GP @nazarevc investing in AI infrastructure & compute | ex-Pantera Capital | Cambridge PhD | Essays on AI → robotwave.substack.com
Joined November 2006
- Tweets 1,456
- Following 1,613
- Followers 9,105
- Likes 3,418
114 Photos and videos
this
The risk of the government deciding that a model is too dangerous should only add to the reasons why open source models running on local hardware can be a reasonable alternative.
148
Oops
Jun 10
In addition to transparency, I now believe frontier models should face mandatory third-party testing for cyber, bio, and autonomy risks—with the power to block or revoke deployment of models that pose catastrophic risk.
219
The Off Switch - new essay
On Friday at 5:21pm, one letter from the US government switched off Anthropic's two newest models for every customer on earth.
1
180
Great piece
176
And this marks the end of the era when everyone had access to the SOTA frontier models. From here on the top models will always be reserved for governments and labs.
Jun 13
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: anthropic.com/news/fable-myt…
2
163
Dr. Steven Waterhouse retweeted
Jun 12
huge new agentive manipulation features new supported robots dropping in production soon
Jun 12
Claude Fable 5 integrated reachability analysis into our G1 Manipulation framework
Now the humanoid knows what its arms can reach before it ever tries
Features:
5D G1 arm capability map in a gravity-aligned built from 2M FK samples per arm with self-collision checking
Planning context creation dropped from 100ms to 62µs
14 joint dual-arm control, solving QP differential IK for both arms
2
12
1,182
Dr. Steven Waterhouse retweeted
Jun 8
Recursive self-improvement going from being a ridicule-worthy fringe sci-fi concept to a completely normalized part of the discourse which is “obviously the plan” is one of the more dramatic Overton window shifts I’ve experienced
There’s something disorienting about it, like if the sky suddenly turned red, and everyone acted like it had been that way all along
69
105
1,314
73,113
Strangely was the same regulatory capture game SBF was trying to play with FTX. Turn all defi into FTX regulated trades and then push regulation to remove the rest. Is this an EA skill they are taught?
Jun 10
i read through dario's blog post, anthropic's blog post, and their 19 page "advanced ai framework" so you don't have to.
regulatory capture. they are going for regulatory capture.
if i had to guess:
* raise money through their ipo
* use the money to push their regulatory capture agenda
* force lock in on enterprise customers (this is coming)
* build kill switches into competitor models
* etc.
the main thing they wanna control is automated r&d, which is why they conveniently bury it at the very end of all their fear mongering as the fourth pillar of fear or existential threat or whatever
3
1
12
2,476
2/ Anthropic shipped Claude Fable 5 this week: its most capable model, built to run for hours and check its own work. It also arrives with parts removed. Ask about cybersecurity, biology, or distillation and you get routed to the older Opus 4.8. The full model, Mythos 5, you can't touch.
1
216
5/ Anthropic named its tiers. The one you can use is Fable. The one you can't is Mythos. A fable is a fiction with a lesson; a myth is a story everyone agrees to treat as true. They kept, for themselves, the word for the story you're meant to believe.
1
74
6/ Full piece: AI Waves # 13: One Claude To Rule Them All
robotwave.substack.com/p/one…
59
open.substack.com/pub/robotw…
1/ Fable:
a short, fictitious story designed to teach a moral lesson or nugget of wisdom
1
3
444
Dr. Steven Waterhouse retweeted
Jun 11
.@ylecun is promoting a decentralized AI training protocol.
“Project Tapestry will build a new open-source platform to enable distributed, globally federated training of frontier open models.”
thealliance.ai/blog/ai-allia…
Jun 10
Everyone, please join Project Tapestry
thealliance.ai/projects/tape…
6
3
54
8,679