AI, science, geopolitics, and the data behind the story. Building a knowledge graph that powers global coordination. Head of growth @geoprotocol
Paris
Joined December 2011
- Tweets945
- Following215
- Followers138
- Likes326
Adam retweeted
"Men don't want an AI companion to have a conversation with it"
37% of people voted that AI companions will replace human relationships.
Watch the full debate & cast your vote on geo: geobrowser.io/space/41e85161…
Adam retweeted
Introducing Geo Debates! 🎉
Let’s get closer to the truth, together.
Explore short-form video debates on the questions that matter. Hear different perspectives. Make your case. Vote on what you find convincing.
Join the conversation today!
geobrowser.io
You’ll be able to try the coolest product I’ve worked on next week. It’s fun, unlike anything you’ve seen before, and it makes the world a better place.
More on this in a few days! @geoprotocol
"As 2025 came to a close, Oxford University Press chose the phrase “rage bait” as its word of the year. Its use had tripled over the previous 12 months, a clear sign that a growing number of people were aware of how they were being manipulated online through outrage."
The world is becoming polarized. So how do we meaningfully fix it?
We believe three important steps are necessary:
- Broadening our worldview
- Cultivating curiosity
- Building bridges and common ground with people we don’t agree with
That's why we're building Geo debates.
ualberta.ca/en/folio/2026/01…
particularly interesting question. why have people stopped believing in science?
The US is having its worst measles surge in decades.
At the same time, one of the most persistent health beliefs in America is back at the center of politics:
“Vaccines cause autism.”
Scientifically, this question has been studied extensively.
Socially, it is still very much alive.
That gap is interesting.
How can a claim remain politically powerful long after the evidence has mostly settled?
The current context is real: the US has recorded thousands of measles cases this year, while vaccine coverage has fallen below the level generally needed to prevent sustained outbreaks.
Adam retweeted
AI companionship is way further along than most people realize.
There are now dedicated AI dating apps where users can create a girlfriend or boyfriend, choose their personality, and build an ongoing relationship over time.
Replika is one of the biggest examples. EVA AI has even opened an AI date café in New York.
The category is also moving off-screen.
Sharp just unveiled a physical AI companion designed to remember you, respond emotionally, and grow closer over time.
Governments are starting to react. China is restricting some AI companion products over concerns about emotional dependence and isolation.
And the numbers are already strange:
• 24% of US social/emotional AI users say they’d miss their AI companion “a lot” if it disappeared
• 20% say AI can be more reliable than friends or family
• 20% have asked AI whether their relationship with AI is becoming unhealthy
• 15% say AI has sometimes made them feel less in touch with reality
This isn’t really a niche “AI girlfriend” story anymore.
People are using AI to talk through loneliness, relationship problems, stress, difficult decisions, and things they might otherwise tell a therapist or close friend.
That raises a genuinely new set of questions:
When does a tool become a relationship?
Can something be emotionally real if only one side is conscious?
Should companies be allowed to design AI to make users more attached?
And if someone becomes deeply dependent on a companion, who gets to decide when that relationship has become unhealthy?
Useful tool, genuine relationship, or something entirely new?
Adam retweeted
“Killer robots” are no longer a sci-fi debate.
The UN and Red Cross are now calling for binding international rules on autonomous weapons, warning that we’re getting close to a “moral red line”: machines independently selecting human targets.
The technology is moving faster than the law.
Ukraine is already scaling armed robotic systems. Militaries around the world are investing heavily in drones, autonomous targeting, and AI-assisted warfare.
And the core debate is harder than “robots bad.”
The case for autonomy:
Machines can react faster than humans, operate where soldiers would die, and perhaps even make fewer mistakes than exhausted or frightened people.
The case against:
If a machine identifies the wrong person and kills them, who is responsible?
The commander?
The engineer?
The manufacturer?
The state?
The algorithm?
And there’s a deeper line:
Should any machine ever be allowed to decide that a human being should die?
news.un.org/en/story/2026/08…
What happens when a community creates a dream together?
Reflections on our relationship to time, myth-building, and community by my friend and poet Emma!
poemma.substack.com/p/a-myth…
"People don't know what to believe anymore" seems to ring true with a lot of people I've been chatting to recently. From academics to content creators - there's a feeling that we're living through some sort of polarization & truth crisis.
Adam retweeted
Was fun hopping on with @itsciaranlyons to talk about agents, knowledge graphs, and the state of the crypto industry.
Green shoots are popping up! 🌱
xitter.cf/i/broadcasts/1RJZzBopv…
Human sovereignty is usually discussed in terms of money, identity, privacy, or political power.
But there is another kind:
Epistemic sovereignty.
The ability to understand how you came to believe something.
To inspect the evidence.
To compare interpretations.
To recognise uncertainty.
To revise your beliefs without surrendering the process to an authority.
For most of the internet era, we navigated information ourselves.
Badly, often.
But we still searched, compared, and constructed our own mental models.
AI now offers to do that work for us.
It gathers the sources.
Synthesises the arguments.
Resolves the contradictions.
And returns a clean answer.
The convenience is undeniable.
But every cognitive task we delegate is also a skill we may practise less.
A world where humans and AI reason together could greatly expand our intelligence.
A world where AI generates our conclusions and humans merely consume them could create unprecedented intellectual dependence.
The distinction will depend on how we build the knowledge layer underneath AI.
Can we inspect the evidence?
Can we see meaningful disagreement?
Can we challenge the model’s framing?
Can communities correct the underlying knowledge?
Can we build upon it collectively?
Or do we receive an answer and trust that someone, somewhere, has already done the thinking correctly?
AI should extend human judgment.
It should not replace our responsibility to form one.
The idea of composability is core to what we’re building at @geoprotocol. It was already a core part of @timberners_lee’s vision when he created the WWW!
The web’s breakthrough was more than making information public.
It was making information reusable.
A page created by one person could be linked by another, indexed by a third and become the foundation for something its author never imagined.
Tim Berners-Lee called this the web’s “unexpected reuse of information.”
AI needs the same property at the facts and ideas.
Knowledge shouldn’t just be accessible. It should be composable.
When should the government get to decide what happens to your body?
@realDonaldTrump has overhauled childhood vaccine policy, emphasizing parental choice and freedom from government coercion.
But his administration also supports restricting abortion access.
If bodily autonomy matters for vaccines, should it matter for abortion too?
Quoting a professor of logic and philosophy I was speaking with yesterday: “respectful, well reasoned debates with evidence, is how we talk about complex moral and social issues. It’s what we’re missing in today’s America”
Our environment promotes certainty, outrage, and people we already agree with.
An idea from @DAcemogluMIT new book feels especially relevant to the internet today:
A healthy democracy depends on accepting that we might be wrong.
That requires exposure to disagreement, open debate, and communities where ideas get challenged.
Better public discourse doesn’t require less disagreement. It requires making disagreement useful.
Plot twist - the only ressource aliens need is compute & they’re just waiting for us to build all these data centers… 👀
Adam retweeted
AI isn't just changing where we look for answers. It's changing how early we stop thinking.
Microsoft and Carnegie Mellon surveyed 319 knowledge workers. The more confidence people had in the AI, the less effort they spent checking its work. The buried assumptions, the open questions all quietly drop out of view.
The fix isn't distrusting AI. It's giving it knowledge that shows its seams:
- Where a claim came from
- Who disputes it
- What changed, and why
Friction isn't a flaw but how understanding happens. That's the bet behind Geo.
