Jay Springett

Episode 302: Monsters In The Mirror Permanently Moved is back! It's an hour long techno-gothic essay on consequences of the silicon mirror: How inside it, we are now confronted by entities containing the liquefied dead of all human culture, wearing language as their skin.

If this kind of exploration intrigues you, you can listen here: experience.computer/p/marco-gianco…

Jay knows how to nudge you to dig deeper and find words to express the inexpressible. The moment you start this kind of mental spelunking you realize how deep, varied, and mysterious it all is.

Turns out I was the first aphantasic guest on the show. Even between aphantasics, the way the mental world is experienced can vary a lot. "A lack of imagery" is too reductive a term for this strange trait.

Jay's podcast is specifically about what it's like to *imagine* things. He has aphantasia like me, and SDAM like me, so we had a great time comparing notes.

I was recently interviewed by @thejaymo on his podcast Experience.Computer, all about what it's like to imagine things when you have aphantasia. It was a lot of fun.

@matdryhurst 😭

@thejaymo Thanks just trying to communicate what words mean

They clearly didn’t read the same report I did The report said the government now has no preference, having prior had the preference of a copyright exception with opt-out That could just as well mean we see a copyright exception with no means to opt-out, just as the treasury today announces a clear statement inviting every major AI lab to open offices in the UK

@matdryhurst Geep doing the good work Mat ♥️

Which in practice is a hand wave Because in truth everyone is terrified to scare away AI business and there is no evidence licensing will produce meaningful income for anyone but the biggest media organizations Opt-outs framed as the worst case scenario is nonsensical

@js_horne Question is how do you have the sub skill agent operate inside the parent's governance envelope? Ensure token budget and permissions are still scoped etc?

1) Saves context 2) Can be integrated into a workflow 3) Preserves the data and context for the creator 4) Comes with a better business model for the creator

Skills are better off packaged as specialized agents

@deontologistics read it back to back with this verysane.ai/p/polly-wants-…

@deontologistics I learnt a great deal reading this! Tnx <3

I think a big divide at the root of arguments about LLM output is whether to treat machine language as socially meaningful or not.

The city centre car park has had its day. There is now a significant opportunity to put the land previously used for car parking to much more productive use. Land that could be used for tens of thousands of homes in city centres. bbc.co.uk/news/articles/…

Coding agents mean that we are about to enter the golden era of TempleOS style projects. This is a good thing.

Today is the last day to pre-order

I'm still fine tuning the "Forge" prompt to fully populate metadata to my satisfaction, and have the by-sa licence etc survive the remixing of two or more Knowledge Objects being combined.

Which you can read in prose here x.com/thejaymo/statu…

Forthcoming essay on Ontological Hardness and AI agent harnesses design principles, in machine-readable DSL.

Not right now tho. I need to go to bed lol.

Gonna start making lots of my long form blog posts available as Knowledge Objects in YAML DSL in a repo, so you can throw them into your project or context and think with them.

THIS IS THE WAY