tkocmathla

It's a fair point, but for any well-established capability, the plugin with the most monthly downloads (overwhelmingly) is generally going to be "the one". For MCP, the top plugin [1] has 1M/mo, and the next one down has less than 2k/mo.

I do agree that in the case of benchmarking, it's a bit silly to put the onus on the benchmark runner to try and pick the right plugin for each capability that won't make someone point to /that/ as the reason Pi didn't do as well as expected.

[1] https://pi.dev/packages/pi-mcp-adapter?name=pi-mcp-adapter

How do you mean, gentrified? I went for the first time early last year and it was an amazing experience. It's definitely very museum-like, and not cheap to enter, but the rooms all look like perfect time capsules and the audio tour is excellent.

That reminds me a bit of The Harmsworth Room [1] in the Churchill War Rooms, where you can dine while you stare at the "original 1940s electric generator wall complete with dials and switches". Very cool space!

[1] https://www.churchillwarroomsvenuehire.co.uk/spaces/harmswor...

It would be really funny, and sociopathic, if every state change registered as a browser history entry so I had to click Back 50,000 times to get back to Hacker News.

Incredible write-up, as usual. I still fondly remember discovering Red Blob Games' Hexagonal Grids [1] guide while building an implementation of the Tzaar board game [2]. The illustrations are enormously helpful!

[1] https://www.redblobgames.com/grids/hexagons

[2] https://boardgamegeek.com/boardgame/31999/tzaar

This is a cool idea, but even more than playing it and learning about NYC, I'm curious to know how you built it, and where the data came from.

At 5 years old, in the US and UK at least, they're either in preschool/kindergarten or not in school at all yet. Grades aren't a thing at that level and I have to imagine the same is true in other areas of the world.

> I'm reminded of an old friend from long ago. She was an early music major at Harvard, and graduated with a MFA. She was very good. She read and wrote Latin and Greek, could compose and play music using medieval notations, and published a book on early needlepoint.

This sounds to me like a life well lived!

What?

https://github.com/apple

Thanks, typo on my part.

The median income in the UK is currently sitting at £2,627 / week or £31,524 / year [1]. This is advertising more than double that at £64,189, not quite graduate wages!

[1] https://www.ons.gov.uk/employmentandlabourmarket/peopleinwor...

EDIT: £2,627 / month, not week!

Reminds me of Microsoft Research's TextWorld [1] (from 2018!).

[1] https://www.microsoft.com/en-us/research/project/textworld

I've been using superpowers [1] for this purpose, and have really appreciated how it guides the model to use careful, methodical approaches to answering my prompts. It's great for multi-step planning, design, and implementation, but also has guidance for debugging, accepting a code review, etc.

[1] https://github.com/obra/superpowers

Yes, they can be anonymous in the sense that you can buy one in person and top it up without an ID [1].

[1] https://tfl.gov.uk/fares/ways-to-pay/where-to-buy-tickets-an...

HEIR [1] is a homomorphic encryption compiler built on modern MLIR.

IREE [2] is very actively developed ML compiler + runtime, also MLIR-based.

[1] https://github.com/google/heir

[2] https://github.com/iree-org/iree

At London Heathrow too, the 100ml limit was scrapped early this year.

https://news.ycombinator.com/item?id=46736815

They briefly address this in the article:

> Violin bowing, the researchers say, is a much more complicated interaction to model.

> AI also forces folks to be online to code

This isn't true in the broad sense you've used. It's true that most people don't have the hardware to run the bleeding-edge foundation models, but with a modest Macbook you can run very capable local models now (at least capable for coding, where my experience is).

The KelpDAO Incident Statement from LayerZero:

https://x.com/LayerZero_Core/status/2046081551574983137

How "fat" are the packed machines? In other words, how much bloat is inevitable, or is that entirely controlled by the base image + the user's smolvm machine spec? How does smolvm's pack compare to something like dockerc [0] in terms of speed and size? Disclaimer: I just learned about dockerc!

I can't actually create and test a pack right now because of [1], but I love the idea of using this to distribute applications you might otherwise use a Docker image for.

[0] https://github.com/NilsIrl/dockerc

[1] https://github.com/smol-machines/smolvm/issues/159

It's very distracting to have every sentence in this article be its own paragraph.

I love this, from a comment on the article:

  He had in his path a script called `\#` that he used to comment out pipe elements like `mycmd1 | \# mycmd2 | mycmd3`. This was how the script was written:
 
  ```
  #!/bin/sh
  cat
  ```

Cool idea, but of course the very first image I clicked into was a dick.

I got this itch too when I came across tinyrenderer [1] and worked through the early lessons through shading, but didn't quite finish the texture mapping yet [2]. It was fun to work in pure C from first principles, even side-questing to write a simple TGA file reader and writer.

I'd be very interested to see your tutorial when it's done!

[1] https://haqr.eu/tinyrenderer

[2] https://github.com/tkocmathla/tinyrenderer

The "locakly" typo is perfectly placed in the comment thread of this article!

And 3 'i's?