder

you are right! Power management improvements are what really enable these form factors... being able to run a wifi sensor on a coin cell for a year makes applications possible that were unthinkable just a few years ago

theory is one thing but the cpu cache is the real bottleneck here... here is a small visual breakdown of how these arrays look in memory and why pointer chasing is so expensive compared to the actual logic: https://vectree.io/c/json-array-memory-indexing

basically the double jump to find values in the heap is what slows down these tools most

seafoam green choice was also influenced by eye rest studies... since our eyes are most sensitive to middle wavelengths. just keep the room dimmer without losing detail. It reduces fatigue for operators on long shifts.

Anyhow, Elixir has it's rebirth now due to Nx