What a few Cybertrucks taught me about statistics, AI and logical thinking

During a road trip through California, my children started counting Cybertrucks. That seemingly simple observation shows why good thinking is more important than quick conclusions.

During a road trip through California, my children started counting Cybertrucks. That seemingly simple observation shows why good thinking is more important than quick conclusions.

A striking count

During our road trip through California, my children started counting Tesla Cybertrucks, not because there was a serious research design behind it, but simply because it’s such a striking car that you can hardly miss it when one drives past. In San Francisco and the Bay Area, we saw about ten; around Santa Barbara and Los Angeles, about five; but near Lake Tahoe, suddenly dozens. That difference stood out, and that's precisely where the interesting part began for me: not the counting itself, but the question of what such a seemingly coincidental observation might mean.

First, doubt

Of course, this isn't a representative sample. We didn't spend the same amount of time everywhere, didn't visit randomly selected areas, and didn't systematically track how many cars drove past in total. So, it could very well be a coincidence. Perhaps near Lake Tahoe, we simply happened to be in places where many Cybertrucks drive; perhaps purchasing power and second homes play a role; or perhaps a large pickup truck is simply more practical in that environment than in an urban area. All these explanations are plausible, and that’s precisely why it’s interesting not to immediately choose one narrative.

How statistics begin

For me, statistical thinking doesn't start with a spreadsheet, a regression model, or a significance test, but with an observation that forces you to ask better questions. You see something remarkable, formulate an initial hypothesis, and then try not to prove it, but rather to disprove it. What other variables could explain the same pattern? Is there a selection bias? Are you looking at a specific group of people? Is there a third factor behind the correlation? Or are you simply seeing noise and attributing meaning to it retrospectively?

More than just a car

With the Cybertruck, this becomes extra interesting because, culturally, the model is much more than just an electric car. For years, electric driving in the United States was strongly associated with progressive, urban, and predominantly Democratic consumers, while pickup trucks had a much stronger cultural connection with more conservative and Republican America. The Cybertruck strikingly brings these two worlds together: electric technology from Silicon Valley, packaged in an extremely distinctive American truck.

Tesla changed along the way

At the same time, the meaning of Tesla has also changed. Due to Elon Musk's increasingly visible political positioning, the brand has come to represent something different for various groups. Republican voters have become more positive about Musk and Tesla, while among a segment of the Democratic base, the brand has lost appeal. As a result, the Cybertruck becomes interesting as a potential proxy for a broader cultural shift: electric driving is no longer automatically linked to a single political identity.

A proxy

Of course, that doesn’t mean someone driving a Cybertruck is automatically a Republican, just as a house price doesn't precisely tell you someone's income. But proxies work precisely because they indirectly reveal something that is more difficult to measure directly. For example, search behaviour can indicate consumer confidence, job growth can reflect economic expectations, and house prices can show the socio-economic development of a neighbourhood. A good proxy is never reality itself, but it can be a signal of an underlying movement.

From signal to research

That's also why such an innocent count becomes interesting to me. Not because dozens of Cybertrucks around Lake Tahoe prove that the political market for electric cars is shifting, but because the observation evokes a hypothesis that could then be investigated with better data. You could look at regional sales figures, political preference per county, income, vehicle type, urbanity, second home ownership, and other characteristics to see which explanation holds up once you control for alternatives.

Fewer bad explanations

That's precisely where the charm of statistics lies for me. Not in producing one definitive answer, but in systematically narrowing the scope for poor explanations. A good analysis not only shows you what is likely true, but also why other explanations are less probable. This requires discipline, because the first explanation often feels the most attractive, especially if it neatly aligns with the narrative you already had in your head.

AI makes this more important

With AI, this only becomes more important. AI can quickly find patterns, combine datasets, and formulate plausible explanations, but precisely because these explanations can sound so convincing, the risk increases that we confuse a good story with a good analysis. A correlation is not yet causation, a pattern is not yet a mechanism, and a prediction is not yet an explanation. AI doesn't change these basic statistical rules; it mainly highlights how important they are.

Learning to think better

That's why I ultimately see statistics and AI primarily as tools for thinking better. They don't automatically produce the truth, but help us formulate sharper hypotheses, seek alternative explanations, and systematically test which interpretation is most logical. Sometimes it starts with millions of data points in a model, and sometimes it just starts with two children in the back seat suddenly saying: “Dad, there’s another Cybertruck.”

Maybe it's a coincidence. Maybe not. And it’s precisely between those two possibilities that analysis begins.

Job van den Berg is an AI keynote speaker, tech entrepreneur, and author of five books on AI. He deploys agents weekly and gives 150+ keynotes per year on AI agents and agentic commerce.

After half a year back in Silicon Valley, one thing stands out: the conversation is shifting from 'what can AI do?' to 'how do we truly build on it?'. Billboards about safe AI, control, and infrastructure indicate a new theme.

Hallucinations can be controlled. Much harder is reasoning that sounds logical, is partly supported by research, yet goes further than the evidence allows. On generational differences, statistics, and methodological caution as an AI skill.

The promise of AI agents is zero friction, ease first. But it was precisely in that friction that the contact that forges our social capital always arose. An essay on connection, technology, and the default setting.

Answers are more accessible than ever. The real skill will not be in formulating the question, but in opening the hatch under the bonnet.