@cichuck

Founder, Collective Intelligence - #ML, #AI - a somewhat #unaligned data scientist and frustrated #economist - exploring agenticism

 Pennsylvania, USA
collectiveintelligence.com
Joined October 2010
  • Tweets152K
  • Following15.4K
  • Followers19.9K
1.5KMedia
Ask three people what #RSI means and you’ll get three different answers: a self-rewriting machine, AI doing AI research, or an agent editing its model weights. We’re debating whether RSI has arrived before agreeing what it is. SOP in #AI land
156
Navier-Stokes and 400 additional proofyarifacts were shat out by OpenAI indicating that vibe mathing is a thing. vibe physics is next?
109
Open Shell is back. Jensen is resurrecting it and wrapping it into an agent safety platform. We saw this coming in March: medium.com/@chuckrussell/dec…

DeClawed — NVIDIA’s Less Chaotic Agents

Our Assessment · March 2026 · Collective Intelligence Inc.

chuckrussell.medium.com
134
nor'easter break - no token Saturday
221
I just published The Programmable Probabilistic Decision Primitive chuckrussell.medium.com/the-…
97
Is Jev 'just' a zero shot classifier? An instance of a new class of AI decision primitives? It behaves like a programmable, probabilistic decision primitive (PPDP) It might find a home in your Agent's harness.
112
A safer frontier AI.
1
1
73
more here: chuckrussell.medium.com/the-…

The Frontier

Every frontier gets settled. Not every settler makes it.

chuckrussell.medium.com
28
Brakes aren't what slow a car down. They're what let you drive fast.
1
215
Artificial intelligence giants Anthropic, OpenAI, SpaceXAI and Google violated antitrust laws with a coordinated effort to decelerate AI development, according to a new lawsuit.
2
1
1
197
#TSLK: an AI doomer metric measured in hours that satisfies FSD, quad/bipedal robotics and agentic AI use cases. Every AI product has a TSLK. For most, TSLK approaches infinity. Battlefield metrics are likely classified but TSLK rapidly approaching zero here.
83
I admire #alexkarp For his insights regarding AI, AI safety, and concerns related to income inequality. @PalantirTech
154
Today, I'm calling on Congress to press pause.
167
the AI Doomer economy is in a bubble literally and figuratively
3
193
The frontier labs don't need to pause AI model development. They do need to reconsider performing irresponsible experiments conducted by amateurs. Negligence caused the Hugging Face hack. OpenAI should be held accountable for that.
1
237
The vendor demos underplay the part I value most. Grok Bot is supposed to stop when it needs approval, not because the model is unsure of itself, but because some steps aren't the bot's to automate. A payment. An email send. A sentence you'll have to defend. #GrokBot @bot
🤖 Made with AI
1
135
chuckrussell.medium.com/a-ho…
🤖 Made with AI

A Hot Desk of Its Own

Grok Bot is the first assistant I’ve used that completes useful work in the systems-of-record that I live in.

chuckrussell.medium.com
41
Grok Bot calls itself a container. The kernel says it's a virtual machine. #GrokBot @bot
🤖 Made with AI
1
147
chuckrussell.medium.com/agen…
🤖 Made with AI

Agents Perceive the World in Pixels

Exposing the Grok Bot terminal.

chuckrussell.medium.com
50
An AI agent has no body. It's weights and a context window. To make it open a browser or click a button, you have to build it hands. We pulled a process map from inside a running Grok Bot sandbox. 154 processes. The machine is more interesting than the pitch. #GrokBot @bot
🤖 Made with AI
2
191
chuckrussell.medium.com/agen…
🤖 Made with AI

Agents Perceive the World in Pixels

Exposing the Grok Bot terminal.

chuckrussell.medium.com
41
Load more