Inside the den

Trust: Offer It, Build It, Keep It

Information, computation, and intelligence are abundant now. Trust is the scarcity of this era — which makes it the opportunity.

Start Speaking Free Structural Positioning

By Greg Gregory, Founder, FluentFred

Every era's giant invention took something scarce and made it abundant. The internet made information abundant. Cloud computing made computation abundant. AI is making intelligence abundant. And with each wave, the money and the meaning moved to whatever remained scarce.

Look around at what's left. Attention is scarce. Genuine connection is scarce. But above all, trust is scarce — and getting scarcer. In a world where anything can be generated, faked, or spun, the businesses that win the next era won't be the ones that shout the loudest. They'll be the ones people believe.

Bitcoin understood this before most. Strip away the speculation and its core product is trust itself — trust that the coin is scarce and the ledger is honest, without needing to trust any person or institution. It didn't eliminate trust; it relocated it, from institutions to mathematics.

Every major invention is secretly a machine for relocating trust.

The question for any founder is: where does trust migrate next, and are you building the rails for it?

I've spent the last two years building FluentFred, an AI language tutor. On paper it's a language learning app. In practice, it's a trust product wearing a language app's clothes. And the lessons I've learned about trust apply to nearly anything you might build or sell. They come down to three verbs: offer it, build it, keep it.

Verb one

Offer it

Trust begins with a promise — but not the kind most companies make.

There are two ways to promise. The first is the virtue promise: we're trustworthy, we're ethical, don't be evil. These fail, and worse, they write the headline for your worst day. When a company that promised not to be evil stumbles, the promise itself becomes the stick it's beaten with. Never make a trust claim whose violation is more memorable than the claim.

The second is the outcome promise: a specific, personal result, underwritten by a specific, accountable human. This is the one that works.

The masterclass came from an unlikely place: a suit retailer. For decades, Men's Wearhouse founder George Zimmer looked into the camera and made men a promise about how they'd feel in his clothes — and he guaranteed it, personally. The genius wasn't the wording. It was that the promise spoke to the insecurity, not the product. Men weren't avoiding suits; they were avoiding being seen in them. Zimmer named the fear and put his own name against dissolving it.

Language learning has the same hidden wound. Learners don't quit because grammar gets hard. They quit because they don't like the way they sound. Speaking is the most feared skill by a wide margin — the adult hears their own clumsy accent, feels foolish, and goes quiet. Then the cruel loop closes: embarrassed to speak, so they don't speak, so they don't improve, so they're more embarrassed.

So FluentFred's promise is aimed at the wound, not the wares:

You're going to like the way you speak.
— Greg Gregory, Founder

Signed by me, the founder, because people are wary of trusting a machine — but they'll trust a person who vouches for the machine. A named human staking his reputation on an outcome is the oldest trust technology we have.

And a promise needs teeth. Ours is the simplest mechanism there is: if you want a refund, you let us know. No forms, no proof of completed lessons, no windows in fine print, no retention specialist talking you out of it. Every hoop a company puts in front of refund-seekers is a confession that they expect people to want out. A guarantee you can invoke with one message says the opposite: we expect you to stay, because it works.

That's the offer. A specific outcome, aimed at the real fear, signed by a real person, backed by a mechanism a customer can actually use.

Verb two

Build it

Here's the paradox at the heart of trust: it is never built by talking about it.

The moment a website says "trust us" and plasters shields and badges everywhere, the visitor's guard goes up — because that's what untrustworthy operators say.

Think about the companies you actually trust. You probably trust your email provider, your search engine, your login screen — not because they asked you to, but because they've done exactly what they said they would, every single day, for twenty years. Trust isn't earned by being lovable. It's earned by being reliable to the point of boring. People trust what never surprises them badly.

That principle translates into concrete design decisions.

Show the method, not the magic. Untrustworthy AI products hide behind "our revolutionary AI." Trustworthy ones open the bonnet. Raw AI, it turns out, is a poor teacher — its instinct is to be maximally helpful, but good teaching is almost the opposite. A real tutor deliberately withholds. It lets you struggle for exactly the right number of seconds before hinting. It ignores your past-tense errors today because today you're only ready for the present tense. It sequences, recycles, and reviews at expanding intervals. None of that lives in the model; it has to be built on top of it, and building it is where my two years went. That's why the whole climb is written down in public, level by level, in The Bridge to Fluency. Showing the work is the credibility. Experts showing their working read as trustworthy; wizards asking for faith read as salesmen.

Receipts over promises. The claim that builds trust isn't "become fluent fast" — every liar in the industry says that. It's evidence: a learner's week-one recording next to their month-three recording, in their own voice. A streak says trust me, this is working. A recording says see for yourself. Wherever possible, convert your promise into something the customer can verify personally.

Prove you know them. People don't trust software in general; they trust things that demonstrably pay attention to them specifically. The moment a tutor says "you made this same mistake on Tuesday — notice the pattern?" something flips. It stops being an app and becomes my tutor. Human trust works identically: we trust those who show they've been listening.

Remove the judgment. The deepest structural advantage of an AI tutor is not that it's cheap — it's that it's shameless. No face to wince, no classmates to smirk, infinite patience for the same butchered sentence attempted twelve times. The learner's core fear — being heard sounding bad — has no oxygen. Whatever your product is, ask what fear stands between your customer and the behaviour that would make them succeed, and design that fear out of the room.

Let the product admit limits. A tutor that occasionally says "that lesson didn't land, let's try it differently" is more trusted than one that's relentlessly positive. Small, candid imperfections are precisely what make the big claim believable. Flawlessness reads as marketing; candour reads as competence.

Experts showing their working read as trustworthy. Wizards asking for faith read as salesmen.

Do all of this and the tone takes care of itself: calm, specific, a little understated, allergic to hype. In a market screaming about revolutionary AI, the quiet site that shows its method and its receipts stands out the way a plain white search page once did against the cluttered portals of the nineties.

Verb three

Keep it

Now the hard part, and the part most companies get wrong: trust is rented, never owned.

The clearest cautionary tale in my industry is the learner with a 500-day streak who quietly realises they still can't hold a conversation. That's not churn from boredom — it's churn from betrayal. They trusted the product to get them to the finish line, kept their side of the bargain daily, and the result never came. When they leave, they don't leave angry. They leave quietly, and they don't come back at the price their trust was first bought.

The moment a customer suspects you're optimising for their subscription rather than their outcome — one dark pattern, one guilt-trip notification, one lesson that feels like padding — the spell breaks. And here's the uncomfortable truth: engagement-metric-driven companies structurally cannot stop doing these things. Their business models won't let them. That constraint is the whole argument of Structural Positioning: an incumbent can see a gap perfectly well and still be forbidden by its own economics from naming it. Which means the discipline of refusing these things is itself a moat. Every manipulation your competitors rely on is a trust deposit you collect just by abstaining.

Keeping trust also means being ruthlessly honest with yourself about whether you deserve it. In our pilot, the feedback has been glowing — and glowing feedback is precisely the signal a founder should distrust. Pilot users are friendly by nature. Humans are wired not to tell a creator that his two-year baby has problems. The comments that decide a product's fate are the ones people don't say out loud; they just quietly stop opening the app.

So we hold ourselves to behavioural evidence, not sentiment: unprompted usage in week four, whether sessions are deepening or thinning, and the golden number — how many users voluntarily told a friend. Real trust converts into unprompted return visits and unprompted referrals. Praise converts into nothing. And when we ask for feedback, we don't ask "do you like it?" We ask "what almost made you stop?" Questions that presume a flaw give people permission to reveal one.

The standard is symmetry: we want users to trust the tutor because it provably works, not because it feels nice — so we must trust our own product for the same reason, held to the same evidence we'd show an investor.

Companies that keep trust behave, every day, as if they could lose it tomorrow. Because they can.

The whole loop

Trust as an architecture

Put the three together and trust stops being a value on a poster.

Offer it — a specific outcome promise, aimed at your customer's real fear, signed by an accountable human, backed by a guarantee with no hoops.

Build it — through behaviour, not claims. Show the method. Provide receipts. Prove you're paying attention. Design the fear out of the room. Admit limits.

Keep it — by refusing the manipulations your competitors depend on, distrusting your own applause, and measuring what people do rather than what they say.

Information, computation, and intelligence are abundant now. Trust is the scarcity of this era — which means it's also the opportunity. Whoever figures out where trust migrates next, and builds the rails for it, owns what comes after.

I'm betting it can migrate to a tutor with no face, no judgment, and a founder's name on the promise.

You're going to like the way you speak.

Don't take our word for it. Take a lesson.

An essay about receipts should end with an offer to produce one. Start speaking free, and judge the promise against your own voice.

Start Speaking Free Structural Positioning