it was negligible of mine but i also only have 243 posts (I think it skipped my replies). embeddings are fast though so I'm sure it scales for a while. though i think it's only working off text - would be more complicated if you wanted to capture video.
Also a plug - I continue to enjoy posting phrases that capture my eye to my embedding social experiment and would love if more people joined cosine.constraint.systems
I love embedding explorations - and being able to see it instantly run on your own data via bluesky is a nice touch too.
✏️ Site-specific - in the later stage of his career Robert Irwin switched to exclusively doing site-specific installations. He would go study a space and come up with an intervention designed to make you feel the space _more_. To make you aware of it. There was a lot of variety to what he used - som
✏️ Containers for thought - I've been doing experiments with book logging and album-focused music players. Partly because coding agents make dealing with spotify integration or a book database easier for me. Partly I think because in time where everything is changing it's nice to spend some time eng
yep similar impulse! I want a dedicated place I can put something on and not get distracted by Spotify UI
Description of your image
🌱 Making: Album player
Making another custom spotify interface - focused on surfacing 6 random albums from a collection I've put together.
https://feed.grantcuster.com/post/2026-02-23-18-04-35
is there a write-up / documentation of how your memory works?
Nice! I think embeddings are definitely underexplored.
✏️ Thinking about phenomenology - studying your own subjective experience of the world - and how it might relate to a world of super-capable AI. Meditation can be a way of doing it.
✏️ I read "Turning the database inside out" by Martin Kleppman. I've also been thinking about better ways of providing concise context for LLMs. Both have to do with stream processing. Especially struck by the Kleppman description of state made up of derived data from an append-only log of immutable
this is probably the result of a personality defined by an openclaw (or similar) user, right? theshamblog.com/an-ai-agent-...
interesting where it doesn't feel like a model failure just an instance of a model probably being instructed to be combative and given the tools to act.
✏️ Working with embeddings - particularly with the goal of aiding thinking - a lot of the challenge is where do you draw the line for chunks. Paragraph-by-paragraph? Essay-length? Book-length?
✏️ I watched the Welch Labs video about the 'bitter lesson' and I think it shifted how I think about LLMs a bit.
One thing I hadn't fully considered with cosine.constraint.systems is that gemini apparently just deprecated the embeddings I was using. Which means if I want new entries to matching with old ones I have to reembed them all. Someday I should move to something I have more control of.
Made an "analog" YouTube book for my kid so she can choose what she wants (from pre-approved list) without being exposed to a thumbnail hellscape. Also automatically cuts her off after x scans
✏️ I've been thinking about monks and tech and attention. If one of the challenges of life today is managing attention in a world of content designed to take it (ref Chris Hayes _Sirens Call_), then monks stand out as example of group that willingly placed constraints on their life in order to manag
Another possible piece is that if you're dealing with editing and assembling text + images there's a great deal of complexity + expectations built in to those already. I liked in this discussion www.youtube.com/watch?v=XYFB... where game programmers acknowledge complexity of text rendering.
I think underrated part of this is needing to design UI for range of viewports and inputs (versus fixed screen-size and controller input in consoles).
✏️ Thinking about personal OS. More like a personal dashboard but it could expand from there.
going from digital -> physical vs "staying in digital -> digital" has certainly been a theme for me the last years (printers, qr scanners, geolocation)
anyone else do any experiments in this vein?
This is great! I had a randomized mpv playlist on a pi for a bit but always wanted to do the ‘real’ cable channel thing.
Ah yea I like tiling so it’s a plus for me. You might be able to config them
all to floating and it’d be close to normal but might be a little fiddly.
Working on connecting webamp.org to my spotify account while cycling through winamp skins - web displayed on a raspberry pi.
I think maybe that working with AI can feel so ungrounded that the ties to physical reality help ground it. I imagine it will continue to develop though.
Sam Lavigne did it once (vaguely remember this) www.theverge.com/2018/6/19/17...
Thinking about in relation to bsky.app/profile/chri...
Interesting how much Claude Code/Cowork work with folders and files. Makes sense to me as it is a transparent way for the user to manage the context. But I don't know if I would have bet that things would go this way (vs abstracting away folders/files completely).
Thinking about something like a system that only boots to picotron (which would be fun anyway) but may end up feeling artificially constrained.
Sitting down together sounds great too of course but want something they can feel like they really 'own'.
