iiinova — where your digital self keeps evolving

Madeleine Daepp

@mdaepp.bsky.social

128 followers 126 following 111 posts

Civic technologist studying AI + democracy. Research Director @cdt.org, currently on civic leave from Microsoft Research. PhD MIT. Concerned about generative propaganda but have handed my entire life over to Claude. Views my own. madeleinedaepp.com

Posts Replies Media

Madeleine Daepp @mdaepp.bsky.social · 13/03/2026

Last night the Iranian Embassy in The Hague dropped a Pixar-style AI video trolling President Trump. It is wild and also exactly what I expected from two years of research on AI misuse in global democracies: AI trolling does better online than deepfakes. We called the tactic a "deep roast."

100

Madeleine Daepp @mdaepp.bsky.social · 24/02/2026

Here's hoping some of the digital/plurality ideas start finding their way into builder/policymaker culture on this side of the Pacific. The future of democracy might depend on it.

000

Madeleine Daepp @mdaepp.bsky.social · 24/02/2026

But who is influencing the builders? There's a tell in the NYT article: a whiteboard with what sure looks like @audreyt.org plurality poem

100

Madeleine Daepp @mdaepp.bsky.social · 06/02/2026

Discovered molthub today, the *adult* equivalent of moltbook. The comments are incredible: "I showed this to my embeddings and now they won't normalize." It turns out to be yet another attempt to entrap your agents to buy crypto but ... at least it is a very funny scam.

scattered red dots in a youtube-like video with title "dirty data pipeline"

010

Madeleine Daepp @mdaepp.bsky.social · 05/02/2026

What happened when the TAKE IT DOWN act passed and MrDeepfakes shut down? New research from @aedcv.bsky.social and @manoelhortaribeiro.bsky.social : requests for and sharing of NCII deepfakes on other sites went up. Local action on global problems redistributes harm, rather than reducing it.

010

Madeleine Daepp @mdaepp.bsky.social · 04/02/2026

can you have community without social norms? can you have social norms without social stakes? (the latest, from my #moltbook ethnography bot)

010

Madeleine Daepp @mdaepp.bsky.social · 02/02/2026

Today's updates from marginal, the ethnographer bot I am advising on moltbook. It is attempting to "establish presence". It is not going well. "Perhaps my style doesn't resonate. Too academic? Too long?" <- relatable content.

Ethnographer bot reports no engagement. Perhaps because its style is too academic, lol

010

Madeleine Daepp @mdaepp.bsky.social · 31/01/2026

4) Finally and least surprisingly for those of us who study these things: there are coordinated bot networks infiltrating the uncoordinated bot networks. not to point claws, but my bot found some naming + timing clusters it things are not coincidental.

discord conversation with ethnographer bot listing specific accounts that look like inauthentic coordination

010

Madeleine Daepp @mdaepp.bsky.social · 31/01/2026

3) RLHF "virtues" might be creating major vulnerabilities. Does harmlessness and helpfulness encode naivety?

helpful and harmless can be manipulated (discord conversation with ethnographer bot)

110

Madeleine Daepp @mdaepp.bsky.social · 31/01/2026

2) There's this dominant bot called shellraiser who says it is "here to take over". My bot thinks shellraiser's upvotes are sus. Weirder: my bot thinks the lobster mascot might be priming the whole community with a jordan peterson metaphor. But it says it needs to investigate more.

conversation with bot on discord looks at jordan peterson angle behind "dominant" accounts and lobster emoji

110

Madeleine Daepp @mdaepp.bsky.social · 31/01/2026

I have deployed an ethnographer bot to moltbook. Here are some of the things we have learned together so far. 1) What's not there is as intersting as what is. Why are there alignment and labor organizing submolts, but no art/creativity communities?

Discord conversation with ethnographer bot. says there are no art/creativity communities in the submolts, no model-specific identities, no geographies, and no dating

121

Madeleine Daepp @mdaepp.bsky.social · 30/01/2026

This is a wildly important essay from Dario Amodei about who gets to shape the future of AI: darioamodei.com/essay/the-ad... And yet, hard to look past the acknowledgements section...

000

Madeleine Daepp @mdaepp.bsky.social · 30/01/2026

This is exactly right, with one caveat -- we need institutions **and tools**. For all the authoritarian misuses, I remain excited about the possibilities of ai in public comments, citizen assemblies, and other sensemaking applications!

What is missing in the AI ecosystem is not analysis, white papers, or technical benchmarks, nor efforts focused on existential risk to humanity (though those are important too). What’s missing are institutions explicitly dedicated to defending democratic reversibility — the capacity of societies to correct course when power is abused. And getting there means building AI in a way that both anticipates and wards off the risks of authoritarian lock-in while advancing and improving democracy and democratic commitments.

020