kidnoodle

Been a fun opportunity to play around with ai in various ways - mixed but definitely powerful experience in using it for data research (finding libraries and their locations), very impressive on reverse engineering apis to get book availability. And then a lot of interesting stuff around getting good behaviour from LLMs as chat partners, for example using jev to improve results from using openrouter by getting it to check for common failure modes and ban providers based on their output.

I’m still working on https://leafle.org which is a tool for getting more interesting reading (or listening) suggestions that one can borrow from the library (or Libby). I’m done for the moment with building it, so I’ve been doing things like making flyers and leaving them in local libraries.

I have made extensive use of this while building my library book recommender doodad https://leafle.org

The LLMs are very good at reverse engineering undocumented apis!

This is very much part of the theory I have with https://leafle.org which uses a similar talk about books, find deeper patterns than the same author or people who read X also read Y, approach to suggest new ones.

I feel like books are an interesting one because the LLMs have ‘read’ so many of them and hence have substantial latent knowledge of them. My sense is that a profile built from books is maybe something that would generalise as well.

My observation from just talking to Claude and co about books is that they kind of wanted to collapse it all to a central explaining idea, which was interesting in a sort of horoscope like way, but just elided too much nuance and complexity to really be useful.

An alternative interface I built for my kids’ Yoto players which adds an intercom feature so I can broadcast messages like “Come have breakfast!” without yelling upstairs.

In the course of other matters, I found myself wondering how many places around the world one can walk to the library from. So I compiled a list of all the libraries in OSM, plus officially listed ones in China and Mexico, and then ran that through mapbox’s isochrones api to see.

Naturally to some extent this follows population density and gdp, but there are quite a few interesting places - Kerala for example has a phenomenal number of libraries!

Is a river alive by Robert McFarlane

A beautifully written travelogue/nature book/eco activist screed that tours the world visiting rivers and the people engaged in living with and protecting them. Fascinating, hopeful, heartbreaking and lyrical. Really enjoying it (evidently)!

At work, standing up openmetadata to capture data lineage from airflow.

At home, increasing the number of libraries I can check live availability for for Leafle. (Library software is a wild place man)

No probs! I’d be super interested to hear how it does if you give it a try.

It is gonna be a bit variable - it has definitely got less knowledge of non-English literature. Couple of French friends tried it out and it did a reasonable job though!

Morning hacker news folks. I made this thing I call leafle - it is a bot you can talk to about books, that can give you suggestions for books you can borrow right now at your local library (at least for 6,900ish libraries it can: https://leafle.nanosheep.net/coverage).

It leans very heavily on the fact that LLMs have very broad latent knowledge of.. well, most books written in English, really. So it tries to do a good job of finding you books which aren't just another one by the same author, or happen to be in vogue right now.

You can try it out a little without signing up, import your reading history from other services (Goodreads, storygraph, particularly nicely with hardcover), or just take a pic of your bookshelf. If you do sign up, it'll get better at suggesting you things as you talk to it more about things you've read (hopefully it does a reasonable job of holding an intelligent conversation on that score), and you can also then connect it to claude and friends via an MCP or play around with the API.

My local library collocates with a cinema these days - I, and many others, were originally super sceptical about this, mostly because the arrangement made a public service and space sort beholden to a commercial entity.

Happily in practice it has turned out to be great, and the combined space operates really effectively as a community ‘third space’. Obviously because of the private sector aspect of this there’s a significant risk in terms of how sustainable it is, but moderated somewhat by the council owning the actual building both live in.

Love a serif (but then I miss writing my docs in latex so..)

I’ve carried on working on my llm powered book recommendation engine, Leafle (try it out if you like! https://leafle.nanosheep.net).

Mostly this week I’ve been wrestling with a better solution to filtering openrouter providers because a common failure mode is getting stuck on one which is either very slow, or doesn’t actually meet the spec it promises in terms of tool calling.

I suppose potentially the upside with books is it takes a while to exhaust the shallow wins compared to music and movies because the investment of time is that much greater.

Oh for sure. I think there’s a lot of potential in llms for this (lot more sophisticated approaches possible than what I have here). Ultimately it should be like one of those fantasy setting librarians or book sellers who looks you up and down and then just hands you the right book for the moment.

Edited to add - or just a really good real life one, it would be awesome if everyone had access to a librarian or bookseller with like an hour to sit and talk with you about books and then help you find some.

A few things - it checks the books it suggests actually exist is maybe the most important (I was really bummed to discover some Claude suggested for me did not). My main reason for building it was to hep me find stuff from the library, so there’s that as well. It also tries to do a little better than more of author X, by using the llm’s ‘understanding’ of the books and what actually mattered to you about them to build up a little profile of your taste and find maybe slightly unusual things that would work for you.

Oh, and it definitely does not pat itself on the back for suggesting books for you.

(You can see the profile it makes and correct it - it did a pretty good job on me)

Oh how annoying! Sorry, that should be fixed now - that was a flow I didn’t account for originally (I figured people would chat first an upload once it gave them suggestions).

Thanks! I’ll have a look

You should be in now! Just trying to manage my limited credit that I’m funding the openrouter account with.

Aha! I was posting about a thing I made for just this purpose the other day - it is called Leafle. You talk to it about books you’ve read and enjoyed or not, and it uses that plus the fact that llms have very extensive knowledge of pretty much all books, to suggest ones you’ll enjoy (or in fact, that will meet your need - reading isn’t just about ‘enjoying’ after all).

It can suggest them to buy, or as audiobooks, or it can find ones currently available to borrow from your local library!

It is at https://leafle.nanosheep.net if you’d like to give it a go :)

I’ve been playing around with exploiting LLMs having ‘read’ every book ever written to make a chatbot to help people choose library books. You talk to it a bit about things you’ve read, and it uses the llm’s latent knowledge of books to suggest ones you might like.

It’s pretty fun! It might even work on your local library (I got carried away and spent a while making it work on a few thousand round the world - you can see the coverage here: https://leafle.nanosheep.net/coverage).

https://leafle.nanosheep.net

I’ve been playing around with using llms to recommend me books I can get from my local library based on the model’s latent knowledge of books it has ‘read’ and a conversation with it about books I’ve liked.

I had a theory that the way to solve this was a location intelligence data union which sold safely anonymised aggregates and shared the profits, while also litigating on behalf of members under available legislation to stop other people using their data.

Alas, I was stymied by not having any cash to work on it, and the unit economics were not very VC friendly (at least I assume that’s one of the reasons why I didn’t get any traction from VCs).

A new spin on my slow baking location intelligence data union (https://wherelabs.info). This week I’m thinking about whether it makes sense to provide a location history ‘vault’, designed to let users expose their location history to LLM’s as context.

A new spin on my slow baking location intelligence data union (https://wherelabs.info).

This week I’m thinking about whether it makes sense to provide a location history ‘vault’, designed to let users expose their location history to LLM’s as context.

Still slowly working away on my location intelligence data union..

I’ve spent a while understanding what sort of market would make it viable. I think it does actually work if you can square: 10K participants per major metro area, revenue of about 2.9M per metro area (so say, 5K monthly recurring with about 50 customers).

At that point you could pay data union participants about $5 a week to share their location data with you.

From talking to some previous data union folks, the major challenges are paying out (my target is much higher than any union managed), and people dropping out over time.

My bet is that these are both solvable things by selling data products rather than just bundles of data, and the data source being very passive.

I’m also interested in the idea that such a union should act more like a union than previous efforts in this space, by actively defending members’ data from brokers.

I’ve been working more on the unit economics of my data union/trust idea (https://wherelabs.info/).

What I’m trying to understand is whether it is viable to pay people ~$5 per week for sharing their location data and demographics based on a 90% share of revenue from sales of data products built on that data. (But without ever selling or exposing individual level data).