This was desperately needed and looks great! Excellent work, mailbox.org team!

If you are looking for an alternative to Gmail or Fastmail, do have a look at mailbox.org. It's a great service, reasonably priced, and hosted in Europe.

@heshsum https://chaos.social/@heshsum/114534310116706697

yeah I agree. I hate this. https://vitonsky.net/blog/2025/05/17/language-detection/

I’m starting to think the second half of ‘have you tried turning it off and on again’ is really bad advice.

@david_chisnall @catch56 @davidgerard oh man, I feel this so hard. I've been coding in Python for a couple years now and the documentation is **terrible** it's definitely written for people who already have comprehensive knowledge. (pandas is still dense but much better. PostgreSQL is **excellent**) LLMs are so convenient for looking up little bits of syntax or asking "show me the common ways people solve x problem."

@catch56 @davidgerard

I originally thought the second would be a benefit. A tool that let people, who otherwise would write no code, write bad code, might improve productivity. After seeing people use it that way, I think a tool that lets people write subtly incorrect code that gives them a wrong but plausible answer is probably more harmful than nothing. And the fact that these are seen as the only two options is a damning indictment of the state of end-user programming.

I have seen LLMs help novices somewhat because a lot of documentation is bad. Asking it questions about Pandas or R seems to help more than the docs for two reasons:

• The docs assume a load of background knowledge. I hadn’t really appreciated this until I asked a novice to look at some of them. Every third word was a piece of jargon that they didn’t know. Most of these have detailed Wikipedia articles and a load other explanations, so there’s a lot of training material for an LLM to give plausible responses that are sufficiently accurate that you can figure out what they’re docs actually mean.
• The documentation is structured as a reference for people who know what they’re doing. If you have problem X and want to know what functions help solve it, you’re stuck. Ask ChatGPT to give you a solution to X and the names of the APIs in the buggy nonsense it produces probably tell you the right bit of the docs to read.

Your future doctor is using ChatGPT to pass medical school, so you better start riding a bike and eating healthy now.

found a weird little guy

this has always been the main problem with an LLM for business use: yes, it can do surprising and delightful things that seem almost human, but the ability to do those things is directly proportional to the likelihood that it will say the N-word or tell customers to kill their parents.

I tried to fuck with the bot a little bit, but they had it completely locked down, which is hilarious because that means it can't actually do any of the improvisational problem-solving that supposedly makes LLMs better than the normal customer service chatbots we've had for years. just a few canned answers, plus some very off-putting personal touches that actually make the experience worse for the customer.

lmao just had my first experience with AI customer support. went great.

this #art got memed and went viral because it fits the moment, but it also is someone's craft, here is a version that credits the artist #GoSlowAndFixThings

Lily Seika Jones
https://www.lilyseikajones.com
@rivuletpaper on insta
purchase a book of Lily's artwork: https://www.gallerynucleus.com/detail/37865/

h/t @jeridansky

Tfw you’re open -0 hours because of a signed int overflow

If I were an 82 year old man, I would simply not run for President again

@Will @ouddin nope! but the Rice Museum was pretty cool, I have a lot more respect for Valencian rice now.

@davidgerard I'm a **little** concerned about the guys running their models at home to make CSAM, but I guess that's a different issue.

i'm not really concerned about people at home running their 3-5 tokens a second, but ~0% of LLM users and advocates and apologists and "but it's useful for" guys are doing that

“…NGOs led by Reporters Without Borders (#RSF) called for the release of Chinese journalist #ZhangZhan... On 14 May 2020, Zhang Zhan was arrested[as a reprisal for] covering the early stages of the COVID-19 outbreak in Wuhan... Her health has…deteriorated due to her intermittent hunger strikes…. Despite having already served a 4-year prison sentence, the journalist is now being prosecuted a 2nd time for reporting on human rights violations…” | RSF https://rsf.org/en/china-coalition-more-60-ngos-including-rsf-demand-release-ailing-journalist-zhang-zhan-5th

LLMs exist to crush labour. That's not even my surmise, the people paying billions of dollars to fund this stuff are extremely open and explicit that this is their goal.

There are no ethical use cases for LLMs at this point in time. Maybe when the bubble has popped thoroughly. Not before.