Like an orange
Everything here is my opinion. I do not speak for your employer.
January 2009
February 2009

2009-01-01 »


It's the beginning of a new year, and that's made me think a bit about thresholds - those moments where things suddenly change from one way of being to another.

If you're starting a company, the route to massive overwhelming success (as opposed to normal success, which is easier) is to correctly predict and bet your product on one of those threshold transitions. Before the transition, your product was impossible, so of course there are no competitors; after the transition, your product is critical, so you'll sell a lot.

Unfortunately, there are different kinds of transitions. Sooner or later, people will get completely sick of advertising, and advertising-based business models will crash... but we don't know when. It might be next month, or it might be in fifty years. If you bet on the death of advertising, you're most likely going to lose.

Other thresholds are based on surprise scientific discoveries; for example, someone discovers a new super-high-density chemical for making batteries, or discovers the secret of nuclear fusion, or cures cancer, or whatever. Maybe the experts in a particular field can make some kind of guess at when these will happen, but it's tough. If you don't time it within two or three years at worst, your company will be dead - or obsolete - by the time the transition comes.

But some kinds of transitions are easier to predict: the ones that follow something like Moore's law. In the book The Innovator's Dilemma, Clayton Christensen discusses several of these situations, from disk drives to printers to hydraulic cranes. It's like magic; you can graph the progress of a new technology, and at the point where its capability exceeds the capability of an older one, suddenly the whole world is different.

Here are a few of those upcoming transitions. I won't try to tell you when they'll come, but perhaps they'll give you some ideas. For context, I'll include a few that have already happened.


Handheld storage becomes large enough to store your whole music collection. (This is really what put the early iPods over the top compared to the silly 256MB players around at the time.)

Laptop hard drives become larger than anybody needs. (Desktop drives already have.)

Handheld storage becomes larger than anybody needs.

Solid state (flash) disks get so popular that optimizing operating systems for rotational latency becomes irrelevant.

Solid state disks get big enough to store most databases, so optimized high-end database engines based on disk latency become irrelevant.


It becomes cheaper to buy a new laptop than to replace the video card in a desktop PC. (I don't know how this will happen. Economies of scale as fewer and fewer people buy desktop-size cards?)

Electronic displays become easily readable in sunlight. (Supposedly the Amazon Kindle has this, but it's only black and white?)

Electronic displays become clearer (contrast, DPI, colour accuracy) than paper.

Electronic displays get about as cheap and reliable as other materials - for constructing interior walls.


Computers get so fast that you can't tell the difference in speed between dynamic and static languages. (We're right on the edge of this.)

Programming (and automated testing) becomes so easy that it's almost always easier to rewrite code for a new platform than maintain it on the old one.

Virtualization can run any DOS application at its original native speed or better. (Done: DOSBox.)

Virtualization can run any Windows 95/98/ME application at its original native speed or better. (Almost done? It seems graphics are still a problem.)

Virtualization can run any Windows NT/2000/XP application at its original native speed or better. (Not yet.)

Windows Vista actually runs on normal computers at a speed that makes it more pleasant than Windows XP. (That happened this year! I saw a sub-$1000 PC with 6 gigs of RAM and Vista ran great on it.)

Microsoft .Net becomes fast and ubiquitous enough that people stop making native Windows apps. (Slowly but surely.)

Computers become fast enough that all native Windows apps ever created will run fine under virtualization, so you can drop Win32 entirely.


The Internet becomes sufficiently fast and widespread that it's cheaper to collaborate on software across the world than to write your own separate implementation. (This is what allowed Open Source.)

The Internet becomes sufficiently fast, and disks get sufficiently large, that giving the entire development history of a project to every developer is a good idea. (We're on the edge of this: distributed version control is catching on.)

It becomes sufficiently cheap to develop and distribute software that you no longer need significant financing for most projects. (That's really the Web 2.0 movement in a nutshell.)

Wireless networking becomes fast, reliable, and cheap enough to replace wired networks to the home.

Wireless devices become so easy to build that your home entertainment centre no longer has its components wired together. And the clocks will be right.

Professionally-run Internet-based services have higher uptime than the server in your office. (This is already true for the servers themselves, but often not your link to them. Then again, small business servers have notoriously low reliability and high maintenance costs.)

Professionally-hosted Internet-based desktop applicatons have higher uptime than apps running locally on your PC. (This will never happen since your PC is needed to run Internet apps. Note the asymmetry with Internet-based servers.)

Latency of an Internet-accessible server is as low as a LAN-connected one. (This will never happen, dooming various efforts that still depend on the assumption that it will.)


Batteries last so long that you no longer think about whether you're killing the battery. (Blackberries already have this; iPhones are reputedly close; laptops not at all, except maybe the old PPC-based Macs.)

Solar power saves more in electricity fees than it costs in up-front investment.

Electronics become sufficiently lightweight and low-power that you can make remote controlled flying toys using insect-like aerodynamics instead of man-made style.

Power density in batteries gets high enough to make electric cars sensible.

Thanks to computer-controlled guidance and diagnostics, cars become so safe that they become essentially uncrashable, and the (physically heavy) crash-safety features are no longer needed.


I'm CEO at Tailscale, where we make network problems disappear.

Why would you follow me on twitter? Use RSS.

apenwarr on gmail.com