Shoutouts to the EA Impact Hubs! 
"Impact Hubs are affordable, temporary coworking and coliving setups for EAs and impact-focused people. The idea is to make it easier to do focused work and build connections with other EAs without having to be in an expensive EA hub like London or SF."

It's pretty impressive to see an online course for an area with only ~2.5 FTEs working on it. This does seem like one way to get more people to work on it, so kudos to you!

I attended some PauseAI UK protests a year or two ago and wanted to share my worries and experience attending my first AI protests. I felt reluctant but was glad I went.

I was worried that attendees would be angry or radical or extreme, that it would be loud, and that nuanced communication would be sidelined in favour of emotional persuasiveness. Relatedly, a part of me was worried what others would think of me if they thought of me as 'the kind of person that goes to protests'. And something about me fighting for proposals that I wasn't myself sure about.

But it seemed like it would be impactful to attend and could be valuable for me. So I went, and I even made and brought a sign!

In actual fact, I was happy to meet other people that I thought were nuanced thinkers and whose work I respected. I enjoyed talking to like-minded people. I liked flyering and I was able to have fine-grained discussions with a range of passers by and other protesters.

It was at times loud or featured talking points that I felt simplified an issue in a way that I didn't like, but overall it seemed like a good way to hang out on a Saturday and contribute to something important to me.

Thanks Michelle!

I'm very happy that PauseAI UK are doing this work and I've enjoyed attending their protests.

I have also been extremely impressed with ControlAI's work. Similarly, they make it easy to contact your representative (in a range of countries). Myself and a bunch of my (non-EA) friends have contacted our MPs through ControlAI's tool.

People can also use Microcommit to receive regular suggestions of very short advocacy activities, from ControlAI, PauseAI, and others.

More emotional reaction options could help other users to differentiate between well-thought-out responses and responses published while still in an emotional state. They could also be used to gauge the emotional state of the community.

This comment was written hastily without being carefully considered.

We've had a bunch of people stay with us over the years. It's been very nice to meet people and to be able to offer them help. It's been especially rewarding to help those who would find it difficult to stay in London otherwise. Finally, it's great to help people meet other EAs and have a special experience when they visit London.

As a step towards some kind of 'centralized' plan, BlueDot/Adam Jones just posted an overview of many different AI plans, grouped into categories.

Key paths, plans and strategies to AI safety success – June 19, 2025

How does Animal Welfare/Global Health affect AI Safety? Very brief considerations.

I think someone might build super strong AI in the next few years, and this could affect most of the value of the future. If true, I think it implies that the majority of any value from an intervention or cause area comes from how it affects whether AI goes well. Even if that's very slight and indirect. Relatedly, I think whether AI goes well depends on whether states will be able to coordinate.

How do Animal Welfare interventions affect whether AI goes well? 
– I think the Moral Circle expansion is relevant.
– Helping reach climate targets seems relevant to help with international coordination.
– But I think that Animal Welfare interventions place a cost on society such as by raising the price of food and increasing pressure on our governments in high-income countries.

How do Global Health interventions affect whether AI goes well? 
– I think that it reduces the pressure on governments in LMICs and gives them a safer society. This gives their Governments slightly more room to come to peaceful international agreements. 
– But it may also enable more people to contribute to AI, whether that be AI capabilities development, chip manufacture (or AI safety/governance)

Overall, I slightly lean towards global health being better. Perhaps RP's tools shed light on this. (I haven't checked!)

Yes, very happy to hear about this! I've reached out to you on here @matthes 

Nice post!

Thank you! Since it's (my) first post, it's helpful to have some positive encouragement.

1. We actually intended not to give examples of those.
2. That's useful feedback on the diagram, thanks.