quamserena

This is trivially preventable via affine types

You can also read the text yourself and draw your own conclusions...

These discussions always focus around enforcement and never on alignment. The moat for this stuff historically has never been strict enforcement; it has been that the people who have the know-how on how to do it have nothing to gain by doing it, since they are well-educated and benefit from the current socioeconomic order (they have no motive to change it; rather, they want to climb it).

This is shifting. First, economic stratification is getting worse, and as economic mobility declines people start looking for alternatives. (See all of Gen Z cheering for Luigi Mangione). Second, AI will enable people who are less educated to build these kinds of weapons.

For example, you can use a Kalman filter to greatly improve the data you get from an IMU and GPS via sensor fusion. Before, this required a specialist skillset; now you can get a "good enough" implementation by prompting Claude.

I really wish the debate around this stuff wasn't framed in terms of preventative enforcement because it naturally leads towards more enforcement (when your only tool is a hammer...). The root of the issue is that the government does not trust its citizenry to follow the law without Big Brother watching. That in and of itself is a symptom of a larger grave political crisis in America: the decay of the state's political legitimacy.

This is why you use uBlock Origin.

I wish it would just say "k, updated xyz to 1.2.3 in Cargo.toml" instead of the entire pages it likes to output. I don't want to read all of that!

By scanning downloaded binaries for known viruses?

You can simply ask the model to point out if there are any problems and then fix them yourself. You don't have to copy and paste its output into your book. You can also pay for an actual copyeditor to edit your book.

This has happened multiple times [0]. You shouldn't put your money into Polymarket

[0] https://old.reddit.com/r/CryptoCurrency/comments/1jki1lj/pol...

I can second this, after finishing my intro Japanese classes I was able to parse the grammar of most sentences. Memorizing vocab was the hard part, so I used OCR on manga pages and then Yomitan to hover over and see word definitions (in English).

Most existing mainstream languages aren’t expressive enough to encode these invariants. For languages outside of the mainstream, Lean 4 is a language supporting verification, and it’s also a full programming language, so you can write your proofs/theorems in the same language that you program in.

Not really. Here’s a comparison of different languages: https://notes.eatonphil.com/parser-generators-vs-handwritten...

Most roll their own for three reasons: performance, context, and error handling. Bison/Menhir et al. are easy to write a grammar and get started with, but in exchange you get less flexibility overall. It becomes difficult to handle context-sensitive parts, do error recovery, and give the user meaningful errors that describe exactly what’s wrong. Usually if there’s a small syntax error we want to try to tell the user how to fix it instead of just producing “Syntax error”, and that requires being able to fix the input and keep parsing.

Menhir has a new mode where the parser is driven externally; this allows your code to drive the entire thing, which requires a lot more machinery than fire-and-forget but also affords you more flexibility.

He was bought out by private equity.[0] That’s why the channel is different now.

[0] https://www.electrify.video/post/electrify-completes-majorit...

3blue1brown has great math visualizations. I find the top 10% of YouTube videos are worth the time over reading, and the bottom 90% are comparable or slower. Those are also nice though because you can put them on while doing other stuff, like eating or doing the laundry.

Part of the public pushback is that people almost always drive the “feels like” speed and not the posted speedlimit. We build 6 lane roads and then wonder why people go 50mph when it’s 35 posted, it’s because it’s 6 lanes and 35 feels slow. Cities profit from this in the form of speed cameras, which is why they’ve been outlawed in a lot of places.

Omg I thought this was just me. How do I turn this off? On iOS, this has been bugging me for a long time.

this sounds nice, but neglects the fact that (1) materials cost has gone up and (2) zoning requirements exist. (1) means its just more expensive to build overall, and (2) means that a lot of proposals for apartment complexes get voted down.

Yes, I agree. Maybe if you’re a fast reader icons don’t do much, but for people who are illiterate (20% of America) they figure out how to use tech by memorizing the icons and locations of buttons.

Thank god, toasts are so annoying. Every little action in Google Calendar has an associated toast/snackbar to go with it that tells you exactly what you just did and asks if you want to undo it. Like wtf? I can’t use my calendar app without these stupid toasts flying in and out and trying to draw my attention to read some irrelevant text. They go away too quickly for anyone not technically literate to click on them, and they are too slow to keep up when you’re creating a ton of events (they just fly in and out). I hope these go away, they add nothing to the application.

Zig’s announcement[0] might provide some insight

[0] https://ziglang.org/news/migrating-from-github-to-codeberg/

I am in university currently and I run Linux, and the only thing that I have needed Windows for has been SolidWorks. Everything else has worked just fine. We’re actually provided Linux VMs because so much software development happens in Linux (or MacOS); you need to know *nix to be job-ready in CS. I’m not sure what world you live in.

This works for ASCII, and you could just “smush” these special Unicode chars into ASCII lookalikes but then your AI won’t be usable by people who actually use these chars as part of their language.

I think the bigger problem is that if the dataset was sufficiently poisoned, LLMs could start producing Greek question marks in their output. Like if you could tie it to some rare trigger words you could then use those words to cause generated code not to compile (despite passing visual inspection).

Including RTL-LTR flips, character substitutions etc? I think Unicode is vast enough where it’s possible to evade any filter and still look textlike enough to the end user, and how could you possibly know if it’s really a Greek question mark or if they’re just trying to mess with your AI?

Why not just rip the audio file off? Would make for a funny DC request

Edit: On second thought this could be an OPSEC problem. Sorry but I don’t know if anyone can help you :(

I have seen engineers slap Teensies on a PCB and call it a day, so it’s definitely normal. It’s faster than having to route your MCU, USB, debugger, etc. manually, so there isn’t really a drawback as long as it physically fits there.

It clearly hasn’t been passed through a PR team or ChatGPT; if it had been you’d expect them to fix the grammatical errors. It’s an honest stream-of-consciousness blog post almost certainly written by Charli XCX herself and herself alone about her thoughts, and it is honest and unapologetic. What word more describes this than “raw”?

I’ve never had AI understand a problem of that depth though. It can maybe surface the right part of the docs to reference but in the times I have used it it leads me down the wrong path half of the time.

I don’t understand how this can be more optimal than just reading the documentation that a human already wrote. All “normal” uses can be answered by reading the docs, everything advanced you can just read the code. I’m not sure when I would ever use this?

For Rust, you can generate offline documentation (as HTML and JS files) for your crate and all dependencies. I’ve used this for working in airgapped environments.

Absolutely. Party has no leadership. The progressive wing has been organizing their base, protesting, and calling congressional offices, trying to encourage centrist Dems do something. They finally do this but then get fold immediately when the market starts to dip and flights get canceled. The progressive wing will now proceed to spam call their offices, protest book tours, and call for Chuck Schumer to step down. Rinse and repeat at the next deadline on Jan 30.

If Mamdani’s victory is anything to go by it’s that the centrist Dems are toast. They will get primaried. They know this, it’s why the fall guys that voted to end the shutdown are not up for reelection during the midterms.