The Truth Wager: would you bet against the God of the singularity?

Does hype even work? A team led by Petroc Sumner and Christopher Chambers looked at 462 press releases about health research from 20 UK universities. There was "little evidence that exaggeration in press releases increased the uptake of news" (Sumner and colleagues, 2014). The same group repeated it on new press releases and again found "no evidence for higher news uptake for exaggerated press releases" (Bratton and colleagues, 2019). In separate experiments, adding a caveat didn't make readers any less interested (Bott and colleagues, 2019).

That's one field, press coverage rather than your feed, and a record of what happened, not a controlled test. Sumner's team say they can't know what the same release would have got without the exaggeration. Still, the trade people assume, a little truth for a lot of attention, may not even be on offer.

I'd already decided not to make it. I call the reason The Truth Wager, or Watters' Wager if you like a bet with a name on it.

We are not the story we tell

I wrote my book, The Way Within, to help people see who they really are. The essence is that we are not the story we tell about ourselves, and trying to live as if we are doesn't end well. When you see who you really are, being truthful follows. You can't fake it till you make it, unless what you want is to be a fake, with everything you build built on sand.

Aristotle has a version of this. The truthful man, he says, "inclines rather to understate the truth; for this seems in better taste because exaggerations are wearisome". And the man who claims more than he has "for money, or the things that lead to money, is an uglier character" (Aristotle, Nicomachean Ethics IV.7, Ross's translation).

If it needs hype, it isn't that good

I've made a commitment to be my best self and live my best life, and hype doesn't help with either. It's weak. If I have to hype something, I lack the craft to make people see what's interesting about it, and the answer is to work on the craft. If people still don't get it, either my writing isn't good enough yet, and better writing comes from building that muscle, or I'm spending my time on something people aren't interested in. Both are worth knowing.

So here's a rough rule: if something's being hyped, it probably isn't that good. The fair objection is that good work goes unseen without promotion. True, but promotion isn't hype. Putting good work in front of the right people is craft.

That doesn't mean dull. I'll use everything that's known about psychology and persuasion to write a title people want to read. What I won't do is exaggerate, inflate, take out of context or misrepresent. Craft, yes. Distortion, no.

Distortion is usually small. In my piece on the $5 trillion AI roll-up pitch, McKinsey's summary said "more than one million firms are viable candidates for sale, representing up to $5 trillion in enterprise value" (McKinsey, 2026). A pitch on X turned it into "~$5 trillion worth of American businesses will change hands by 2035, according to McKinsey" (Isenberg, 2026). Nobody invented a number, and McKinsey drops "up to" itself further down its own page. What the pitch added was certainty: "viable candidates for sale" became "will change hands". The honest version can still pull. My title for that piece was a question: "Is the $5 trillion AI roll-up opportunity real?"

Before you post: craft or distortion?

Start: you've written a title, hook or claim, about someone's research or your own work, and you want more people to read it.

Put it next to its source: the study, the report, or your own numbers. Does it say more than the source says? That's exaggerating.

Has a qualifier fallen off: "up to", "may", "in this sample", "could"? That's inflating.

Read the sentences either side of the source's words. Does the meaning change? That's taking it out of context.

Read the source's own one-line summary, such as an abstract's conclusion. Would its authors say your version isn't what they meant? That's misrepresenting.

Pass rule: any yes, cut the claim back to the source's size. Four noes, and every bit of craft you like goes on how it's said.

Done when: the claim is no bigger than its source and you'd be happy for the source to read it.

The rule: make the truth compelling; never make it bigger.

The wager

Blaise Pascal's wager was about God. "If you gain, you gain all; if you lose, you lose nothing" (Pascal, Pensées §233, W. F. Trotter's translation).

Mine is about the record. What you publish now is kept: pages are archived, posts stay searchable, and text on the open web is used to train AI models. Maybe, probably, AI will one day be asked to judge what each of us said: call it the God of the singularity. I can't know that, and the wager doesn't need it. It only needs the record to be kept, which it already is, and someone, some day, to be able to check it.

The Truth Wager.

If I tell the truth and nobody ever checks, I've lived as someone who told the truth and did work that has value. If I tell the truth and somebody checks, my record stands. If I hype and nobody checks, I may profit, but what's it worth? If I hype and I'm found out, the reputation goes, and reputations, like china, once broken are gone.

Pascal had already written the first box. "Now, what harm will befall you in taking this side? You will be faithful, humble, grateful, generous, a sincere friend, truthful. Certainly you will not have those poisonous pleasures, glory and luxury" (Pascal, Pensées §233, Trotter). Glory is exactly what hype is selling.

Two places where mine differs from his. First, I can't promise the truth is free. In a study of rumours on Twitter between 2006 and 2017, false news reached more people than the truth and spread faster (Vosoughi and colleagues, 2018). Those were different stories, so it doesn't measure what telling one story straight costs you. But it stops me claiming Pascal's "lose nothing": this is a bet with a possible small cost against a large, lasting one.

Second, Pascal's critics asked which god. Diderot: "An Imam could reason just as well this way" (Hájek, Stanford Encyclopedia of Philosophy). The same question applies here: which judge? Some do reward hype: a feed that pays for outrage, a crowd that wants its side to win. The judges I care about are different. A client deciding whether to hire me. A reader deciding whether to come back. An AI asked whether my work can be relied on. Each of them has a reason to ask the same thing: is it true?

Honest mistakes and lies leave different marks

Over time, a record can show more than who told the truth. It can show who passed on something false by accident, and who kept doing it on purpose.

The first kind is common and forgivable. I've done it. This month I repeated a story about rioters stripping an old couple in France. The photos were AI fakes. I said so in public, and told the friend I'd been wrong. In Pennycook and colleagues' studies, most people said sharing only accurate news mattered to them. Yet whether a headline was true made little difference to whether they'd share it, often because their attention was on something other than accuracy (Pennycook and colleagues, 2021).

The second kind, distortion for money or status, is evidence of something else: that you can't be trusted. I think owning a mistake and correcting it shows the opposite: maturity, integrity and humility.

In a lab trust game with 262 participants, almost all aged 19 to 22, trust damaged by untrustworthy moves came back with a run of good ones. Add deception and, in the authors' model, trust "never fully recovers". That held even after a promise, an apology and a run of good moves (Schweitzer, Hershey and Bradlow, 2006). The game lasted seven rounds, so "never" is a projection. An apology on its own made no significant difference. It's one lab study, with young adults, comparing bad moves with and without a lie, not mistakes. But it puts the line where I put it: at the lie.

Azoulay, Bonatti and Krieger followed 376 US medical school faculty who had a paper retracted, against similar scientists who hadn't. Citations to their earlier work on other topics fell 17.6% where misconduct was alleged, fraud and plagiarism included. Where it wasn't, a group that takes in negligence as well as honest mistakes, they fell 8.2% (Azoulay, Bonatti and Krieger, NBER working paper, 2017). An honest mistake still cost something. The estimate for misconduct was about twice as big.

Without a lie and with one. Sources: Schweitzer, Hershey and Bradlow, 2006; Azoulay, Bonatti and Krieger, 2017.

So the china is the lie, not the mistake. That's also the wager's honest limit: not every bit of hype is a lie. Drop an "up to" through carelessness, own it and fix it, and you're with the honest mistakes, which cost less. The china is for distortion you knew was distortion and published anyway, for the money or the status. From outside, the two can look alike, so run the check.

"The liars are winning"

That's the strongest objection, and it has evidence. Five survey experiments with over 15,000 American adults tested what happens when politicians falsely call real scandal stories fake news or deepfakes. The politicians gained support, more than by staying silent or apologising (Schiff, Schiff and Bueno, 2024). Lying about being lied about paid.

Look at where it stopped paying, though. Against video the trick was "largely ineffective": no gain at all in one study, though one version of the claim still worked in a later one. And in the one study that tested it, a fact-check rejecting the politician's claim took the gain away (Schiff, Schiff and Bueno, 2024). When the claim was checked, the lie stopped paying.

My answer is that the liars can win if you're playing one hand, or a short game. I'm not, and I can't promise they lose the long one. The wager doesn't need them to.

If no judge ever comes, I've lost some reach and told the truth. If one does, I've a record I'm happy to have read. Make the truth compelling. Never make it bigger.

Sources

Sumner and colleagues (2014), BMJ: https://europepmc.org/article/PMC/PMC4262123
Bratton and colleagues (2019), Wellcome Open Research: https://europepmc.org/article/PMC/PMC6833989
Bott and colleagues (2019), Journal of Experimental Psychology: Applied: https://doi.org/10.1037/xap0000232
Plato, Republic, Book VII, Benjamin Jowett's translation, Project Gutenberg: https://www.gutenberg.org/ebooks/1497
Aristotle, Nicomachean Ethics, Book IV, W. D. Ross's translation, MIT Internet Classics Archive: http://classics.mit.edu/Aristotle/nicomachaen.4.iv.html
McKinsey (2026), The Great Ownership Transfer: https://www.mckinsey.com/institute-for-economic-mobility/our-insights/the-great-ownership-transfer-a-new-era-of-business-stewardship
Isenberg (2026), "$5T opportunity: AI Roll Ups", X: https://x.com/gregisenberg/status/2103927365977928019
Pascal, Pensées §233, W. F. Trotter's translation (Dent, 1910), as quoted in Hájek, "Pascal's Wager", Stanford Encyclopedia of Philosophy: https://plato.stanford.edu/entries/pascal-wager/
Vosoughi, Roy and Aral (2018), Science: https://doi.org/10.1126/science.aap9559
Pennycook and colleagues (2021), Nature: https://doi.org/10.1038/s41586-021-03344-2
Schweitzer, Hershey and Bradlow (2006), "Promises and lies: restoring violated trust": https://faculty.wharton.upenn.edu/wp-content/uploads/2014/06/Promises-and-Lies.pdf
Azoulay, Bonatti and Krieger (2017), NBER Working Paper 21146: https://www.nber.org/papers/w21146
Schiff, Schiff and Bueno (2024), American Political Science Review: https://doi.org/10.1017/S0003055423001454