ChatGPT, Claude, and Grok All Crashed. OpenAI Had Declared the AGI Era Hours Before.

OpenAI spent this week telling the world it built something close to artificial general intelligence. Within hours, ChatGPT, Claude, and Grok all crashed at the same time. That is not a coincidence worth ignoring, it is the whole story of this newsletter in miniature: a huge claim, followed almost immediately by a reminder that the plumbing underneath is still held together with duct tape. Here is what actually happened today, and what it means for a business that cannot afford to bet everything on one tool having a good day.

Small teams don't have time to stitch together five tools and hope it works.

Apollo gives you everything you need to find leads, reach them, and close deals — all in one place:

Stop juggling tools and start building pipeline that scales.

With Apollo, the AI revenue engine powering 4M+ users.

Sign up for free

What happened: OpenAI released GPT-6 Astra this week, and company president Greg Brockman told reporters he personally believes it represents artificial general intelligence, calling it the start of the "AGI era," according to Axios. OpenAI says Astra scored 98.6% on ARC-AGI-3, a test built to measure reasoning on problems a model has never seen, up from 7.8% for the model it released six months ago. But The New Stack found the comparison was not apples to apples: Astra ran on a customized evaluation setup that the other models in the comparison did not use. Access is rolling out first to trusted partners, then to ChatGPT Plus, Pro, Business, and Enterprise users over the coming days.

Why it matters to you: Every few months a lab declares the "AGI era" has arrived, and every few months the benchmark turns out to have an asterisk. That does not mean Astra is not a real jump in capability, it likely is. It means you should not restructure your business around a press release. Judge a new model by what it does on your actual work, not by what a company says about itself on launch day.

What to do about it: If you use ChatGPT for coding, writing, or research, run your usual tasks through Astra once it reaches your plan and compare the results to what you already use, before you pay for an upgrade or change your workflow.

What happened: Hours after the Astra launch, ChatGPT and Codex, Anthropic's Claude, and xAI's Grok all reported outages within the same window on September 3, according to Axios. Microsoft Azure, which provides cloud infrastructure behind several of these services, was also having problems at the same time, and multiple outlets pointed to it as a likely shared cause.

Why it matters to you: If your business leans on one AI chatbot for customer replies, content, or code, and that one service goes down, you have no backup by default. The "smartest model ever" announcement and a multi-company outage happened on the same day. Intelligence and reliability are two different things, and only one of them keeps your business running on a bad afternoon.

What to do about it: Keep one alternative ready, whether that is a second chatbot subscription or a saved template you can use by hand, for the tasks that cannot wait if your main AI tool goes dark.

What happened: Nvidia has agreed to acquire Hugging Face, the platform that hosts the open AI models many free and low-cost tools are built on, for roughly $12.9 billion, according to Nvidia's own announcement. CEO Jensen Huang said Hugging Face will "remain an open platform for the entire AI ecosystem" and that developers can keep using any cloud, framework, or hardware they choose, not just Nvidia's.

Why it matters to you: Hugging Face is the quiet backbone behind a lot of free or cheap AI tools small businesses use without ever seeing the name. Nvidia already controls most of the chips that train and run AI models. Now it also owns a major piece of where those models get shared and downloaded. Nvidia says nothing will change for users, but a company that owns both layers of the stack has a lot of room to change its mind later.

What happened: Senator Bernie Sanders and Representative Greg Casar introduced the Ban Artificial Superintelligence Act, which would prohibit developing or deploying AI systems that match or exceed human cognitive performance across the board, pause advanced AI development until a new federal AI safety agency is running, and set penalties that include shutting down violating companies and up to 20 years in prison for individuals, according to a summary from Senator Sanders' office.

Why it matters to you: This bill is unlikely to pass as written, and Congress has floated aggressive AI legislation before that went nowhere. But it shows that "let the labs regulate themselves" is losing support in Washington, on the left and increasingly on the right too. If your business plan assumes frontier AI keeps improving on schedule with no interference, this is one more sign the regulatory ground could shift faster than that plan expects.

What happened: Los Angeles Unified School District barred all students, from kindergarten through 12th grade, from using generative AI tools on district devices for the 2026-27 school year, according to K-12 Dive. That is broader than New York City's policy from the day before, which pauses AI only for elementary and middle schoolers while leaving high schoolers eligible for supervised pilots.

Why it matters to you: If your business touches schools, tutoring, after-school programs, or youth products in any way, the ground under "just add an AI feature" is moving fast. Two of the country's biggest school systems moved to restrict AI within a day of each other.

What happened: Thinking Machines, the AI startup founded by former OpenAI CTO Mira Murati, is reportedly in talks with Accel to raise $1 billion at a valuation of at least $40 billion, according to TechCrunch. The company's annual revenue run rate is a little over $100 million, and it raised $2 billion at a $12 billion valuation less than a year ago.

Why it matters to you: A company can now be valued at roughly 400 times its revenue on the strength of who founded it and what it might build someday. That is not a verdict on whether Thinking Machines succeeds, it might. It is a reminder that a lot of today's AI valuations are bets on a story, not on current numbers, and stories can deflate as fast as they inflate.

What happened: Google is rolling out Gemini-powered "Live" voice modes for Gmail, Docs, and Keep, letting users search their inbox, draft documents, organize notes, and manage tasks by talking instead of typing, according to 9to5Google.

Why it matters to you: If you already run your business on Google Workspace, this is a free upgrade worth trying, not a reason to switch platforms. Voice search across your inbox and docs can save real time on the boring parts of running a business, like digging up one email or invoice from three months ago.

What to do about it: When the feature reaches your account, try it on one repetitive task first, like finding an old email or drafting a routine reply, before you build it into your daily workflow.

Every AI company wants you to believe today's launch is the one that changes everything. Sometimes it might be. But the pattern this week tells you what to actually watch: not the press release, but what happens when the servers get busy, when Congress gets nervous, or when a school board says no. Stay skeptical, keep a backup plan for the tools you depend on, and let the AGI debate be someone else's argument to have on a Friday night.

If this issue saved you from falling for one AI headline today, forward it to one person who needs the same reality check. And if someone forwarded this to you, welcome aboard, you can subscribe to get tomorrow's issue in your inbox.

Talk tomorrow,
Mark Shilensky

Follow me personally on Social Media:
Facebook: https://www.facebook.com/MarkShilenskyPage
Instagram: https://www.instagram.com/markashilensky/
Threads: https://www.threads.com/@markashilensky
LinkedIn: https://www.linkedin.com/in/shilensky/
X: https://x.com/markshilensky
TikTok: https://www.tiktok.com/@mark.shilensky
BlueSky: https://bsky.app/profile/markshilensky.bsky.social