oatsandsugar

It looks like no LLM has seen a pelican before at all.

A taxonomy of bread bag tags :)

As in ETFs that include no oil or oil processing companies

That's super! Do y'all have no petroleum etfs on offer?

This book was a book I always looked at on my dad's bookshelf, and which I read after he passed away. The distinguishing between bullshit and lying is one I think is more and more critical.

I'm at a very explosive part of Akira in my life. Feels right.

*Greater Toronto Area, not the GTA we were all thinking of

Incredibly effective data storytelling in this article. Found this via Kottke.

Absolutely gorgeous imagery, and it seems to have a functional purpose as well, as a digital twin for structural modeling.

Incredible work.

The work in the related stories are equally gorgeous. Thanks for sharing mate.

They trained a model to create embeddings that were a 1024 dimension vector, with each vector being a floating point with 32 bits.

This gave them a baseline performance of 100% with an embedding size of 4,096 bytes.

They then experimented with lopping off the second half of the embedding, leaving 512 dimensions, at 2048 bytes.

They also experimented with just flattening each dimension to 1 bit, 0 or 1 (0 for negative, 1 for positive), reducing the size of the embedding to a minuscule 128 bytes.

Counterintuitively, the "binary" simplification was not only way smaller, but ended up being slightly more performant (96.46% c.f. 95.22%).

This result is wild to me.

* Gravity isn't changing, this is about the science of gravity propagating

Sure, also if the framework you are using makes it easier to create logs that are helpful

*imperceptibly

I wonder how much less secrecy they have in their vote.

I definitely think the complete lack of ergonomics is a much worse sin than the charging port.

This is a bit depressing to me. If you live in a community where you don't trust your children's friends and their families enough for a sleep-over, maybe that's not the best community to live in.

Easy example is clear errors and logging making it easier to run, debug and amend software. That's part of the DX, but not necessarily part of the DX of writing the software for the first time.

I question the thesis that "writing" code productively is the only part of DX. When I consider DX, I consider that code is written less often than it is amended, than it is read, and than it is run.

Getting all of those things right is DX, not just writing.

Wow, this is his whole life.

the "importing voters" thing? That is absolute garbage

This feels like a non-financial markets, betting kind of business that I don't love is mixed in with financial products.

Incredible, thank you for sending that over!

The level of meat in this article is higher than expected, I especially liked the extraction of corporate use-cases from job postings.

Tasks were thinking makes human worse

> Three such cases are implicit statistical learning, visual recognition, and classifying with patterns containing exceptions.

Fascinating that our lizard brains are better at implicit statistical reasoning

I mean, I struggle even getting Dall-E to iterate on one image without changing everything, so this is pretty cool

This is why the internet exists. A person sharing their odd delights.

The proposal shared here minimizes the managerial overhead of Scrum, but I think it dilutes one thing I really like about it, which is allowing for early course correction on larger projects. If a sprint is a month and standups are minimized, then it could take a month before you course correct

This makes it seem like the whole strike was just about pay. I understand that safety standards was a key thing they were striking for too.

What is the purpose of a video of a word cloud populating?

I think this is a well trodden bit of thought—but I wonder if we've explored the other side of the coin enough. All that bureaucracy must have been created for some reason, and that inertia that it creates might somehow contribute to the big company's longevity?