Hey {{first_name | there}},

By now, you have probably seen this video.

OpenAI gives GPT-6 Astra a task, and instead of just telling you how to do it, Astra starts doing it.

It opens software. Clicks around. Searches the web. Works with files. Writes code. Fixes things. And keeps going through a task without needing instructions every few seconds.

When OpenAI launched Astra on September 3, its president Greg Brockman said:

“Welcome to the AGI era.”

A few days later, Nvidia CEO Jensen Huang went even further: “AGI has arrived.”

Those are two of the most powerful people in AI saying something people have been predicting for decades.

So…Is AGI actually here?

I think the answer is still no, at least not in the way most people understand AGI.

But something important has definitely changed. Let's talk about it.

AGI stands for Artificial General Intelligence. And this is already where the problem starts.

There is no universally accepted test where an AI takes an exam, gets 90%, and everyone says: “Yep. That's AGI.”

Even OpenAI has used different ways of describing it over the years.

One of its clearest definitions is: AI systems that are highly autonomous and can outperform humans at most economically valuable work.

The important words there are most work. Being better than humans at coding is not AGI. Being better at maths is not AGI. Being an incredible writer is not AGI.

AGI is supposed to be broad. Give it accounting. Then coding. Then legal work. Then strategy. Then a completely new problem it has never seen before.

And it should be able to learn, reason and perform across all of them at roughly human level or better.

That's a much higher bar.

A few years ago, OpenAI reportedly created a simple five-stage framework for thinking about progress toward AGI.

It makes this whole debate much easier to understand.

Level 1: Chatbots

AI can talk to you like ChatGPT. You ask something. It responds.

Level 2: Reasoners

AI can solve difficult problems at roughly human level.

It does not just repeat information. It can think through maths, code, research and unfamiliar problems.

Level 3: Agents

Now AI can actually take actions.

Instead of telling you how to book a flight, it can use the browser and do the work.

Instead of telling you how to fix your website, it can open the code, change it and test whether it works.

Level 4: Innovators

AI starts producing genuinely new discoveries and ideas. Not remixing what humans already know. Actually helping push science, maths and engineering forward.

Level 5: Organizations

This is the wild one. An AI system could effectively perform the work of an entire organization. Planning. Research. Engineering. Operations. Finance. 

All working together with very little human help. So where are we today?

And perhaps we are starting to see pieces of Level 4.

Astra can already browse websites, operate software, fill forms, work with spreadsheets, build websites, test code and complete long computer tasks.

That is very clearly agent territory. But it gets more interesting.

OpenAI also says Astra has helped with long-standing open maths problems. On September 8, OpenAI published an AI-generated proposed solution to the Navier–Stokes Millennium Prize Problem, including a formal proof.

That starts sounding a lot like Level 4.

So if I had to place Astra on that old OpenAI ladder, I would put it here:

Level 3, with early signs of Level 4. Not Level 5.

And that distinction matters.

This is the part I think matters far more than the AGI label.

For most of ChatGPT's life, AI mainly answered. You asked for an email. It wrote an email. You asked for code. It gave you code.

You still had to take that output and do something with it.

That boundary is disappearing. Astra can now use the computer itself. Claude can use browsers, terminals and other tools. Coding agents can work through tasks for long periods.

The shift is: AI that tells you what to do to AI that does it.

And that changes things much faster than another 5% benchmark improvement ever could.

My answer is: Not yet.

At least not according to the stricter definition of a general system that can reliably outperform humans across most valuable intellectual work.

Astra is extraordinary. But it can still fail. It still needs humans.

Its strongest abilities are uneven.

And nobody has demonstrated an AI that can reliably take over the full range of work a capable human, never mind an entire company, can perform.

But I do think Brockman's wording may age better than Jensen's.

Maybe AGI hasn't arrived. But the AGI era might have.

We have moved from chatbots…to reasoners…to agents that can actually operate computers… and now we are beginning to see systems contribute to new scientific and mathematical work.

The line is getting blurry very quickly.

You do not need to spend the next two years guessing the exact date AGI will arrive.

It does not matter whether someone eventually decides September 2026 was “the month AGI happened.”

Something much more practical has already happened.

AI can do real work. Not someday. Now.

The person who learns how to delegate research, coding, writing, analysis and repetitive computer work to these systems today will have a much bigger advantage than the person waiting for some official AGI announcement.

So I wouldn't ask: “When will AGI change my job?”

I'd ask: “What part of my job can these systems already do today that they couldn't do six months ago?”

That answer is changing very, very fast.

And right now, that matters far more than what we decide to call it.

— Aashish

AI made PMs faster. Multiplayer mode is still broken.

A PM can summarize research, draft a PRD, and mock up a prototype before lunch. The hard part starts when the team has to decide what actually gets built.

Jira Product Discovery gives product teams one place to capture insights, prioritize ideas with consistent frameworks, and build living roadmaps stakeholders can rally around.

And because it’s connected to Jira, the context behind every decision stays with the work—so developers and their agents know not just what to build, but why.

AI helps PMs move faster. Jira Product Discovery helps the whole team build with confidence.

Get Started for Free.