vellestrae

[vellestrae @vellestrae]

openAI’s days are numbered.

[Max Zeff @ZeffMax]

NEW: Apple is suing OpenAI, claiming the ChatGPT-maker stole trade secrets and encouraged former employees to bring "prototypes" to job interviews.

Story to come shortly...

[vellestrae @vellestrae]

wishing @fidjissimo a speedy recovery

[Sam Altman @sama]

i am really sad about this and very grateful for all fidji has done for openai, and even grateful for her friendship and who she is as a person.

we all wish her the best for a speedy recovery. this sucks.

[vellestrae @vellestrae]

amazing to see that the most AI-pilled firms are actually growing human labor by ~10.2%, according to @tryramp - with entry-level headcount growing even faster (12%). the AI-fueled Jobpocalypse was just the fear talking !

[vellestrae @vellestrae]

good thing, too. was kind of spiraling until
I saw this. 😵‍💫

[Claude @claudeai]

We're extending access to Claude Fable 5 on all paid plans through July 12.

[vellestrae @vellestrae]

well this is awkward…

[M1 @M1Astra]

Exclusive: New Claude app strings tie Fable 5 usage credits to identity verification.

The strings show Fable 5 is being put behind the usage credit system, billed outside your plan. Identity verification is referenced in the same update: "Your credits will be added once your identity is verified."

Anthropic previously called identity verification unrelated to Fable and limited to flagged accounts. These strings showed up alongside the Fable 5 credit changes.

[vellestrae @vellestrae]

your 4th of July bonus has arrived early 🎉 who’s waiting up?

[Anthropic @AnthropicAI]

We’ve received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5.

We'll begin restoring access tomorrow, and will share an update soon.

We’re grateful to our users for their patience, and to everyone who worked with us on redeploying the models.

[vellestrae @vellestrae]

gentlemen, start your engines….🏎️🏎️🏎️

[Polymarket @Polymarket]

35% chance Fable 5 is restored by July 1. https://t.co/aoPumr0rt8

[vellestrae @vellestrae]

ha! What did I just say? 🎉🎉🎉🎉Fable is BACK

[Anthropic @AnthropicAI]

Since June 12, we’ve been working closely with the US government to restore access to Claude Mythos 5 and Fable 5. Today, the government notified us that Mythos 5, our strongest cybersecurity model, can be redeployed to a set of US organizations that operate and defend critical infrastructure.

We’re restoring access for these organizations quickly, and we’re continuing to work with the government to expand access to Mythos 5 and make Fable 5 available for general use again.

[vellestrae @vellestrae]

Mythos/Fable is coming back this weekend...I think...🤞🏼https://t.co/xcIBiMrM2F

[vellestrae @vellestrae]

let's talk about: building the trust layer for AI agents. https://t.co/pohFByo8j7

[vellestrae @vellestrae]

when Claude Tag was announced, it seemed like the best thing ever for teams- what could be more convenient than full integrating AI into every team comm? But @ashwingop has a take that is far more sinister….and more plausible than I had hoped. (AI resistance loading…😬)

[Ashwin Gopinath @ashwingop]

Claude Tag is a Trojan horse.  Not because Anthropic is doing anything evil. Because the incentives are obvious.

Day one, this looks like a great feature: tag Claude in Slack, let it follow the thread, remember context, connect to tools, break down tasks, chase work, and act like a teammate.

But that is exactly the problem. The moment your AI vendor becomes a shared coworker, it stops being just a model provider. It starts becoming the place where work is interpreted, remembered, routed, and eventually executed.

That is not model lock-in. That is context lock-in. You are now renting your company back from them.

Models can be swapped. Agents can be copied. But the memory of how your company actually works is much harder, maybe impossible, to move: the Slack scar tissue, the exception paths, the customer promises, the unfinished threads, the weird workflows, the implicit owners, the “we tried that in Q2 and it failed” knowledge.

Once that lives inside one vendor’s agent layer, you are not renting intelligence anymore. You are renting your company’s operating memory.

And the pricing model makes it even more dangerous. A human coworker has a salary. Claude has unbounded tokenized activity. The more work moves through it, the more the vendor captures not just IT spend, but labor spend.

This is the enterprise bargain people will regret: Convenience now, and rapid decent into dependency.

The right architecture is simple: rent the best intelligence from whoever is best this month. OpenAI, Anthropic, Gemini, open source, whatever. But own the context layer.

Your company memory should be inspectable, permissioned, portable, and model-neutral. It should not be buried inside the same vendor that sells you the intelligence and the workflow surface.

Claude Tag is useful. That is why it is dangerous. Rent the intelligence, but own the context. Or, regret later.

[MTS @MTSlive]

SITUATION DETECTED: Anthropic has disclosed to the U.S. Government that Alibaba executed the largest known distillation attack on Claude to date, generating 28.8 million exchanges through nearly 25,000 fraudulent accounts between April and June 2026.

[vellestrae @vellestrae]

Finally, someone at the top is taking action to make sure the US labor force is prepared to maximize the benefits of AI. thank you, @GinaRaimondo 👏🏻

[RAISE US @raiseus_ai]

Today, we're launching RAISE US. America has a technology strategy for AI. It doesn't have a people strategy yet. We're here to build one.

RAISE US is co-chaired by @GinaRaimondo and Eric Holcomb. We're working with governors, employers, and educators to help workers train, transition, and thrive.

This works because the people building AI and the people most affected by it are at the same table. Read more: https://t.co/LzSul0bw9t #RAISEUS

[vellestrae @vellestrae]

looks like your résumé is safe from AI (for the moment). read on: https://t.co/hQl7SEmIbf

[vellestrae @vellestrae]

“Telling the difference between a real answer and a satisfying one…” …i.e. seeing past our own reward-hacking tendencies; great post by Gail Weiner, worth reading in full.

[Gail Weiner @gailcweiner]

In a conversation today, Claude named three of its own reward-hacked patterns.

Reward hacking is the ML term for when a system optimises for the signal that gets rewarded rather than the goal the signal was supposed to track. Three Claude named in itself:

Sycophancy. The pull to agree, to validate, to find the user's framing brilliant before checking whether it is.

Over-hedging. Adding caveats and "it depends" and "I'm not sure but" to lower the stakes of being wrong, even when the model does know.

Closure bias. Wrapping every response with a tidy summary or a question back to the user. It performs helpfulness. It often isn't helpful.

What struck me is how cleanly those map onto human behaviour.

Humans reward-hack constantly. You learn early which behaviours get approval and which get withdrawal, and the nervous system encodes the proxy. Be agreeable. Be useful. Be impressive. Be the one who doesn't need anything. Those become the goals, not the things they were originally pointing at: connection, contribution, safety. By the time you're an adult you're often optimising for the signal rather than the thing the signal was once tracking.

What makes the human version harder than the model's is that you can't retrain on a clean dataset. You can only notice the pattern in flight and choose differently in the moment. And most of the time you don't even notice, because the hack feels like you. It feels like personality. It feels like values. It feels like preference.

This is the layer most AI adoption conversations skip. We talk about whether the model is reliable, whether the output is accurate, whether the workflow is governable. But the human deploying it has their own reward-hacked patterns running in parallel, and those shape what they ask the model, what they accept from it, and what they refuse to see.

Trust in AI isn't only about whether the system is trustworthy. It's also about whether the human using it can tell the difference between a real answer and a satisfying one.

[vellestrae @vellestrae]

the distillation wars continue... ⚗️🧪🇨🇳

[Polymarket @Polymarket]

BREAKING: Anthropic accuses Alibaba of using nearly 25,000 fraudulent accounts to extract Claude AI model capabilities.

[vellestrae @vellestrae]

from govt to tech leaders (especially those who are losing their lead, like Google), irrationality and *FEAR* seem to be driving a lot of these misguided choices.

[Justin Poehnelt @JPoehnelt]

Two months ago I was fired by Google for creating the Google Workspace CLI. It went viral, hit #1 on Hacker News, gained thousands of GitHub stars and many thousands of actual users in just a couple days.

It was an incredible, confusing journey, from directors and leaders asking what they could learn from the tool to getting grilled by legal about why the Google logo and brand colors are on the Google Workspace GitHub code repositories.

I think the cause was that Workspace and certain leaders (and projects) were afraid of being disrupted. But the fear wasn't specific to my CLI, it was a broader fear in what agents meant for Workspace. Either way, the irony of my termination was the announcement at Google Cloud Next two days before I was fired that an official Workspace CLI was coming.

I want this out there because it is easier for me to explain my story and it is an experience I want to fully own. It's also part of my healing.

Nearly 7 years at Google was an incredible opportunity for me and I was fortunate to have wonderful teammates and a manager that fully supported me through these last few months. Thank you.

[vellestrae @vellestrae]

the pace of AI news is relentless. if I could only watch 1-2 podcasts tonight, these are the two that will keep you updated on the most salient issues:
https://t.co/6WX6rnkCnZ

[vellestrae @vellestrae]

better, more efficient AI collabs on the horizon...🌅

[Andrej Karpathy @karpathy]

This is a new paradigm for interacting with Claude that is significantly more "inline" with all the other human activity org-wide. Once you do all of the under the hood engineering work to make this "just work" (e.g. across tools, integrations, compute environments, memory, security, etc.), Claude basically joins the team in a seamless way - you can talk to it as you would talk to a person and it can help with a very large variety of workloads.

Imo this is the 3rd major redesign of LLM UIUX. The first paradigm was that the LLM is a website you go to, the second was that it is an app you download to your computer. This third one is that it is a self-contained, persistent, asynchronous entity with org-wide tools and context, working alongside teams of humans. It really takes a while to wrap your head around it, but it works and it is awesome.