The Window Clock Is Shrinking

A few months ago I was talking with a technology law firm about how fast a competitor could duplicate a piece of software once they’d seen it working. Their estimate: about six months. These are smart people who work on technology deals all day. The real answer, for the thing we were discussing, was about six weeks. And nowadays, with current tooling, I think that timeline could be as short as two or three weeks.

They weren’t being careless. Their intuition was calibrated in an earlier era, and it was correct for that era. The problem is that the era changed and the intuitions didn’t. I’ve been watching this same miscalibration show up everywhere: in legal advice, in investor diligence habits, in hiring timelines, in how founders plan their own launches.

Here’s the thing I think we should be talking about a lot more than we are. Adoption windows for new technologies that become standards have compressed roughly 100x in fifty years. I don’t mean how long it takes to invent something, and I don’t mean how long a given product takes to succeed. I mean how long it takes a genuinely new technology to go from public debut to the thing everybody has standardized on. That number used to be measured in decades. It is now sometimes measured in days, and essentially every professional habit we have was built for the old number.

So here’s a measuring instrument I’ve started using instead. I call it the window clock.

The compression is real and it’s measurable

Call the time from a technology’s public debut to industry-wide standardization its adoption window. Here’s what that window has done over the last fifty years, using five well-documented cases:

Sixteen years, then six, then four, then months, then a decisive step measured in days. That’s roughly a 100x compression over one working lifetime, and 30x of it happened in the last decade alone. The trend line hasn’t flattened.

Every institutional habit around technology was calibrated when the window was four years. Six-week diligence periods. Quarterly planning. “Let’s circle back after the holidays.” The habits are still here. The window they were built for is gone.

W is not the same everywhere

I’ve worked in IT, media technology, biotech, and AI. The windows in those fields are nothing alike, and the difference isn’t the era, it’s what blocks technology adoption.

The pattern: a sector’s window is set by its slowest mandatory gate. Therapeutics has clinical trials and regulators, so no amount of software speed compresses it below a decade. Consumer media has hardware replacement cycles and licensing negotiations. Enterprise IT has procurement and migration risk. AI tooling has none of those. A developer can adopt a new standard in an afternoon, so the window collapses to the speed of attention.

This is why cross-sector intuition transfers so badly. Someone whose instincts formed in biotech will find AI timelines reckless. Someone from AI tooling will find biotech incomprehensibly slow. Both are correctly calibrated to their own gates. The mistake is carrying one sector’s clock into another, and the more successful you were in the first sector, the more confident and wrong the transfer tends to be.

It also means a sector’s window can shift when a gate is removed. If a regulatory pathway opens, or a hardware dependency disappears, or distribution moves from procurement cycles to instant download, that field’s window can shorten by an order of magnitude while everyone’s habits stay where they were.

The fix: stop measuring in calendar time

The instrument is simple. For whatever technology you’re dealing with, estimate its window W. Then express every schedule quantity as a percentage of W instead of in days or weeks.

For a market with a 90-day window:

This is the conversion that makes eras comparable. A week of delay isn’t a week. On a fast-window technology it’s eight percent of the period in which the outcome gets decided. Nobody in 2015 would have defended taking eighteen months to clear a check. That’s what a six-week process is now, in window terms.

This framing doesn’t require anyone to panic or move recklessly. If you’re an investor, I’m not asking you to be faster than you’ve ever been. I’m asking you to spend the same fraction of the window on your process that you always did. Your habits were four percent of the window when you formed them. Keep them at four percent.

Two clocks, not one

Not everything compresses. Sorting activities onto the right clock is most of the skill:

Window-clock items are denominated in %W and compress or die: shipping the reference implementation, naming the category, closing committed money, filing the paperwork that protects the work.

Wall-clock items run on calendar time no matter what era it is: building trust, enterprise procurement, human relationships. These were never fast. They will never be fast.

The operating rule: never let a wall-clock dependency sit on the window-clock critical path. Do the trust-building early, so it’s already banked when the window opens.

This compression also creates a new variety of the classic build-or-buy decision. When the window was four years, you could build your wall-clock assets inside it: grow the relationships, learn the market, mature the processes. In a 90-day window, a single wall-clock item can cost more than the entire window. So the question becomes: do you have the time to build those relationships and processes at wall-clock speed, or do you buy them? Buying doesn’t just mean acquisitions. It means drawing on relationships you’ve spent years banking. It means hiring people who already carry the trust, the domain knowledge, and the access you’d otherwise spend a decade earning. The window is too short to grow those assets from scratch, but it’s not too short to bring in people who already have them.

Everything you complete before the window opens costs zero percent of the window. The window starts at public salience, not at conception. Work done quietly beforehand is free in window currency. If you’ve ever wondered why some companies seem to materialize fully formed the week their category gets hot, this is why. They loaded the window before it opened.

Who starts the clock

Here’s the part that matters most if you’re the one holding something new. The clock doesn’t start itself. If you’re introducing the category, or revealing a fundamentally better way to build things, the window opens when you say it does. Unless someone else opens it for you.

That turns release timing into an investment decision. Before the clock starts, everything you accumulate appreciates. Every tool you finish, every relationship you bank, every process you mature will be worth more inside the window than it cost to build outside it, because inside the window everything competes for the same ninety days of attention and capacity. Waiting is not idleness. It’s compounding. The discount rate on that compounding is the risk that someone else starts the clock while you wait, at a moment you didn’t choose and in a position you didn’t pick.

Which is why letting someone else start the clock can be the winning move, but only under one condition: you are the best-prepared party in the room when it happens. Docker started the container clock in 2013. Google had spent the previous decade quietly running containers at scale internally, and when it released Kubernetes into Docker’s window, the standard went to the player with the deepest stores, not the one who fired the gun. The starter gets the attention. The best-stocked player gets the market.

The clock-starting process is effectively a game of chicken. Every prepared party would rather keep accumulating, and every one of them knows that whoever starts the clock commits the whole field to the sprint at once. Hold too long and someone starts it on you. Start too early and you’ve spent your accumulation advantage to fire a gun anyone could have fired. Which player wins depends on the prize. Attention and naming go to the starter: OpenAI started the consumer AI clock with ChatGPT while Google sat on comparable capability, and the mindshare never came back. Standards and markets go to depth: that’s the Kubernetes story above. Theoretically the strongest position is the one where you’d win either race. There’s a lot you can’t see from inside the game, though, so treat that as the aim rather than a guarantee. If you can start the clock yourself, on your own timing, that should be the best way to win.

Hold the clock while you accumulate, but never hold it so long that you get out-accumulated, because the moment someone with deeper reserves can start the clock, your patience has been banking value for them. The whole skill reduces to one honest question, asked continuously: am I still the best-prepared party in the room? Wait while the answer is yes and your reserves are growing faster than anyone else’s. Start the clock the day that stops being true.

Caveats, because every instrument has them

W is technology-specific. Estimate it per case, from named precedents, with a range. Some technologies still have multi-year windows. And percentage framing can turn into urgency theater if you let it; the discipline is anchoring W in documented cases like the ones above, not in adrenaline.

But once you start denominating plans in window percentages, you can’t unsee it. The lawyer’s six-month estimate, the investor’s six-week process, the founder’s “we’ll launch next quarter”: all reasonable statements, all denominated in a currency that’s been quietly devalued 30x.

Check the exchange rate before you spend.

Sources: SQL’s history from Codd’s 1970 paper to the 1986 ANSI standard, HTML 4.0 at the W3C, 1997, container history at TechTarget, Kubernetes history, A Year of MCP, MCP one-year anniversary, SiliconANGLE on the Agent Skills open standard, Agent Skills interoperability across 32 tools, FDA approves the first CRISPR gene therapies, translational science timelines from technology initiation to FDA approval, history of streaming media, DVR timeline 1999 to 2026.