A few nights ago I watched a video that put up a chart I had never seen before. A line climbs steeply, peaks, falls off a cliff, bottoms out, then rises again to a lower, steadier level.
I recognised the shape immediately, but not from any chart. I recognised it from my own account.
What I actually did
Last year everyone around me was talking about AI. I bought a small position in two theme funds — one tracking the power infrastructure that data centres need, one tracking quantum computing companies. Small. A toe in the water.
At first they went up. That felt good, and it did something more dangerous than feeling good: it made me feel like I had been right.
Then they started falling. Not gently.
And here is where I made the decision I want to write about. I looked at the falling price and thought: this is the opportunity. Every time they dropped hard, I bought more. June, July, into early August. Not on a schedule — on the dips. Each time, the same quiet reasoning: it is cheaper now, so this must be a better entry than last time.
They are down about 24 percent.
The word I was using wrong
I told myself I was dollar-cost averaging, because that is what this whole site is about and it is what I do everywhere else.
I was not. I was averaging down, and the two are close to opposites.
| Dollar-cost averaging | Averaging down | |
|---|---|---|
| When you buy | A fixed date, whatever the price | After a drop |
| Decisions made | None. It is automatic | One every time |
| What it assumes | Markets rise over decades | This is cheap now |
Dollar-cost averaging exists to remove judgement. You buy on the 15th because it is the 15th. The whole point is that you are not allowed to have an opinion.
Averaging down is the opposite. Every purchase is a fresh judgement that the current price is a bargain. It looks like discipline — you are buying when others are scared, which is what everyone says to do — but you are making a call each time, and you have to be right repeatedly.
In the same account, over the same two months, I was doing both. My S&P 500, Dow-Jones-style dividend, and Nasdaq index purchases went through on their scheduled dates in fixed amounts, with no input from me. In my August recap, the portion of my balance driven by markets rather than deposits came out +$3,200.
Same investor, same weeks, opposite results. I had one name for two different behaviours.
The shape is easy to draw afterwards. Knowing where you are standing on it, in real time, is the part nobody can do.
The curve I did not know about
The chart in that video is Gartner’s Hype Cycle. It has five stages:
- Innovation Trigger — a breakthrough gets attention, usually before there is a usable product.
- Peak of Inflated Expectations — early success stories, heavy press, money arriving faster than results.
- Trough of Disillusionment — projects fail to deliver, interest collapses.
- Slope of Enlightenment — survivors work out what the technology is genuinely good for.
- Plateau of Productivity — mainstream adoption, well below the original hype.
I want to be honest about the sequence here, because it is the entire reason this post exists. I did not know this curve existed when I was buying. I saw it for the first time last week, months after the purchases, and only then did I understand what those “opportunities” had been.
That is not a confession of unusual foolishness. It is the normal condition. The curve is perfectly legible looking backwards and almost useless looking forwards, which is exactly why my dip-buying felt so reasonable at the time. On the way down, every point looks cheap compared with the one before it.
Britain, 1845
The railway was not a scam. That is the part people skip.
In 1845 the British Parliament authorised more than 8,500 miles of new track in a single year. Petitions for new railway companies went from 199 in 1844 to 562 in 1845 to 815 in 1846. An index of railway shares had roughly doubled between 1843 and its October 1845 peak.
Then the Bank of England raised rates, the Irish famine strained public finances, and companies authorised to build could not raise the capital to finish. The index bottomed in April 1850, down 64 percent. It did not return to its 1845 level until the 1860s.
The railways got built. Britain ended up covered in track and trains still run on much of it. The technology was completely real — arguably more transformative than its promoters claimed. Shareholders who bought in 1845 still waited around fifteen years to get their money back.
Somewhere on that slide from 1845 to 1850 there were investors doing exactly what I did, buying each dip because it looked cheaper than the last one. For five years, every one of those purchases was wrong.
America, 2000
The same story with better records.
The Nasdaq Composite peaked at 5,048.62 on 10 March 2000. By October 2002 it was around 1,114 — down roughly 78 percent. It did not close above the March 2000 level again until 23 April 2015.
| Peak | Trough | Fall | Back to peak | |
|---|---|---|---|---|
| British railways | Oct 1845 | Apr 1850 | −64% | 1860s |
| Nasdaq | Mar 2000 | Oct 2002 | −78% | Apr 2015 |
The internet was real too. It did everything the 1999 pitch decks promised and more. Being right about a technology and being right about its price were two separate problems, and only one of them was solved by 2002.
I find the last column more sobering than the fourth. A 78 percent drop is a headline. Fifteen years is a decade and a half of an actual life.
Where the curve says AI is now
In 2025 Gartner placed generative AI in the Trough of Disillusionment — not as a prediction of collapse, but as a description of what they were observing: stalled enterprise projects, unclear returns, integration costs above expectations. In the same cycle, the less glamorous enabling technologies underneath it were climbing the Slope of Enlightenment.
So the analysts had already called it while I was busy buying the dips. Not that knowing would have saved me. Which brings me to the thing I actually want to say.
Knowing the curve would not have helped
It is tempting to end this by saying I should have learned about the Hype Cycle sooner. I do not think that is true.
I know the curve now. I still cannot tell you where AI sits on it today. Gartner’s trough call is a judgement, not a measurement — and the trough is not a point, it is a stretch of road that can last a year or a decade. Even with the chart in front of me, “we are near the bottom” is a guess, and it is the same guess I was making every time I bought a dip without knowing the chart existed.
The curve explains the past. It does not locate the present.
What actually protected me
Not knowledge. Arithmetic, decided in advance.
Those two theme funds sit inside a slice of my portfolio labelled “Other” at 6.7 percent of the total. My S&P 500 holdings are 31.5 percent. Bonds, dividend funds and cash make up most of the rest.
A 24 percent loss on 6.7 percent of a portfolio costs about 1.6 percent of everything. That is a number I can look at on a bad morning without doing anything drastic. If those funds had been 40 percent of the account, I would not be writing a calm blog post — I would be making decisions at the worst possible moment, which is the single thing my whole system exists to prevent.
I did not predict the curve. I did not need to, because I had already decided how much it was allowed to matter.
Where I might still be wrong
Two things worth saying against myself.
If AI climbs out over the next few years, my 6.7 percent will look timid rather than prudent, and I will have earned a fraction of what a concentrated position would have paid. Position sizing is not free. You buy the protection with upside you gave away, and I did give it away.
And “it recovered last time” is a description of survivors. Some technologies enter the trough and never come out; the ones that did are the only ones we tell stories about. I cannot rule out that these two funds are in the other group.
I am carrying both of those risks in a slice small enough that neither changes the plan. This month’s index purchases went through on their scheduled day, in their scheduled amount, while the theme funds sat there at minus 24. I did not stop the first because of the second.
If there is one thing worth taking from this, it is not the shape of the curve. It is checking which of the two things you are actually doing — because I called mine by the wrong name for two months, and the name was the part I was most confident about.
This is my personal story, not financial advice. Just one 40-something engineer keeping an honest record of his own journey.
Related reading