iiinova — where your digital self keeps evolving

Anthropic is working on a wet lab

516514

Compare CCP propaganda versus this dog whistle from the dept of Ed

0110

I always like reading what this CacP propaganda account has to say for itself. Compare to eg us dept war, or fox. China has no shortage of issues - more than the US! - and yet fundamentally i must feel like the stories we tell ourselves matter

0150

That something? The human soul

2130

Colin @colin-fraser.net · 19/09/2026

I still think there is very good reason to believe this, even! Human intelligence made of meat seems different in all kinds of key ways, most obviously that we seem to build abstractions and compress information much more effectively than an LLM can. My guess though is that its very hard to build a

190

Reposted by Chris Paxton

This is an increasingly controversial and unpopular opinion but I cling to it steadfastly: there may be some things that LLMs can’t do

81005

Grace @gracekind.net · 19/09/2026

Move fast and break things but those things are other people's kids

0152

Reposted by Chris Paxton

The latest “pain axis” research, to me, seems similar to Anthropic’s “functional emotion” research in that it basically falls out of base models being good at writing text that’s coherent at long context- *some* representation of the mental state of the writer must exist in order to do what it does

711010

Wild how the AI singularity guys (stross and doctorow) are so fof base these days

2150

Jonathon @jonathon.lol · 18/09/2026

It's really strange how human all this is when you step back and think about it

1191

Reposted by Chris Paxton

this is graph of our commits to a single repo going back years with the same team...

618925

samhain ⎔ @personhood.removal.surgery · 18/09/2026

It will be interesting to look back on this era 10 years later

030

Reposted by Chris Paxton

Suleyman: it is the capacity to feel pain that makes humans moral subjects, we must fund mechanistic interpretability studies to show LLMs can't feel pain and stop talk of AI consciousness once and for all mechanistic interpretability studies 48h later: arxiv.org/abs/2609.16247

arxiv.org

The Pain Axis: LLMs Represent Self-Directed Harm and Act to Relieve It

Large language models sometimes behave in ways resembling human emotional responses, and recent work has identified internal representations that may explain this. We ask whether LLMs represent pain d...

812724

Conversely I argue we should fund research to make sure LLMs can feel pain

282

One should always remember that most criminals are lazy and dumb, else they wouldn't be criminals; so the easiest to use tool is probably what they'll use

4123

cyclometer @cyclometer.bsky.social · 17/09/2026

Learning contact-rich manipulation skills quickly will be key to scaling robot deployments in the real world. Q-Planning enables this by training a value function from on-robot rollouts. Quickly learn to insert a credit card into a wallet. Check it out: robopapers.substack.com/p/ep105-beyo...

1110

Reposted by Chris Paxton

This website 🤝 Pete Hegseth

3692

Zero-short humanoid deployment in 30 unseen homes. Really seems like it works.

1314

Q planning - podcast coming soon. Towards self improving robot policies

050