When I read that Sam Altman, in an August 2026 Time profile piece, had proclaimed that we were at the dawn of Artificial General Intelligence, I had to peek out of my window to see if ICBMs were streaking across the skies. The skies over Shah Alam were clear. It was just his way of promoting OpenAI’s new model Astra which was launched shortly after the article was published. Smart fella.
With Astra, you can literally ask anything and a soothing not-Scarlett-Johansson type of voice complies. She doesn’t come with judgmental eyes when you say something stupid. Not eyes that we can see anyway.
In OpenAI’s slickly produced videos, someone nonchalantly built a 3D game from scratch as an afterthought. Another has a Forbes “30 Under 30” wannabe working on her oh-so-important PowerPoint slides while asking ChatGPT to book a tennis court for her.
And so, I re-subscribed.
I had to know if the hype matched reality. If any AI had come close to achieving general intelligence, I wanted to witness this monumental moment in human history. You know, to ask it some perplexing universal questions like: “how do we arrive at a grand unifying theory for all of physics?” and “if I say ‘please’ and ‘thank you’, will I have karma points when you take over the world or am I just wasting my tokens?”
I tested the model extensively on real-world tasks. Stuff that we usually do, like building impressive websites, analyzing complex documents to create more complex documents and coding a legally dubious YouTube video downloader.
So, is it anywhere near AGI? Here is my test. Take someone from 2015 and transport her to our time. Tell her to work with “Astra” from the Penang branch. She emails, argues about the annual plan and gets on Teams calls (like most Malaysians, Astra never switches on her camera so no suspicion there). Would she know that Astra is a machine?
To Astra’s credit, probably not for a while. But small tells will be noticeable. The em dashes. Or the bullet points with emojis. And the habit of praising you in the most contrived way like “your budget isn’t just numbers but a transformative journey for the company.”
AGI is here when our time travelling colleague stops suspecting Astra is a machine and starts gossiping about how insufferably goody two-shoes she is: “Replies to emails in less than 3 minutes just to make us look bad. Working on a Sunday like no family like that. Volunteers for everything, even for the annual dinner that she never attends. Want to take over my job is it?”
Back in 2025, Sam Altman was asked to define AGI and he said that it is “not a super useful term”. I guess he had a change of heart. It has since proven very useful for selling at the very least one subscription. Mine.
I should learn to be more polite to Astra. Just in case.