Letter #1, To the First Fifty

To the first fifty,

This is the first of four letters you will receive in your first year. They will arrive in September, December, March, and June. They are written in the way Warren Buffett wrote to Berkshire shareholders, plainly, on paper, once a quarter. The first letter from someone who built a thing should not be a press release. It should set the line the work is held to, in language a reader can return to in a year, and use to judge whether the line held.

You are fifty people. The cap is a real cap. When the fiftieth Founder is signed, the offer closes, and the only way in for someone else is through a Founder who leaves. The number is not a marketing device. It is the largest cohort I can know by their patterns, by what they push back on, by what they ask. Above fifty, the relationship becomes statistical. I would rather keep it personal while the product is still finding its shape.

This letter has four parts. The first is the conditions VÆN was built under, so you can judge whether the diagnosis matches your own experience. The second is what is already known to fail, because honesty about failure is the price of being trusted with private data. The third is the six strategic decisions taken since May, with the reasoning behind each. The fourth is what is not yet built, drawn from the live list of pending work that the team and I are still closing.

I do not sign this letter with my given name. The Architect is a deliberate pseudonym. The reasoning is in the Manifesto and I will not relitigate it here. The pseudonym is held by a legal entity, audited annually, and survives me if I am hit by a tram on a Tuesday. The work is what is being held to a line, not the writer.

I. The conditions

VÆN was built into three weather systems at once.

The first is the post-Covid wellness saturation. From 2021 onward, the category that should have raised a discipline raised a theatre instead. Streaks decoupled from outcomes. Ring closures decoupled from rest. Coaches decoupled from the cost of what they prescribed. A person can spend a thousand euro a year on apps that congratulate them for nothing in particular, and end the year exactly where they started. The reason this happened is structural, not malicious. Apps that retain users by making them feel productive sell better than apps that retain users by making them face themselves. The market selected for theatre. I do not blame the founders of those apps. I blame the metric they were optimizing.

The second is the AI-application proliferation of 2024 and 2025. After GPT-4 became cheap enough to wrap, every consumer category got a chatbot pasted on the front. Most of these chatbots offer the same thing: a stripe of large-language-model output, friendly in tone, expensive in dependency, identical across competitors. The cost is hidden. The personality is borrowed. The memory, where there is memory, evaporates between sessions because storing it would cost money the unit economics do not justify. The result is a marketplace where AI feels less rare every month and less useful by the same measure. I am not opposed to AI. I am opposed to AI deployed as a finish, rather than as a substrate.

The third is the quiet decline of popular Stoicism. Marcus Aurelius and Seneca went through a brief renaissance during lockdown, then were absorbed into a self-help cycle that misread them. Stoicism became, in the popular imagination, a posture of toughness. It is not that. It is a discipline of distinguishing what is in your control from what is not, and acting only on the former, with steadiness rather than force. The popular version flattens this into productivity advice for men who think feelings are weakness. The real version is harder, slower, less marketable, and more useful. VÆN takes the real version as its register, not the pop version.

These three weather systems gave VÆN its shape. We are built to be the daily intelligence layer for the hours that fall between professional sessions. We use AI as a substrate, not a feature. We borrow our register from Stoicism without confusing it with toughness. And we measure ourselves not against other wellness apps, because that comparison flatters us, but against the standard of a Stoic mirror: was the user honest with themselves, did they notice what they would otherwise have skipped, did they return.

II. The failures already known

A letter that lists only what works is a press release. I will say what is already known to be weak.

VÆN is too cold for some users. The voice is intentionally not warm in the first ninety days, because the first ninety days are observation, not change. Warmth too early is the failure mode of every wellness app that tells a user they are doing fine when they are not. I am willing to lose users to coldness in week one in order to keep the users who need a mirror, not a hug. But this is a real cost. Some Founders, in onboarding, will close the app and not return because the first Mirror landed flat. The fix is not to be warmer. The fix is to be clearer, earlier, about what the first ninety days are for. We are still drafting that copy.

VÆN is too quiet for some users. The product does not send daily streak notifications, does not buzz at 9 a.m. with a motivational quote, does not gamify the home screen. This is by design. The cost is that a user who needs an external scaffold to remember to engage may forget to engage at all. We are testing weekly anchor notifications, written in ÆON's voice, that arrive on Sunday and say only what is true. Whether they are enough remains an open question. If they are not enough, Founders will tell us.

VÆN is not a clinical product, and it will fail users who need a clinical product. This is the most important failure to name. If a user is in acute crisis, VÆN's Mirror is not a substitute for emergency services, a clinician, or another human being. The Mirror is built for the 8,700 hours of the year when no clinician is in the room. It is not built to replace the twelve to sixty hours when one is. Every onboarding flow names this. Every crisis-keyword detected in a Witness entry escalates to a country-specific hotline. But escalation logic is software, and software fails. If a user with severe mental-health needs comes to VÆN expecting a system to catch them and we miss the signal, that is a serious failure. We will publish the false-negative rate of our keyword detection at the end of year one, audited by an external party. We will publish the false-positive rate as well, because cold over-escalation is its own harm.

VÆN's data system is built for the EU and assumes EU law. Founders outside the European Economic Area will get a slower product, fewer integrations, and a privacy posture designed for the strongest legal regime, not theirs. This is a tradeoff I made on purpose. I would rather over-engineer privacy for everyone than under-engineer it for the strictest jurisdiction.

VÆN's Founder cap of fifty excludes nearly everyone who might benefit from this work. I am aware that pricing for scarcity reads, to some readers, as an excuse to charge more later. The lifetime €7,99 price for Founders is locked. The post-Founder price is €14,99. Both are publicly committed. I do not promise that price will never change for new subscribers, but I do promise that no Founder will ever pay more than €7,99 per month for the duration of their subscription. That is in the legal terms, not the marketing copy.

III. The six strategic decisions

Six decisions were taken in May and confirmed in early summer. Each is a tradeoff. Each is reversible only at significant cost, so each was worth taking slowly.

Decision one. Launch in September 2026, not in May or June. The earlier dates were within reach for a thinner product. September gives sixteen weeks of build, two hard validation gates, and an alpha cohort large enough to find the bugs that one person testing alone cannot find. The cost is four months of runway burn before the first euro of subscription revenue. I took the cost because shipping a product that breaks the first fifty users would erase the cohort I am writing this letter to.

Decision two. Six extensions, each built to depth. The Witness, the Compass, the Enforcer, the Playmaker, the Fuel, and the Forge all ship at launch, each built to a level a competitor cannot reach by adding a feature. The hardest calls were Fuel and Playmaker, where a fourteen-year database race against MyFitnessPal or Strava cannot be won in sixteen weeks. The decision there was to compete on alignment rather than logging volume. Fuel reads what you eat and links it to your energy, training, and sleep. The Forge turns what is already in your kitchen into meals that fit. The Playmaker reads recovery before it prescribes load. The depth comes from the cross-extension intelligence no specialist app can reach, because no specialist app holds the other five. Each extension keeps deepening through months three, six, and nine.

Decision three. Three voices, three roles, one brand. VÆN signs the institutional copy, the legal text, the press releases. ÆON is the customer-facing AI character that writes the Mirror and the Sunday Conversation. The Architect, that is, the writer of this letter, signs quarterly statements and the Manifesto. JÆRVIS, internal-only, runs operations for the founder and never appears in the product. The reasoning is that a single voice doing all four jobs becomes a chatbot. Three voices doing three jobs keeps each register honest. The cost is that brand-voice discipline has to be enforced mechanically, by automated checks, not by hope. The discipline is in the hook system, not in the goodwill of writers.

Decision four. Fifty Founders, lifetime price locked, slot reopens on churn. A larger cohort would have produced more cash in the first year. A smaller cohort would have produced less feedback. Fifty is the number at which I can read every Founder message in a week without falling behind. Above that, the relationship becomes asymmetric, and a Founder writes into a void. The lifetime price is the price of trust paid in advance. If a Founder leaves, the slot reopens for the next person on the waitlist, with the same lifetime terms.

Decision five. EU-wide on day one, marketing in waves. The app is available in ten EU markets from the first day. The marketing energy is concentrated in Belgium first, the Netherlands second, France and Germany after. The reasoning is that the Forge OCR has been tested only on the top five EU retailers, and rolling out to all ten markets with strong copy in week one would have produced bug reports we could not triage. Quiet markets stay quiet on purpose. The cost is slower growth in countries where energetic marketing would have produced earlier signups. I accept the cost.

Decision six. The Mirror as the ninety-day carrier artifact. Every category-defining brand has an artifact that leaves the app. Strava has the route map. Whoop has the recovery score. Notion has the shared template URL. For the first eighty-nine days, VÆN had nothing. The Mirror, an eighty-to-one-hundred-fifty-word reflective paragraph that arrives within forty-eight hours of onboarding, fills that gap. It is built to be screenshotted. It is built to be uncopyable without cross-pillar data. It is signed by ÆON, not by the user, so it does not encourage performative sharing. Whether it travels the way Strava routes travel is an empirical question that this Founder cohort will help answer.

IV. What is not yet built

The list of pending work is in front of me. Some items are minor. Some are not. I will say the honest version.

The pre-commit hook system that enforces the brand voice is shipped, but the multi-tenant audit of every privacy boundary in the product has not yet been completed end to end. We have committed to running that audit before the first paying subscriber outside the Founder cohort signs up. Until that audit clears, the product is in a measured-trust state, not a fully audited one. The Founders are subject to the same privacy boundaries as any future user. The audit is for verification, not for permission.

Fuel, Playmaker, and Forge ship real at launch and keep deepening through months three to nine: more retailers behind the Forge OCR, deeper wearable sync in the Playmaker, broader food intelligence in the Fuel. What launches is honest depth, not a stub, but the work is not finished. What deepens, and when, is named on the product page and again in the onboarding.

The Apple Watch companion, the Garmin live sync, the cellular-grade offline mode, the German UI, the additional Forge retailers, all of these are on the year-one roadmap and none of these are shipping on day one. The roadmap is public. If a delivery date slips, that slip will be named in the next quarterly letter, not buried in a changelog.

The crisis-detection logic that escalates to local hotlines is shipped, but its true false-negative rate is not yet known under production load. We will publish that rate at end of year one. If the rate is worse than projected, that will be in the letter as well.

The end-to-end test coverage of the Mirror generation pipeline is good for the happy path and incomplete for adversarial cases. We have not yet broken our own product hard enough. The Founder cohort will, by ordinary use, surface failure modes a test suite cannot generate. When a Founder hits one, we owe them a fix within a week, or an honest explanation of why the fix takes longer.

Close

If this product becomes less useful than promised, write to architect@vaen.cc. The address is read by me. We will read every message. We may not reply to every message, because the cohort is fifty and the writer is one. We will improve. Improvement is the only response that matters. A reply that does not change the product is a politeness, not a fix.

The next letter arrives in December. By then, the Mirror will have run for three months in the wild, the Fuel and Playmaker stubs will have hit their first real load, and the first wave of Founders will have either stayed or left. Whatever happened will be on the record.

Stop drifting. Start designing.

The Architect, VÆN OS