@ele Great job, thank you! It’s important to have accessible videos like this to explain the basics

@Em0nM4stodon In my preferred timeline, all those CEOs are talking about their flagship projects to achieve all the SDGs within the decade. #SDG

https://en.wikipedia.org/wiki/Sustainable_Development_Goals

@glennf Nice find, dictation is what I want there

@kattekrab That's up to the teacher. Sometimes the best way is face-to-face.

@cogdog

Agreed, that article is horse turds

https://interactives.cnas.org/reports/sovereign-ai-index/

It wasn’t until I beat the PS5 game #Saros last night (which is really slick btw) that I realised that the whole thing is pretty much the same story AND gameplay as Super Mario.

https://housemarque.com/games/saros

@lightweight @downbeatdan

I made an app that fully maintains my own static music site here: https://mantisaudiogram.music

I was planning to release it openly once I get a bit further with federation ... until we find a way to store music more reliably in something like BitTorrent forever, my dream would be every artist paying only for a minimal basic web server and then registering it to be part of a global federated system for discovery, playlists etc.

Another superb and hilarious read from Cory Doctorow @pluralistic

Thanks, Comrade Trump, for helping to push the world to renewables and sovereign tech.

https://pluralistic.net/2026/04/20/praxis/

@cogdog Also pretty good, free and easy for an agent to use: https://github.com/remotion-dev/remotion

https://dougiamas.com/giving-your-agent-a-mission/

@carnage4life It’s not just him … he has support of US government and was voted in

@dan A good harness like Claude Code or OpenCode is a big part of it (they work with any LLM).

The other thing though is a decent-sized context window. Smaller projects work better on all local models. For larger projects you really need at least 200k token context. Gemma 4 is looking very promising!

@dan Currently the reality is I need to still use some frontier models to get speed and quality on core stuff. But always moving more and more functionality towards local models! The goal is 100% local eventually of course.

@hw @marcusgreen This is the way. Then everyone is responsible for their own power use in a very direct way. (My home servers already run on solar 24/7)

@jbond The Chinese operate like this too. Just makes good sense IMO

https://www.youtube.com/shorts/5vPirAga0SI?si=oJuhmJGtL8tTgFWi