Ambrose

a. piece. of. software. is. not. the. type. of. entity. that. can. gain. consciousness.

MagicSchool uses a multi-model approach and pairs each tool with the model that performs best for its task. Today, we use multiple providers, including OpenAI GPT, Anthropic Claude, and Google Gemini, among others. We test models continually to ensure quality, safety, and reliability.

Also joke is on them because MagicSchool uses ChatGPT.

I agree with most of this, also thank you cause it was insightful, I just wouldn't discount the minor actions currently not policed. It is not a crime to delete one's Instagram account (yet), but this has the tangible impact of reducing the amount of data Instagram can extract from you.

I mean obviously we still need to get policymakers to hold these companies accountable and make meaningful interventions to prevent these harmful designs across every layer of the digital stack. Hard things are worth doing though.

I think the best way to counteract this is by developing the political imagination of various individual & collective capacities of this power to act from our positions within our communities and institutions, and the tangible impacts (however minimal) that taking those actions can actually achieve.

1. We use the expression power to act rather than the more common philosophical term agency. From a Foucauldian perspective (Foucault 1982), power exists only in its exercise and is relationally constituted—it is not a latent capacity that a subject possesses but a situationally embedded possibility (Mühlhoff 2018; Vogelmann 2017; Saar 2007). The notion of power to act (Handlungsmacht) thus enables a graduated and relational description of situations, even where possibilities for action are highly constrained. Unlike agency, it does not imply an individualized inner capacity existing independently of its enactment. By contrast, when we speak of self-efficacy, we refer to the subjective perception of effectiveness—that is, the felt (in)ability to shape outcomes. Yet such perceptions are themselves effects of sociotechnical situatedness. Hence, power to act captures the relational and structural conditions of action, while self-efficacy describes their subjective experiential dimension.

I think they address they with how they use "power to act" rather than agency, to describe these contradictions. We still have power to act, even if those potentialities are highly constrained, but our inability to wield that power further constrains it.

When the Future Feels Foreclosed: AI Resignation and the Power to Act This article develops the concept of ‘AI resignation’ to capture how young people encounter AI not only as a helpful or flawed tool, but as an overpowering and seemingly inevitable force that can for...

Love this thread. You both might like this paper:
onlinelibrary.wiley.com/doi/10.1002/...
I'm wondering how educators can best challenge the feelings of helplessness these sociotechnical contradictions engender.

Crucially, in our analysis, AI resignation is not only an empirical finding but also the effect of a strategy of power: by engaging in AI resignation, subjects unwittingly participate in the wieldings of the power apparatuses of digital capitalism, because these apparatuses tend to benefit from subjects' resignation. This form of power thus operates by hollowing out the very basis of resistance—self-efficacy—so that experiences of participation, creativity, or deviation appear increasingly empty or structurally impossible.

This is so good, I wanna hang it on my wall.

"This article develops the concept of ‘AI resignation’ to capture how young people encounter AI not only as a helpful or flawed tool, but as an overpowering and seemingly inevitable force that can foreclose their sense of political and personal power to act in relation to the future."

because the rest are cowards

This explains a lot particularly around why TikTok's data centre development has been so aggressive and impatient

restofworld.org/2025/brazil-...

www.theguardian.com/technology/2...

Iran attacks reinforce need to reduce US dependence What you missed in February 2026

The attacks on Iran from the US and Israel further confirm that they’re rogue states operating outside international law. They need to be reined in, but we also need to reduce US leverage over us—which includes getting off US tech and developing alternatives.

The latest Disconnect recap is out now:

Global majority countries must embed critical minerals into AI governance Over the past decade, an increasing number of Global Majority countries—broadly characterized as economically developing countries in Africa, the Caribbean, Asia, Latin America, the Middle East, and O...

My latest piece for Science examines the global AI value chain, how critical minerals are reshaping geopolitical competition across Global Majority countries, and why oversight of these resources must be embedded into AI governance.

www.science.org/doi/10.1126/...

a map of openai's influence on other media companies

Amazing site here via @timnitgebru.bsky.social - a map of big tech influence on media and media companies.

imo this goes a decent way to explaining why coverage of AI specifically has been so shockingly bad recently. Very useful resource!!

nananwachukwu.github.io/media-captur...

Big tech is hungry for consumer data. Mass. needs privacy legislation now The heart of the privacy bill that the Massachusetts legislature is now considering is a simple but firm rule that companies should collect and use only the personal information that they need to prov...

Woodrow Hartzog & Neil Richards: "The Legislature should ignore the high-priced lobbyists and pass a law that actually protects us from data-hungry business practices that benefit no one but big tech." @hartzog.bsky.social www.wbur.org/cognoscenti/...

"In attempting to overcome dependency through purely technical means, they inadvertently created new forms of subordination."

Don’t Call It ‘Intelligence’ Humans are question machines. AI is an answer machine.

“Most of what a writer experiences is failure. Developing a voice takes years. The point is not to make it out of the woods quickly or unscathed. Getting lost is not the rough part. It’s the whole thing.”
—Charles Yu
@theatlantic.com

@authorsguild.org @penamerica.bsky.social @wgawest.bsky.social

Can social media age verification really protect kids? As countries start enforcing new age-limit laws, platforms like Roblox use facial technology — but critics warn of privacy leaks and surveillance.

“I used a fake ID of Stalin and it got accepted”

More countries are pushing social media platforms to conduct age verifications, but the methods are far from foolproof
https://restofworld.org/2026/social-media-age-verification-tools/?utm_campaign=row-social&utm_source=bluesky

How many of them are in the Epstein files?

Nine, mostly white dudes on stage in a manel.

Everyone wants to sign letters The Future of Life Institute puts out every few years, it seems.

Take a look at this manel which happened around their first letter 🙄

The billionaires and eugenicists on this manel are the actual existential risks to humanity we should worry about.

Inevitably they will blame psychosis. And we've seen this before with companies and academics claiming lung cancer is caused by stress not smoking!

Remember Hans Eysenck? www.theguardian.com/science/2019...

> This research programme has led to one of the worst scientific scandals of all time

1/n

We went from making up reasons like "weapons of mass destruction" to you know what, fuck it we don't need to even lie, the president "had a feeling" that they pose a threat so we just bomb them so we can. A country doing this can't expect peace inside.

At this point is there a tracker for chatbot related deaths?

Yeah, I'm trying to think about some of the big problems in science right now and be more thoughtful about how I work and how I contextualize my work. It's also really important to me we all critically examine the way academia functions. But that's long term, science should be slow.

"Cognition is a collective activity." Wow, I love this line, I'm less on the philosophy side of things, but love reading it, and especially spend a lot of time thinking about this (recently wrote a piece on this actually), so I'm excited to add Fleck to my reading list, thank you.

ChatGPT uninstalls surged by 295% after DoD deal | TechCrunch Many consumers ditched ChatGPT's app after news of its DoD deal went live, while Claude's downloads grew.

Yeah that's the shit

ChatGPT uninstalls surged by 295% after DoD deal | TechCrunch

techcrunch.com/2026/03/02/c...

Thank you, these all seem explicitly useful for my research! I've encountered similar findings already on the persistence of false beliefs and cognitive dissonance, so I'm keen to see what these add, I'm very interested in the idea of epistemic capture as you mentioned.

People or systems? To blame is human. The fix is to engineer Person-centered safety theories that place the burden of causality on human traits and actions have been largely dismissed in favor of systems-centered theories. Students and practitioners are now taught that accidents are caused by multiple factors ...

How many of us read it and didn't believe for science-based reasons vs not believing it because our social context told us not to?

We tend frame it as if we have one rational side and one set of idiots, but...attribution bias. Context is more complex to tweak. We're entitled to expect integrity!

Wait, do you have any paper recommendations on this?

Slide on Misinformation: Misinformation: the false dichotomy ➢ Most discussion of misinformation is overly simplistic: “With us or against us!” ➢ “Things are OK, back to normal!” vs “Things are not OK (because microchips/5G/etc)” ➢ Often led by institutional voices ➢ Frames misinformed beliefs as “human error” (and claims of stupidity, malice, etc) ➢ Proposed solutions generally involve shouting at (big budgets for PR campaigns) and/or punishing (social media bans) people until they do what they are told ➢ That’s about obedience, not understanding ➢ It also doesn’t work

Slide: Preamble: “human error” is not a cause ➢ One of the most common findings of an incident investigation, across fields, is “human error” ➢ That answers “Who is to blame?” ➢ But humans commit errors – that’s unavoidable! ➢ Preventing failure is a systems problem. We have to ask: “Why does the system allow a predictable event (human error) to lead to a significant failure?”

Slide: Preamble: “human error” is not a cause ➢ This does not mean there is not a place for accountability, but we have an obligation to learn from failure ➢ When the same error is widely repeated, we also have to ask: “Why is the system creating this error?” ➢ “Human error” is at most a component of a bigger systems failure, and often an excuse to stop thinking

Love these three slides especially, can't wait to read this now:
direct.mit.edu/books/oa-mon...
Thank you.