Five Years

My theory: 'Five years' is a particular timeline that has a secret, hidden meaning - especially in technology.

I think I first noticed it when Steve Jobs announced the iPhone - he said that Apple had a five year lead on the competition. I remember thinking, "Huh. That's an oddly specific number to put on what's secretly going on inside companies you don't really know anything about, or projects they haven't even started yet."

Steve Jobs' 'Five Years' prediction

The iPhone announcement was in January 2006 – about six months before it launched.

Google launched the first public version of the Android operating system in 2008¹. By 2010 the combination of Version 2.1/2.2 of Google's operating system (which were significant leaps forwards over the earlier versions) and a growing range of new Android devices from various manufacturers (Samsung, HTC, Motorola, etc.) meant that Android actually looked like a viable challenger. So by 2011 - five years after Jobs' announcement, the competition had the software, hardware, partnerships, consumer base, developer support and hardware logistics needed to challenge Apple's dominance in the – now completely reshaped – smartphone market.

So – looking at what happened in the space of those five years;

• First - someone had to decide that instead of making phones like 'the old ones but better', they needed to make something like the iPhone. (Which was a totally different thing to existing smartphones/feature phones at the time.)
• Someone had to figure out what a competing mobile operating system would look like; how it would work, what the hardware requirements/constraints would be. Not many businesses had an existing operating system that could be reworked. (Arguably, the biggest problem that Microsoft, RIM, Nokia etc. had was they they had an existing operating system that couldn't be 'reworked' to be an iPhone-competitor.)
• Then, handset companies could then design and build the hardware to run the software (that didn't yet exist.)
• No one company was in a position to do both, so this would involve companies working together (the "Open Handset Alliance" was set up by Google in late 2007 to get the relevant companies on board and aligned, along with the first release of the Android software.)
• Once handsets were on sale and people were actually buying them, the handset manufacturers could figure out what scale they could make and sell them, developers could understand whether it was something to invest time/money/energy into.

At that point – once any key bugs/issues have been resolved, there's a virtuous cycle; competition drives improvements in handsets, consumer adoption drives developer interest, better apps make the phones more appealing, which drives handset sales, which drives developer interest, etc. etc.

Its worth noting that Android (and Apple) wasn't a given. There was a pre-Apple smartphone industry; Blackberry were doing well, Windows Phones were useful, Symbian was a good mobile operating system. After the iPhone, it all looked different; Microsoft essentially rebooted their efforts (a massive improvement, but too little too late to effectively compete), Palm came out with WebOS (a genuinely great operating system that never got the adoption it needed to break through). I don't think Jobs could have known who was going to be the competition five years ahead of his prediction – but the fact that he still got the timeline about right was something I always found intriguing.

The thing is, making a product as innovative as the iPhone was isn't just one thing; there's the software, the hardware, and the marketing. As a company that writes the software and builds the hardware, Apple is unusual - no competitor who could handle the marketing (ie. not Palm) could do the software and hardware. Google had no experience in phone hardware, Nokia couldn't do the software, Blackberry were too tied to their keyboard to be able to build a challenger touchscreen (they later tried to make a tablet to compete with the iPad and… well, at least they tried.) Symbian would require a total overhaul to be a relevant platform. Anybody trying to catch up would be working in partnerships - and have to navigate all the meetings, politics and disagreements that come with that approach.

So – before the hardware companies could get started, someone needed to build the software. Thats not a small project. Then, someone else would need to build the actual phones - again, not a small project, and one fundamentally tied to the success of the software design. First, it had to work. That took a couple of years. Then, it had to get good - another couple of years. Then it had to get sold.

Since then, I've built out a (very) rough theory of technology timelines, AKA "how to read corporate announcements";

A Rough Theory of Tech Timelines

• "In the next 3 months": This is announcing to your would-be customers – 'this is real, its ready to roll out, and you should get ready to buy it'.

  • The messaging to customers - 'Put your existing purchasing plans on hold, because you'll regret it.'
  • For hardware: its ready to go – possibly already being manufactured – but perhaps packaging/pricing/marketing hasn't been finalised, or there are still logistics to be worked through. (For example, physically shipping goods from Asia to the US can take 4-6 weeks alone.) But on this timeframe, the 'thing in the box' is almost definitely ready.
  • For software: a bit fuzzier (due to the lack of physical logistics involved in online distribution); some sort of beta version has probably been successfully tested, but there are still some final issues to be worked out (which are expected to be minor…) or the release is part of a longer marketing timeline.

• "3-6 months": Its real, its nearly ready – but there are still some important issues to resolve.

  • Maybe a supplier for a critical part has yet to sign off. Maybe a software component isn't quite finalised yet. Maybe there's a regulatory process that needs to happen before you can go to market. But the product is real and almost ready…
  • This is probably as much of an announcement to partners and suppliers in the industry than to customers; the 'core' product is real and working, but there are still bugs to be ironed out, 'product' decisions to be finalised, regulatory approvals etc.
  • The benefit to the company is that they can get announcements out in a way that they have control over (ie. before they have to deal with supply chain leaks, incomplete/inaccurate information etc.)

• "6 months - 1 year": This is where timelines shift from "plans" to "ambitions"; where the announcement says "we will", you should read it more as "we hope…".

  • Perhaps that 3rd party supplier or software issue is looking like it will be tricky to resolve quickly – but for whatever reason, its important to get the news out sooner rather than later.
  • A big clue is that actual dates being talked about get fuzzy. Instead of dates, announcements talk vaguely about seasons. "Next summer", "Later this year", "Ready for the holiday season".
  • Perhaps the shift here is that this is "news" aimed at investors as much as it is for future customers. "Buy in now, because once this is public everyone will want a piece."

• 1-2 years: Its probably real, but there are big decisions that still need to be made – maybe the basic design still isn't locked down, or key partners aren't fully on board etc.

  • We're well into "we hope" territory once we're counting in years rather than months, and the target audience for the messaging is probably as much about a message to internal staff and stakeholders about the level of commitment to the project as it is to inform people in the outside world about something on the horizon.
  • We're also into 'teaser' territory, with something like "we look forward to telling you more later in the year" (ie. when we've figured out what it is that we're going to say, and we've got a sense of what sort of timelines we might actually commit to.)

• 3-5 years: We've got a good idea of what the end product will look like and how we plan to make money out of it, but we're not even close to the final stages yet.

  • This is where the company knows that there are still important decisions to be made that depend on other important decisions that haven't been finalised. eg. We have a concept design for hardware, but we don't know what requirements the software is going to have. Or, we have a working prototype but still need to figure out how to make the machines that will make the end product, where we're going to build them affordably, still need to agree on a final UI concept, significant infrastructure work still needed to make the proof of concept scale.
  • (This is – I would guess – how Steve Jobs was thinking about the industry's reaction to the iPhone. Sure, Samsung would have something that looked like an iPhone on the market within a year), but worked like was a very different proposition.)

• 5 years and above: Anything on this kind of timescale is essentially a speculative concept, based on anticipated developments that haven't really happened yet and might not be under the company's control.

  • In other words, "we don't really know what the business model is, or who the target customer will be, but we think this is cool."
  • This should be read as an announcement that "we want to be the first to do this"; not that we're in any meaningful way ready to do this.

Zuckerberg's 'Five Years' prediction

In 2021, when Mark Zuckerberg rebranded "Facebook, inc." as "Meta Platforms, inc." at the end of October, he was talking about the metaverse as a thing that doesn't yet exist, but would be here "in the next five to ten years". My read at the time was that this is a complex thing to build, but that in a 'five years' timeframe Meta would be building and releasing the tools to enable others to work on a new, "open" platform.

Its now 2026 – so this October will mark the five year anniversary of that prediction. It's not looking great for the metaverse; Horizon Worlds might not be shutting down the VR app just yet, but the whole idea of the Metaverse (as Meta described it) is probably looking less realistic now than it ever has. (In 2022 I said that the Metaverse is becoming the punchline of an elephant joke; in 2026, its just a joke.)²

Other than the 'Meta' rebrand, what evidence have we seen that Meta have been pushing for an actual "metaverse"? In 2021, Meta were advocating a metaverse built on "open" standards - so digital assets could move between different companies' platforms. Today; Horizon Worlds is still a walled garden. The Meta Horizon store is no more "Open" than the Apple App Store it was supposed to be different from (including the "tightly controlled" approval process before publishing and 15/30% "high taxes" on sales that goes to Meta).²

You can bring your Horizons avatar into other apps – but only within the Meta Quest platform. (So, no more 'open' than your Nintendo Wii avatars being available in different Nintendo Wii games.) There was an announcement that 3rd parties would be building VR heasets using Meta's Horizon OS – but they were 'paused' last year. Meta's 'metaverse' is just as closed as it was five years ago.

Even Bluesky and Mastadon are more 'metaverse-ish' than Meta's own Threads - you could argue that APIs to allow 3rd party integration with WhatsApp and Messenger are one 'opening up' – but that looks more like a result of the EU's DMA than steps towards a Metaverse vision.

I'm pretty sure that by now, we could have seen something at least on the way to enabling a coherent online identity owned by the individual, rather than the platforms. But then, that would enable us to build our own newsfeeds based on our own interests; instead (in the space of five years) we've gone from 100% of content from those you chose to follow to just 30%. The relationship between platform and user has shifted hard in the opposite direction.

Zuckerberg's "five years" looks less like a bad prediction than a promise that was abandoned. (Or, if you really want to be generous, derailed by an even more expensive side-quest.)

AI's 'five years' timeline

So - at a time when AI is disrupting (or threatening disruption) of pretty much anything and everything and the next few years seems to have never looked less clear (which is quite a thing to be able to say 5 years after we came out of Covid lockdowns), how useful is this framework? If AI can write its own code (faster than people), does that change the model?

Looking through the forecasts/predictions in Wikipedia's AGI Timescales page;

• Geoffery Hinton estimated in 2024 that 'systems smarter than humans' could appear within 5 to 20 years
• Jensen Huang (Nvidia's CEO) stated a 'five year' expectation in March 2024 for AI 'capable of passing any test at least as well as humans'.

Elsewhere, Demis Hassabis (CEO of Google Deepmind)

And soon, I think actually in the next maybe 5 to 10 years I think we'll have systems that are capable of not only solving a important problem or conjecture in science but coming up with it in the first place. (Source: 60 Minutes interview)

Dario Amodei (Anthropic CEO) in a blog post from October 2024;

What powerful AI (I dislike the term AGI) will look like, and when (or if) it will arrive, is a huge topic in itself. […] I think it could come as early as 2026, though there are also ways it could take much longer. But for the purposes of this essay, I’d like to put these issues aside, assume it will come reasonably soon, and focus on what happens in the 5-10 years after that.

So; assuming the "AI breakthrough" happens, and looking at the effects on a 5+ year timeline.

Sam Altman (OpenAI CEO) in a blog post;

It is possible that we will have superintelligence in a few thousand days (!); it may take longer, but I’m confident we’ll get there.

That was posted in September 23, 2024 (915 days ago, as of 2026-03-26). "A few thousand days" is an really weird way of talking about time - but to save you the maths, five years is 1,826 days. So we're still talking about the same sort of timeframe.

Yann LeCun (former head of AI at Meta, and known for being sceptical that large language models are the path to AGI/superintelligence) posted on X in October 2024;

I said that reaching Human-Level AI "will take several years if not a decade."
Sam Altman says "several thousand days" which is at least 2000 days (6 years) or perhaps 3000 days (9 years). So we're not in disagreement.
But I think the distribution has a long tail: it could take much longer than that. In AI, it almost always takes longer.
In any case, it's not going to be in the next year or two.

So – its pretty obvious that large language models are going to get bigger and better. Its less obvious exactly what that really means ('bigger' models are one thing – having the RAM to actually make them widely available at scale is another, pricing is another point), but it seems like the truly 'next level' leap to AGI/superintelligence/'powerful AI' or whatever we want to call the next generation isn't something any serious thought leaders in the space are expecting within that five year timeframe.

Elon Musk's Five Years

Obviously, this is a rough framework – not hard-and-fast rules.

For example, if there's a Wikipedia page about a specific person's predictions with that many red crosses against them, then you should probably their predictions about something like AGI arriving in 2026 with a pinch of salt.

But as with many things, I think Elon Musk is an exception that proves the rule. And as a rough mental model, I think the timeframe works pretty well.

How I'm reading 'five year' timescales

There's a famous Bill Gates quote about how 'most people overestimate what they can do in one year and underestimate what they can do in ten years' - which maybe suggests that about five years in the future is a sort of inflection point where expectations and reality come into alignment.

My take is that if there's a thing that you think is possible - you know (you think) what needs to be done, and you know what resources need to be allocated to the task (and you know that they are available to allocate – ie. not huge amounts of cheap RAM/GPUs in the next few years), then a 12 month timeline for a big project is probably feasible. Maybe it will turn out to be a bit more complicated, maybe the project will have some unexpected delays - so 18 months is probably a bit more realistic as a target.

If its a project with several stakeholders with conflicting interests to manage - thats a significantly slower-moving beast that will probably add another year to the timeline. Now you're looking at 2.5 year timeline for what sounds like a 12 month project.

But if you're talking about something that you think might be possible, what you're saying is that there are a couple of stages to the project. Before you can make 'the thing', you've got another project - you've got to build the tools to make 'the thing'. Before you can build a new generation of smartphones, you've got to build a new generation of smartphone software; before you can make the hardware, you've got to build the components; before you build the components, you've got to build the factories; before you can build a community of developers for the software, you've got to build and sell a new generation of hardware - its something thats complex; not complicated.

Before you can have a metaverse, you've got to put the tools that people will use to build the metaverse in their hands. Before you can build AGI, you've got to build a model, then build the computers and datasets that will do the training and inference that will put the intelligence into the model. Again - these are complex, not complicated problems; that all adds to the timeframe.

So now, every time I read someone talking about technology that is "five years away", I read it as "we can't build this today, but we're working on building the tools/infrastructure that we think we will need to build this imaginary thing." Its manifestations, as much as predictions.

But maybe right now, it’s different. AI tools are the tools that we'll use to build the AI tools that will build the tools – and if there's one thing you can say about these tools; they might not be good, they might not be cheap – but they are definitely fast.

We've got five years, my brain hurts a lot
Five years, that's all we've got

1. Although the Android project was underway before the iPhone, early Android was basically a BlackBerry clone. ↩

2. For what its worth, for all my metaverse enthusiasm in 2021 and early 2022, I marked the point where the thing I was excited about had diverged from 'the metaverse' in August 2022. I think there's a similarity between the way that Google 'killed' RSS by embracing it with Google Reader and then extinguishing it, and the way Meta embraced 'the Metaverse' and then killed it. But that's probably the topic for another post… ↩

3. In fairness, if you want to build your own 3D experience in Unity and side-load it onto the headset then you *can*. But while thats fun as a hobby, you'd be crazy to think about it as a business model. ↩