utopcell

until November, probably.

We also live in a very different world now compared to 2021 when the problem was reported. A brief Gemini query to translate the old diff format to the new one returned in a few seconds with what appears to me a correct, < 40-line, solution.

While I agree with the article's point, we should note that collisions do not happen out of the box because Neovim defaults undodir to $XDG_STATE_HOME/nvim/undo/ (~/.local/state/nvim/undo/), while Vim defaults to storing .un~ files alongside the edited file or inside ~/.vim/undo/. Data loss only occurs if both editors are configured to share an undo directory or save inline .un~ files.

Great site, intuitive description. I also found [1] very useful in the past.

[1] https://bbycroft.net/llm

They obviously did not send the proof to Claude.

I'm pretty sure they were sending a prompt for Claude to _find_ the Navier-Stokes proof, using ideas that have been publicly shared before online, but not necessarily used for the problem.

"BaW Music Theory", for that click-bait effect.

What's surprising (to me) is that there exist 3840x2560 28" IPS panel monitors on Amazon for only $150.

External 16" 3840x2400 monitors are ~$150 on Amazon.

> You can ask 100 people and they'll all give you a different list.

True. With 5 choices you need at least 126 people before you can guarantee that two lists are the same.

Definitely in the pool of people that used to do this extensively. I would occasionally increase/decrease the colum limit by a few characters and try again if the words didn't align.

There's no credible reference to this. The most likely scenario is that they worked together in designing the system. They also credited their professors, Motwani and Winograd, for it.

I don't want to alarm people but I hear they now have automobiles that can outrun humans.

How much does the monitor cost?

Too late, already trapped.

Oh no!, another paperclips game. There goes my day.

It is well documented that the Antikythera mechanism could predict solar and lunar eclipses. The device's operation was based on 7th century BC Babylonian astronomy, specifically the Saros cycle, and Greek astronomy. The theory and data was definitely present in Thales' time.

It doesn't. I went to Saragoza in Spain, at 8:30pm on August 12 to verify. Cool project nonetheless.

From context, I assume it's the grace period you are given to find another team after you get laid off.

TheSpectrum is great at capturing the original look&feel. The keyboard feels identical to the original. An emulator captures the software aspect well, but that's not all the speccy was.

What makes you think it doesn't?

I remember a time when a tweet was 140 characters.

Who knows what will happen in 20 years. Right now, it's not the case.

For sure, nothing like the original, but it's a cheap, quick way to get you started.

Demis is great -- this is not a zero-sum game. Who built what, and the relative importance, is an interesting discussion to have, but it's orthogonal to my point.

Which is: When you think about Google, true, old, "don't be evil", tech excellence Google, you don't think about Demis. You think about Jeff's and Sanjay's geeky, technically uncompromising, faces.

Demis stepped down? Who cares? Jeff and Sanjay are leaving Google. We are in uncharted territory.

The Retro Games TheSpectrum is not a bad starting point [1], it's very close to the original and you can use HDMI video output and USB power.

[1] https://retrogames.biz/products/thespectrum/

You probably mean absl::flat_hash_map<>.

> Many security tools for doing security audits with LLMs are based around a "look at this file" loop

This makes sense. At most it converts the question of ability to a question of cost (i.e. fire up a prompt for each file).

This article reduces the hype about Mythos in my mind: A new model that can find 9 new bugs while no previous model can identify them, is a whole different story from what this article demonstrates: that only 2/9 of the detected bugs are new for Mythos.

Great work.

> You can see which bugs each model found in the full report.

This should have probably been surfaced at the text. To answer my own question: If one uses gemma4-26b-a4b, mimo-v2.5-pro and gemini-3.5-flash then 7/9 bugs are covered, while no model can uncover the remaining two.

> And, you have misunderstood what the benchmark does. It tells the model to audit the file, and it is allowed to look at the rest of the repo. It is not pointed at the bug.

I obviously meant that the model was pointed to the bug report. What would have been the point of providing the actual bug. This is still easier than telling Mythos to generally look at a codebase and find any bug.