Notices by Simon Willison (simon@fedi.simonwillison.net), page 6

The whole project took 20 minutes from start to finish - and honestly, I don't value this obscure little feature (on a page of my blog that gets virtually no traffic) enough for it to be worth me spending any more time on than that https://github.com/simonw/simonwillisonblog/issues/391

Glad to see I'm not the only person howling with despair at how difficult it is to build anything that involves OpenID Connect in the comments on this Hacker News thread https://news.ycombinator.com/item?id=38868610

@mfowler @timbray That's great advice for people who know they're going to continue actively caring about their blog for the long-term, but I'm hoping to find a safe default for people who might get inspired, blog for a few months / few years, then lose interest - but still want the content they write to stay online effectively forever

My hunch is that expired domains are an extremely common reason for personal content to blink out of existence

@Dtl Blogger has been around forever, but Google's track record of keeping things running (even things that have been around forever) is still bad enough that I don't trust it - especially since they seem to have lost interest in further developing Blogger over a decade ago

@mikemccaffrey @natbat "digital columbariums" is a fantastic turn of phrase, love that!

Since it's my birthday today, I asked @natbat if she could set me up a non-profit foundation which works as a co-operative where people can pool resources and donate their domain name and cash to keep it maintained (and static files served from it) forever, with a board of directors that appoints their own successors and keeps the thing running in perpetuity without it depending on any single individual

She said she already got me a new watch

I don't trust Medium because they added a paywall / registration wall to content that had previously been published

Substack are too young to have proven themselves as long-term trustworthy stewards of content

I used to trust Twitter for long-term content availability, but then they got bought by someone who clearly doesn't care about that at all (and threads are now hidden behind registration walls)

I have a hunch that domain names eventually expiring is one of the top causes for blog content to become unavailable in the long (10+ year) term, but I don't have any actual data to back that up, just an intuition - partly based on how many of the things I've linked to from my own blog in the 2000s no longer exist

@evan @philippe @timbray Oh wow that's amazing!

Do you know of any examples of sites that are hosted on there? The http://stats.s3.us.archive.org/downloadsPerDay.png in their docs is giving me a 500 error at the moment

@philippe @evan @timbray Archive.org is really useful, but the one thing it doesn't (yet) help with is keeping URLs working - and my personal ideal of blogging is all about linking to things and being linked to

If someone wants to start a blog, and they'd like the content they publish to stick around on the internet for as long as possible into the future, what are their best options?

Buying a domain is tricky because domain names expire

GitHub Pages is pretty great, because GitHub is free and has, to-date, a great track record of not breaking *.github.io content

What are other great options? WordPress.com I think are good on this front

I don't trust anyone with less than 10 years of track record

To celebrate public domain day, here's a fine-tune of Stable Diffusion XL on frames from 1928 Steamboat Willie

https://huggingface.co/Pclanglais/Mickey-1928

(Stable Diffusion's core training data itself is of course crammed with unlicensed images)

Here's a Mickey fighting a walrus, and Mickey fighting a Pelican

To round off the year, I pulled together some notes on "Stuff we figured out about AI in 2023"

We figured out a lot! https://simonwillison.net/2023/Dec/31/ai-in-2023/

Here's the table of contents for my 2023 in AI round-up

"The most surprising thing we’ve learned about LLMs this year is that they’re actually quite easy to build." https://simonwillison.net/2023/Dec/31/ai-in-2023/#easy-to-build

"I like to compare the difficulty of training an LLM to that of building a suspension bridge—not trivial, but hundreds of countries around the world have figured out how to do it."

I've been trying out https://perplexity.ai recently for search - it's the best implementation of AI-assisted search I've tried so far

It runs multiple searches in parallel (using the Bing index) and uses the results to answer your questions, sometimes automatically running follow-up searches too

Similar to Bard and Bing, but it feels like they've built a slightly more effective implementation of the pattern

It's not quite yet my ideal research assistant - I want something I can task with finding information that goes out and automatically does what I'd otherwise spend an hour doing myself - but it's hinting in that direction

If you want to try out Perplexity Pro they're offering a 2 months free holiday deal right now, http://pplx.ai/holidays

(This is very much not a paid endorsement, but it's relevant to this thread so I'm including it here)

If there is a better Christmas movie than The Muppet Christmas Carol, I don't know what it is

Michael Caine: "I'm going to play this movie like I'm working with the Royal Shakespeare Company. I will never wink, I will never do anything Muppety. I am going to play Scrooge as if it is an utterly dramatic role and there are no puppets around me."

Legend.

I suspect a neat benefit of everyone wanting to build "chat with your PDF" things is that there's a sudden flood of extra effort being put into PDF text extraction and parsing libraries right now, across a bunch of different languages