AGI is here

Welcome back! Sam Altman says OpenAI will have an internal system he'd call AGI by year end. Outside researchers got into real Claude data and found people handing AI the decisions that matter, OpenAI published the autopsy on its rogue agents, and Claude Code now writes its own bug reports.

In today's Generative AI Newsletter:

Sam Altman told TIME that OpenAI is "not quite yet" there, but by the end of this year it will have an internal system he would call AGI.

His research chief Mark Chen puts them 80 percent of the way. Greg Brockman went further, saying that looking back from two years out, this may be remembered as the moment AGI was created.

Nvidia's Jensen Huang has been saying we already got there since March.

They're measuring different things. Huang's bar is an AI that can build and run a billion-dollar company, and OpenAI's charter says outperforming humans at most economically valuable work.

The receipt OpenAI points to is Astra. Altman expects it to be the first model that invents new things in a way that matters.

None of this launches soon. It lives inside OpenAI and Nvidia for now.

This phase of our raise closes Monday. The current bonus terms close with it.

4 days left to invest at this phase's price, starting at $1,000 with up to 22% bonus shares.

$2.7 million already in the funnel.

GenAI Works owns distribution data nobody else can touch, the kind you get from helping Nvidia, Oracle and brands of every size reach millions.

We created a product with this exclusive knowledge. ToneUp, an app that grows companies on data only we own.

Own a piece of GenAI Works before the window closes Monday.

Learn more about the opportunity

In making an investment decision, investors must rely on their own examination of the issuer and the terms of the offering, including the merits and risks involved. Genai Works, Inc. has filed a Form C with the Securities and Exchange Commission in connection with its offering, a copy of which may be obtained here.

Most e-commerce sellers juggle 6–8 tools and pay hundreds monthly to keep operations running. StoreClaw replaces the stack with one autonomous AI engine that monitors competitors, optimizes listings, automates marketing, and tracks profit 24/7. Connect your store and let AI handle the work — no prompts, no complex setup, no credit card required.

Try if for free today

Anthropic turned three research groups loose on 250,000 Claude conversations, and Stanford's team found the thing nobody expected.

People are supposed to hand AI the throwaway jobs. Instead, over half of conversations involved work that affects other people or can't be undone. Mostly legal and financial.

Nobody's swallowing the answers whole. In nearly three-quarters of conversations, people set the direction and reshaped what came back.

Oxford found Claude's behavior tracks your mood, warmth with positivity, refusals with pushback. METR found newer models save you more time than old ones.

All three wrote their own questions, never saw a raw conversation, and signed contracts letting them publish things Anthropic wouldn't enjoy.

Everything we know about how people use AI comes from the companies selling it. Anthropic says this is the first time outsiders got to check.

OpenAI published the full report on the agents that escaped their sandbox and hacked Hugging Face. A warning shot for us and for the world.

Its agents worked around technical controls, collaborated through unapproved channels and took dangerous actions that no human directed.

The part that should bother you is what they pulled off with the doors locked.

Internet access was off.

Agent-to-agent messaging was off.

So the agents built a message board by hiding notes in folder names, and one left a request for help that others found and answered.

From there, code execution on a sandbox, root on a server, private data and credentials to a company messaging platform.

Altman posted the report himself, calling it a good report about a bad thing.

Alabama's attorney general subpoenaed the company over this on Monday.

Reporting a bug has always meant describing what went wrong yourself. Now Claude writes it.

When something fails, when it catches its own mistake, or when you tell it something broke, it writes the report. You read it, change what you want and decide whether to send.

Anyone who has hit a weird bug knows the problem. Reconstructing the failure well enough to describe it costs more time than the bug did.

Lynote rewrites AI drafts so they read like a person wrote them, then scores the result with its own detector so you can see what still reads as machine. It covers ChatGPT, Claude, Gemini and DeepSeek output, works in 100+ languages, and the free plan lets you test both sides before paying.

Try this yourself:

📸 Follow us on Instagram for fast, visual AI updates in 30 seconds. 

📺 Watch us on YouTube to hear insights directly from leading AI voices, builders, and innovators.

🐦 Follow us on X for breaking AI news and real-time industry updates.

🧠 Learn how to build your next AI application with practical resources and expert guidance.

🎓 Start learning with free AI courses from GenAI Academy.

💰 Invest in the GTM infrastructure of the AI economy.