cristoperb

I don't mind if models do that but I don't understand why a human doesn't take 2 minutes to proof and re-word/delete them

"there's no ethical production under capitalism"?

ethnographies of the rich and the famous

This is a good metaphor for treating the means as an end, thanks. And I agree with your parent comment that that is largely the misalignment that Terence is pointing out.

They reference this paper which describes a method to decrypt reasoning traces (by sending the encrypted trace back to the model and asking it to transcribe it):

https://stolen-thoughts.com/paper.pdf

Apparently just a TI-84 Evo emulator(?) that you can run in a browser for $20/year:

https://education.ti.com/en/products/calculators/graphing-ca...

I still use my TI-89 from high school, but I'm interested to find if there are any open hardware/firmware calculator projects with basic engineering tools and a CAS.

I can't speak for the person you're replying too, but I use -- for emdash for two reasons: I never remember how to type an actual emdash in linux/X11, and more importantly, I do most of my writing in Asciidoc which converts -- to an emdash automatically. It's nothing to do with bot detection or whatever.

But it does get me confused sometimes because in LaTeX (and other markup languages) -- gets converted to an endash whereas it takes three hyphens --- to make an emdash.

I haven't tried it for anything myself yet. The paper provides several benchmarks. The emphasis during training was on multi-language support (over 1800 languages are represented in its pre-training data, which is 40% non-English) and non-copyrighted training data... and the benchmarks seem to suffer for it.

https://arxiv.org/abs/2509.14233

Apertus is the open source 8b and 70b LLM from swiss-ai. They've published both the base and the instruct sft models. Very cool that projects like this exist.

https://apertvs.ai/pages/documentation/

It looks like the ia's collection of Compute! is complete:

https://archive.org/details/compute-magazine?page=2&sort=dat...

Also Compute!' Gazette:

https://archive.org/details/computes.gazette/Compute_Gazette...

Though see the Plain Text Accounting[0] movement for something maybe more unixy than Gnucash. I download .csv files from my bank and credit card issuers and import to hledger[1]. hledger has its own rules engine for filtering/transforming imported entries, but you could also preprocess the files using any unix tool before importing if you needed to.

0: https://plaintextaccounting.org/

1: https://hledger.org/

I missed the "a" in the title and expected something about Proudhon

> We all know Meta can still read E2EE chats

That can't be true, otherwise in what sense is it E2EE?

Thanks, your install script worked for me.

In case it helps anyone else, the first time I tried to run purr I got "OSError: PortAudio library not found". Installing libportaudio (apt install libportaudio2) got it running.

Yes. Here's an AP article:

https://apnews.com/article/afroman-police-raid-lawsuit-ohio-...

You could do signatures/MAC without encryption to guarantee that the message was not modified

Thanks for sharing, and congratulations!

Interesting that you didn't use dead reckoning at all to return to the nest. I didn't think I could get away with that. And somehow using a second channel to extend the trails never occurred to me.

I just came back to check if anyone else was discussing their entry, so thanks for posting your write up!

I wrote mine by hand just testing it using the in-browser simulator. I initially used subroutines to try to keep things structured but then resorted to inlining most things (and flattening as much state as I could to bare registers) to try to speed things up, though i never quite got my ants to always move within the 64 op step.

My best score on the official test set was 654 (43rd place). I put my code and a little write up here: https://github.com/cristoper/brain.ant

> Thunderbird delivers RSS feed items the same way as email, so you can apply filters to mark them as "read"

This is a good idea. I use Thunderbird only for a small number of feeds I want to read every post from. I used to also use a separate feed reader for my "river of news", but eventually I stopped looking at that and just loadded hackernews or reddit when I wanted a distraction. But I might try Thunderbird for more feeds and just auto-mark most of them as read so I can browse at my leisure.

I guess it's not prefatory remarks or disclaimers that I find so grating, but the explicit "I'll explain" (or worse, faux conversational "May I explain?" "Let me explain") followed immediately by the explanation.

This is a nitpick, but since this essay is over 15 years old now I don't think the author will mind. This phrase always rankles me:

> Let me explain what I mean.

It turns out that if you're writing an essay or a youtube script you don't have to tell me that you're going to explain something to me before you explain it to me. I guess it acts as a "hack" to try to impart some gravity to what follows without actually having to write a convincing introduction, but unlike "it turns out" it can almost always just be deleted to improve the flow.

This is my project/writeup. I think it is too simple for a Show HN, but if anyone actually tries it out and finds it useful I'd love to know!

source code: https://github.com/cristoper/lichess-puzzle-timer/

I don't know when the term became widespread for gui-style terminal programs, but the wikipedia entry has existed for more than 20 years so I think it is an older term than you imply.

https://en.wikipedia.org/w/index.php?title=Text-based_user_i...

I cloned my voice and had it generate audio for a paragraph from something I wrote. It definitely kind of sounds like me, but I like it much better than listening to my real voice. Some kind of uncanny peak.

It's just Gemini. I'm guessing they changes the system prompt for the new year or something, but it's pretty annoying.

I don't know the answer to your second question, but what about "transformer" makes you think "not an LLM"?

Apparently it is the same as the DeepseekV3 architecture and already supported by llama.cpp once the new name is added. Here's the PR: https://github.com/ggml-org/llama.cpp/pull/18936

No, it has always used its own protocol.

But apparently there is an XMPP extension that implements the Double Ratchet algorithm that Signal is based on:

https://en.wikipedia.org/wiki/OMEMO

No worries, I don't think "company" is even technically wrong. But I do think given the nonprofit structure (and Moxie Marlinspike's track record), that there are fewer incentives for Signal to lie about its privacy guarantees than a messaging app backed by a commercially-driven big company.