nsarrazin

Used to work on a chat app where we had full control of the stack from the GPUs to the chat interface and everything in between. We were a small team too so I could be pretty confident that nothing changed in the stack and we still regularly had users complain that this or that model got nerfed. Perceived performance is actual performance over expectations and the latter just keeps increasing over time.

It doesn’t mean the big labs don’t also nerf models! But if they didn’t you’d still have users complaining.

To be completely fair to the birds, I think a lot of this comes from the fact that what matters for flying efficiency gets better as you size up thanks to the square-cube law: the drag is reduced and the engine efficiency increases.

You could not scale down a plane to the size of a bird and match their efficiency. On the other hand you also couldn’t scale a bird to the size of a plane so I guess we can call it even!

Did some quick maths, it’s roughly 10% for the bird (2W, approx. 4g/day of fat, over 10 days) so not negligible but doesn’t change the full story.

And you could argue the plane also has a baseline metabolism ;)

> I doubt our flying machines are anywhere near as energy efficient as birds

This kind of nerd sniped me, so I looked into it and turns out no, planes are more efficient than birds!

You have to look at the energy spent for moving a fixed mass over a fixed distance and for the most efficient birds I looked into, that’s about 300g of fat burned for 300g of “dry weight” for a trip of ~11,000km. (bar-tailed godwits)

For a plane that’s roughly 80 tonnes of kerosene for 170 tonnes dry weight on a similar distance (787). Energy density of fat and kerosene are actually pretty close so you can just look at the kg of fuel burned per kg of payload for the trip. For a bird that’s 1kg of fuel per kg transported and for the plane that’s ~0.5kg per kg. So roughly twice as efficient.

So from a pure energy point of view it’d be more efficient to put all the migratory birds in a 787 than letting them fly on their own!

To me the main appeal of this is not that it makes cutting easier (you already reduce much of the cutting force needed if you slice across instead of pushing down, which this seems to be doing at a micrometer level) but the fact that food doesn't stick to the blade.

All of my knives have sticking issues, potato slices are the worst culprits. When doing a lot of prep work on veggies in a cramped area it's a pain and this seems like it could actually fix things?

OTOH, I'm not a huge fan of the design, I don't mind spending good money on a knife but between the blade profile and the chonky handle, it looks like an hybrid between an 8" fibrox and an electric turkey cutter.

If there was a premium release with a fancier handle (could wood work?) and a nicer blade profile (something like a gyuto or a french sabatier) I would be interested and I think it would reduce sticker shock (looks premium, expect premium price). I love the design of the charging tile though!

You can find the env path with `uv python find --script "${filePath}`

I've been working on a VSCode extension to automatically detect and activate the right interpreter: https://marketplace.visualstudio.com/items?itemName=nsarrazi...

On my free time I've been making a small "game" where you control a spacecraft using a 6502 computer. You can control things like thrusters, antennas and read data from star sensors, gyros, etc. using memory mapped I/O directly and you can't control anything directly.

The whole thing works pretty well, it's built using Godot & a custom Rust extension with the 6502 emulator & assembler. It can even be exported to the web so you can write your code using a CodeMirror editor and send it to the game in the browser.

Now, as expected, the hard part is making it fun. Just giving players an empty code editor and a manual for the sensors/actuators doesn't seem great, and there's also no goal for now. I'd like to take inspiration from Shenzhen I/O to build a sense of progression by tackling more and more difficult control challenges maybe?

It also supports running multiple computers at once so I think there's some fun potential for a "multiplayer" mode where you try to achieve some objective (scanning X, mining Y) against other players by sending spacecrafts with your code on them and all the crafts would run on one server. Anyway I'm curious if people would be interested in such a game!

We just released the latest version of the Llama 3.1 models on HuggingChat! We’re currently hosting the 405B at FP8 and the 70B at full precision. They’re compatible with our tools, so that means you can use them with web search, PDF parsing and image generation for example.

Excited to see what the community thinks of them, let us know if they work well for your use cases. The 405B is pretty good at some advanced reasoning tasks, it's quite impressive actually!

Try them out here:

- 405B: https://huggingface.co/chat/models/meta-llama/Meta-Llama-3.1...

- 70B: https://huggingface.co/chat/models/meta-llama/Meta-Llama-3.1...

HuggingChat is still working though! https://hf.co/chat/