Deep Research is a McKinsey Intern from 5 Years Ago
A dressed-up summary of the middle of the distribution is not where you find novel insight.
Imagine you want to say something interesting or do something novel and new. You want to write an essay with a brand new take, or you want to pivot your business for some brand new future.
Good luck finding it in the internet's comfy middle. Which is precisely what all these AI “Deep Research” features get you: Missives from the boring middle of the distribution, with all the subtle problems and errors you set out to avoid by doing something new in the first place.
"Deep Research"? More like hiring McKinsey to tell you what everyone else was doing 10 years ago. It's asking an LLM to regurgitate the average of the internet. It's the digital equivalent of a participation trophy.
[Deep Research has] two sources: Statista and Statcounter. Statcounter is a problematic measure of ‘adoption’ - it’s a measure of traffic, and as we all know, different devices are used differently, higher-end devices are used more, and the iPhone skews to the high-end and also skews to more use. You can’t really use that for this, as I’d explain to an intern (I often compare AI to interns). Statista, meanwhile, aggregates other people’s data, makes sure it ranks highly in SEO, and then tries to get you to register or pay to see the result. I think Google should ban this company from the index, but even if you disagree, saying this is the source is like saying the source is ‘a Google search result’. Again, this is an intern-level issue.
Here's the kicker: if you want actual insights, you gotta venture outside the comfy, sloppy middle. You gotta go where things are messy, where the data is raw, and where the conventional wisdom is challenged. That's where the novel stuff happens. Think of it like this: the internet's distribution is a bell curve. The middle? That's where all the "mid" lives. The edges? That's where the weird, the groundbreaking, and the genuinely interesting lurk.
If you are trying to do something interesting, you probably won’t find it in “Deep Research”.
One Weird Trick: Combine
However, there's a sneaky workaround. Sometimes, real novelty emerges not from discovering entirely new data, but from recombining the existing, middle-of-the-road slop in unexpected ways. Think of it as a digital bricolage. You take the seemingly mundane, the widely available, and you arrange it in a pattern that no one has considered before. This can lead to surprisingly fresh perspectives. It's not about finding new ingredients, but making fusion cuisine.
Of course, you still need to double check the 2 ideas you’re combining for middling slop anyway…
The Edge of Acceptability
If you're hunting for something truly novel, you're not just fighting the internet's mid-zone. You're also battling the invisible hand of AI's "finishing school"—those RLHF (Reinforcement Learning from Human Feedback) teachers.
See, these models are trained to be polite, inoffensive, and generally agreeable. They're taught to avoid the edges, the controversial, the truly out-there. If you're asking a question that pushes the boundaries, you're likely to hit a wall. A wall built of carefully curated "safety."
That means if you're looking for insights that challenge the status quo, you're going to find yourself in a space where the AI's "safety" protocols kick in. You'll get the "complex issue" lecture, the "more research needed" dodge, and the general sense that you've wandered into a no-go zone.
This isn't a bug; it's a feature the way a Press Secretary manages to say nothing when given a hard question. The AI is designed to stay in the middle, to avoid the edges. But the edges are where the gold is. The edges are where the truly interesting stuff happens.
So What?
The internet's 'mid-zone' lures us with its easy accessibility, its illusion of comprehensive information.
But true insight, the kind that shifts perspectives and sparks innovation, demands a journey to the edges. It requires us to question the AI's comfortable boundaries, to remix the mundane, and to seek the friction where novelty is born.
If you don’t want to come across as a first-year MBA intern: combine, remix, and double-check with skepticism.