EA AI safetyism increasingly looks like Marxism-Leninism for the algorithmic age.
The old vanguard claimed privileged knowledge of the inevitable course of History. The new one claims privileged knowledge of the probabilistic course of Humanity.
Both use an elaborate intellectual framework to reach the same political conclusion: a small group of enlightened people must constrain everyone else for their own good.
That has never lead to anything except monumental human suffering.
my mom is a massive codex/chatgpt user but isn’t on X and has no idea who @thsottiaux is, so every time her usage gets reset she’s completely surprised and so happy she can keep going. it’s genuinely adorable
rough two cents but ai people are too cruel to the artists and mathematicians impacted by ai entering their field
a good comparison i saw was 'imagine if claude code was just outputting a binary asset, not code'
antagonizing fellow humans over ai doing their work is fucked up
@theo If you have 1,000 LLMs rereading the same company data to figure out the same things, you're wasting a ridiculous amount of compute. Memory means figuring it out once, saving it, and keeping the sources so it can be checked (and it could just be markdown that does this)
@theo@Winterrose > What possible purpose would products like this serve that isn't served better with really basic updates to the Markdown files, giving context to agents?
This is memory in its simplest form
just submitted my project to the @OpenAIDevs WebMCP hackathon: Sushi
> Sushi is a music production app where the entire song is written in code
> you can edit that code directly, or arrange the same song visually like a traditional DAW
with WebMCP
> ChatGPT or Claude can read the actual composition, change it, test the result, and play it
try it here: https://t.co/yjeerxiPvf
@Teknium@NousResearch has the cleanup aggregated at all around specific areas or modules in the codebase or has it been fairly broad? Curious about the state of the “god files” lol
@Teknium Hey is there any word on this or am I the only one? The in app browser is so broken on hermes desktop that my agent refuses to use it, usually resorting to headless browser usage instead. (using gpt 5.6 sol, on arch linux hyprland)
@ayyinspector@NousResearch hey! native support for hermes bots is being added to signet, a shared memory layer, if you feel like trying it out :)
https://t.co/78gcsoorFt
@NousResearch hey! signet is the shared memory layer my agents run on. we're adding native support for hermes bots right now, so the bots can share memories with one another and contribute to a shared knowledge base. they shouldn't have to start from 0.
https://t.co/78gcsoorFt
my agents' memory is self-hosted in signet. it aggregates the places i actually work in--obsidian, github, discord, transcripts--into one shared memory layer my agents (pi and hermes) can reach, so my agent knows everything i do. you didn't list it, but it's worth a look if you're not happy with the options hermes gives you by default.
https://t.co/nKfLfdHF91
@raphaelschaad Try this out! it supports having multiple agents connected at a time and once installed will backfill with your agents existing memories: https://t.co/78gcsoorFt super easy to set-up
@mattshumer_@trq212 are you fucking serious?
are you actually fucking serious??
procedural knowledge is dead, everyone
matt shumer has said it so it must be true
this is cope beyond what i know how to describe