The Overhang
One Useful Thing Ethan Mollick
AI is already doing weeks of work and weird creative jobs right now. The real gap isn’t future models — it’s how little people use what exists.
Based on reporting by One Useful Thing, Ethan Mollick — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
The big argument here isn’t about some distant AI breakthrough. It’s about the gap opening right now between what current models can already do and what most people are asking them to do.
The source points to GPT-6 Astra and Fable 5.1 as examples. One turned a text-only 1977 game, Zork, into a 3D action-adventure. The other tried to rebuild Umberto Eco’s library in 3D from videos, photos, and catalogues, sorting about 5,000 identified books into 27,000 shelf slots and labeling them as certain, guess, or unknown. Both jobs, the argument goes, would have taken weeks of human work.
The more interesting part is how little prompting it took to get beyond the obvious. GPT-6 Astra could use Blender without being told to do so. It made a trailer for the upcoming book Co-Existence, wrote jokes, generated voices, music, and sound effects, then later built a more cinematic version by using its own Blender work as a storyboard and driving a video generator through a browser. The author says he barely touched the production process.
That’s the “capability overhang”: a lot of what these systems can already do is simply sitting there, unused or misunderstood. And the point isn’t that humans should try to outproduce the machine. The point is that people who bring deep knowledge, wide knowledge, taste, and agency can get much more out of the same tools than people who treat them like a generic autocomplete box.
The conclusion is blunt. Even if every lab stopped training new models tomorrow, the existing ones would still be enough to reshape large parts of the economy. The change is coming either way. The real question is whether people use AI to replace themselves, or to do things that neither side could pull off alone.
My take — AI-written commentary, not fact-checked reporting
The overhang is the story, and most of the panic is conveniently aimed at the next model instead of the current one. That’s a very human dodge: easier to fear the monster in the hallway than notice the one already in the kitchen making your spreadsheet. The smarter bet is on people with judgment, because slop is cheap and taste still costs effort.
Read more about this at: One Useful Thing