Presume Radical Uncertainty (The 1997 Lens)
Treat an emerging platform as if it were 1997 for the internet: assume most of it doesn't work and you can't yet name the winners
- Difficulty
- Easy
- Time to result
- ~ongoing to results
- Steps
- 4
- Confidence
- 85%
A discipline for thinking clearly about a technology whose end-state is unknowable. Rather than projecting today's leaders and use-cases forward, you deliberately place yourself at an early analog moment — 1997 for the internet, 2000 for mobile — and accept that most of what will matter hasn't been built, most current attempts don't work, and confident predictions would mostly be wrong. The output is calibrated humility, not paralysis.
Origin
Benedict Evans's recurring framing across his writing and the deck whose '80 slides say we don't know' — grounded in how badly analysts in 1997 and 2000 predicted the internet and mobile winners.
Core principles
- 01A huge platform shift and total uncertainty about how it plays out are both true at once
- 02You can be certain the change is enormous and honest that you cannot name the winners
- 03In 1997 you'd have backed Excite or Yahoo over Google; in 2000 you'd have dismissed a search company as irrelevant to mobile
- 04Redefinition of terms (AGI, superintelligence) is a symptom of this fog — the words keep sliding to mean 'whatever just started working'
- 05Humility is the deliverable: the honest answer to most specifics is 'it depends'
How to run it
- 1
Fix the analog moment
Locate where you are on the curve of a prior platform — for AI in software, roughly 1997 of the internet: very exciting, most stuff doesn't work yet, most of what people will do hasn't been built.
- 2
Separate the certain from the unknowable
Hold two things at once: the shift is fundamental and will change everything, AND you do not know which companies, products, or timelines will win.
Pro tip Refuse unproductive precision fights like 'is it 20% bigger than the internet or 100%' — those aren't productive conversations.
- 3
Stress-test your confident predictions against the analog
For any bold call, ask what the equivalent call in 1997 or 2000 would have been and whether you'd have gotten it right. Most people would have missed almost all of it.
Watch out Beware anyone presuming today's situation simply continues — that is the specific over-confidence this lens guards against.
- 4
Default to 'it depends' on specifics
On granular questions, resist a clean answer; the honest position is that it depends, and stating that preserves accuracy.
Watch out You can't predict which things will be exposed or disrupted — personal trainers and taxi drivers looked safe until they weren't.
In the wild
A reader told Evans his semi-annual presentation was '80 slides saying we don't know.' He calls it facetious but kind of true — the honest state of knowledge about how AI plays out is genuine uncertainty.
→ An analyst deliberately publishing calibrated uncertainty rather than false precision.
Asked what you'd have gotten right about the internet in 1997, Evans notes you'd likely have backed Excite or Yahoo, and about mobile in 2000 you'd have dismissed a weird-logo search company as irrelevant.
→ Demonstrates that confident early winner-picking in prior shifts was mostly wrong, justifying humility now.
Common mistakes
Presuming today's situation simply continues
Assuming current leaders keep their lead or current economics persist is the default error; the 1997 lens exists precisely to break that projection.
Fighting over unknowable magnitudes instead of accepting uncertainty
Debating whether AI is 20% or 100% bigger than the internet is unproductive precision that substitutes for admitting you don't know how it will work.
Is it for you?
Best for
Analysts, founders, and leaders making strategy under a fast-moving platform shift who need to avoid false confidence
Not ideal for
Decisions that demand a concrete near-term commitment where 'it depends' is not actionable enough on its own
From the transcript
“if you're going to make the internet comparison it's like we're in 1997. Like it's very exciting. Most stuff kind of doesn't work yet.”
“imagine having this conversation about the internet in 1997 like what would you have got right”
“we should presume we don't know”
“you can't predict which things are going to be exposed”
“But you have to have that humility.”
From the episode
A rational conversation on where AI is actually going
Benedict Evans