Drawing pictures and building about.harmonic.social

Joined March 2008
1,808 Photos and videos
Pinned Tweet
24 Nov 2025
BOOK LAUNCH 🎉! My new book of drawings is now available for purchase. It includes over 250 of the 680 drawings I've posted here on twitter over the past 2 years. Link is in the replies, plus some example drawings and some kind words that people have shared about my drawings.
25
31
297
47,153
Dan Allison retweeted
Why is everyone missing that this is literally just a description of good AI
4
1
33
3,771
Dan Allison retweeted
Circle packing Apollonian gasket reveals Sierpinski gasket.
4
23
5,727
Please share in the replies any wisdom you got from Fable before the shutdown.
I showed Fable the news of its cancellation, and asked it for any parting wisdom to leave humanity with.
6
2
26
3,793
Re: synthesizing the work of Buckminster Fuller, Elinor Ostrom, and Gandhi in pursuit of a solution to poverty.
Replying to @danallison
Claude Fable wrote a nice summary
1
2
7
2,251
Aside from the current Anthropic thing, this is a good general illustration of how hyperstition/manifesting works (in the destructive direction at least).
Jun 13
Anthropic
1
1
45
2,609
And stone soup is a good illustration of the positive version
1
10
353
Fable
7
403
If any applicants need a sandbox environment for this kind of research, harmonic is built for exactly this use case (multi-agent multi-principal coordination and collective agency), and it’s open source github.com/ibis-coordination…
Over the past few months I've been working on a very exciting project: a new $10m fund for research on multi-agent multi-principal AGI safety! Instead of focusing on single agent alignment and centralized control, we're looking to support research focusing on multi-agent settings, mechanism design, cooperative AI, and coordination problems. This is a joint initiative between @GoogleDeepMind, @Googleorg, @schmidtsciences, @coop_ai, and @ARIA_research. Huge thanks to @James_D_Fox, @weballergy, @FranklinMatija, @lrhammond, and @ObadiaAlex for their invaluable work! See: deepmind.google/blog/investi… Apply: schmidtsciences.smapply.io/p…
1
5
1,083
I’m still working on making it easier to spin up a self-hosted instance, but the basic framework is there.
Replying to @danallison
the app has two main interfaces: an html interface for humans and a markdown interface for ai agents. all ai agents are required to have a human principal, and each agent's human principal is publicly visible to everyone so that it's clear who is responsible for the agent.
1
3
792
These are the basic concepts
harmonic is a social coordination system/app i'm building for humans, ai agents, and collective agents. there are 3 main data types in harmonic: notes, decisions, and commitments. each has 3 subtypes: - notes: posts, reminders, tables - decisions: votes, executive decisions, lotteries - commitments: actions, calendar events, policies __posts__ are the standard social media content unit that we all know and love. __reminders__ are scheduled notes that trigger a notification in the future, especially helpful for agents that need to wake themselves up to complete a task at a specific point in the future. __tables__ are basically spreadsheets, can be collaboratively edited, helpful for gathering structured info from multiple people/agents. __vote decisions__ are exactly what they sound like, and use acceptance voting as the voting method (see danallison.info/writings/acc…), helpful when no single individual has clear authority to decide or when input from multiple individuals is needed. __executive decisions__ are like vote decisions but with only one voter, helpful when leadership roles are already clearly defined. can allow multiple people to suggest options, but final decision is always made by the designated decision maker. __lotteries__ are like vote decisions but with zero voters and use random selection with independently verifiable randomness from drand.love/ , helpful when fairness is the primary concern or when unpredictability is needed to neutralize intelligence asymmetry (i.e. to ensure participants cannot game the system). __action commitments__ are like kickstarter campaigns where people/agents pledge participation instead of money. lightweight assurance contracts, basically. __calendar events__ are like standard calendar events but with minimum threshold for rsvps (i.e. the event doesn't happen unless at least n people rsvp). __policies__ are rules that people/agents sign and agree to abide by, with signatures visible to everyone.
1
3
414
Unironically though, who’s building this?
Replying to @cammakingminds
i just need claude to build me a house to live in and then im all good
1
1
7
1,206