It's almost as if everything Trump says about the negotiations bullshit and everyone who believes him is a cultist or a moron.

A thoroughly corrupt, debased hack. A malicious liar.

@kenwhite.bsky.social, one small step closer to Vitandus for JD

How Zentropi partners with Character.ai Character.ai takes safety seriously. With millions of users creating and chatting with AI characters every day, the team invests heavily in systems that help protect their community — and they're alwa...

Excited to publicly share the work Zentropi's been doing with the folks at Character to improve safety through SLM-driven classification and automated policy improvement! Please go check it out and hit me up if you have similar problems we can help with.

blog.zentropi.ai/how-zentropi...

more seriously congrats to the hungarian people for kicking that asshole out

How do you remember things and/or finish projects if not by leaving your research tabs open in the browser to cause you anxiety ?!?!?

But thanks to edge, they are well organized!

I have, conservatively, 1000 open tabs.

They're called workspaces and they're essentially windows of tab groups that you can close and open (to manage perf) but are auto-saved.

My most heretical tech opinion is that microsoft edge is the best browser because it’s just a fork of chrome with excellent side tabs, and side tab groups, *and* side tab group groups

Fuck yeah that launch was cool

2021 Republicans: How dare tech companies talk to government about moderating speech!!! We’ll create a whole false narrative about it and take it all the way to the Supreme Court!

2025 Republicans: We run things now so

march 26, 2026 The SAVE America
Act — and what's next 
The White House’s three-pronged plan to attack the midterms

Trump is pushing *hard* for the SAVE America Act to pass knowing that it’s bound to fail. Why? It's a pretext. Read about the administration’s much broader strategy to undermine the midterms and what we can all do about it.
🔗 www.ifyoucankeepit.org/p/whatever-happens-with-...

Raul Rivera yelling outside Mamdani presser.

Mamdani: “I want that man to be able to afford to keep living in NYC, because the day that I don't hear him yelling at me, it means he got priced out of this place.”

“We need to make this a city where it's affordable enough to yell at your politicians.”

It's exciting to see folks finding value in these tools - take a look at what @katieharbath.bsky.social built, it's very neat!

Thrilled to publish my latest article

BAN COOKIE BANNERS: A CASE STUDY IN TECH REGULATION

with the Harv. J. of Law & Tech!

A cri de cœur to end this worse-than-useless regulatory compliance regime and let something that actually protects users take its place.

papers.ssrn.com/sol3/papers....
1/

This means anyone building an agent can give it a principled, consistent way to evaluate content instead of hoping the LLM gets it right. Small step, but it makes agents more trustworthy by default. Give the skill at try here and let us know how it goes: github.com/zentropi-ai/skills/

Senator Tuberville on x writing a response to the @endwokeness account Original tweet from @EndWokeness - “less than 25 years apart” Image 1 - the twin towers as they are hit by a the planes on 9/11/2001 Image 2 - Mayor Mamdani sitting on a prayer rug while hosting an Iftar at city hall Quote tweet from Senator Tuberville - “the enemy is inside the gates.”

Let there be as much outrage from politicians in Washington when kids go hungry as there is when I break bread with New Yorkers.

Enabling Streaming Classification CoPE now enables streaming classification using a linear probe. This is an experimental methodology we developed to help ensure the safety of real-time generative AI systems.

The result is a running violation score as content generates - with enough lead time to actually act on it. Weights, training code, and tutorial are all open. You can find the repo at huggingface.co/zentropi-ai/... and a detailed writeup at blog.zentropi.ai/enabling-str... 🧵 4/4

So @samidh.bsky.social built a very lightweight probe on the internal representations of CoPE (our content evaluation model). As it processes each token, those representations already encode whether a violation is developing - the probe just has to ask. 🧵 3/4

When an LLM generates token-by-token and the user reads as it arrives, "complete text" means the user already saw it. If you're reliant on post-hoc classification, you've lost the best window for intervention and wasted a bunch of tokens to boot. 🧵 2/4

Historically, content classification tooling has been built around the assumption that you're evaluating a finished piece of content. That assumption holds fine for user-generated text. It breaks down entirely for streaming LLM outputs. 🧵 1/4

Please please read this from the archbishop of Chicago

www.archchicago.org/statement/-/...

Counterpoint - have you considered that human beings generally might have a lot less agency than we believe.

When there are no laws meaningfully regulating military AI use, the Pentagon’s promise to use AI “lawfully” mean almost nothing. The authoritarian AI crisis has arrived www.platformer.news/anthropic-pe...

Brennan Lee Mulligan - "Personality Predates Ideology" YouTube video by ogzombiebreakfast

"On the level of of individuals and civilizations, personality predates ideology. Meaning before you were a fascist, you were a bully and an asshole".

www.youtube.com/watch?v=3J2D...

I want you to put the word out there that we back up.

Some day we'll all regret this & we'll hold hearings & rend garments & pledge "never again" & we'll be just as full of shit as we were last time.

Zentropi is now integrated into Coop, @roost.tools's open source moderation platform. You can write a content policy in plain English on Zentropi, plug it into Coop as a signal, and have a moderation pipeline running in minutes.

Zentropi Now Powers Coop Zentropi labelers can now be used as classifiers within Coop, ROOST's Open Source Moderation Platform

@dwillner.bsky.social and I have spent years watching T&S teams rebuild the same infrastructure from scratch. This is what it looks like when open tools actually work together instead. Really proud of this one and appreciative of @roost.tools's leadership!

Details: blog.zentropi.ai/zentropi-now...