← The Architect
For a decision that matters

Help making a big decision

A career move, a relationship, leaving a city. Not ChatGPT, not a diary — a private space to write what you actually think, and a mentor that reads it, remembers your context, and pushes back instead of flattering you.

Most big decisions do not get stuck because you lack facts. They get stuck because two versions of you want different things, and naming that out loud feels like losing something. So you research instead. You make another list. You ask three friends who all tell you what you want to hear. The decision stays exactly where it was — the subject of a system for people who overthink rather than a problem more thinking will solve.

The Architect is built for the part that actually moves a decision: writing the whole thing down honestly, and being read by something that remembers what you said last time and is willing to disagree with you. Clarity, not comfort.

Why a hard decision stays stuck

It is a clarity problem wearing an information costume

When a decision matters, the instinct is to gather more — more data, more frameworks, more opinions. But past a certain point, more information makes it worse: it gives the overthinking somewhere to go. The decision feels like a puzzle with a correct answer you have not found yet, when it is really a position you are reluctant to take.

Writing changes that, because writing forces specificity. The vague dread of "I don't know what to do about my job" becomes the much more answerable "I'm afraid that if I leave, I'll discover the problem was me." But a blank page only takes you so far — you don't need another journal, you need a mentor that remembers, because the value is not in venting once. It is in being read across time by something that notices the pattern.

What it does

Writes it down, reads it back, then pushes

You write the decision out in full. Not a tidy summary — the actual tangle. What you are weighing, what scares you, the version you would never say to the people involved. Type it or speak it; voice input transcribes with Whisper, so you can think out loud on a walk.

The mentor reads it against everything you have written before. This is the part a fresh chat window cannot do. After three or more entries, pattern detection runs across your whole journal, so the mentor can notice that this week's framing of the decision quietly contradicts what you wrote a month ago — and say so. That is the actual mechanism behind how to stop repeating the same mistakes: seeing the shape of the loop from outside the moment, not just resolving to choose better next time.

Then it pushes back. You pick a mentor persona, and each one has a real point of view rather than a reflex to agree. The Stoic asks what is actually within your control and what you are pretending is. The Sage asks the question behind the question. The Billionaire strips the situation to where the leverage really is. None of them are tuned to keep you comfortable — which, when something important is on the line, is the entire point. If you have not seen that distinction drawn out, what an AI mentor actually does is the longer version of the argument.

How it works

Personas, memory, privacy

Five mentor personas, plus a custom one. The Stoic, The Sage, and The Mystic are on the free tier. The Billionaire and The Traveler from 2075 unlock on the paid tiers, alongside a Custom persona you design — your own mentor, with the philosophy and bluntness you actually want pointed at your decisions. Mentor responses run on Claude Sonnet 4.

Memory that accumulates. Every entry stays as context for every future response. Pattern detection runs across your full journal after three or more entries. A real mentor knows what you said three months ago and notices when today contradicts it; that continuity is the whole difference between a fresh chat and a relationship — which is also why your best thinking is gone by noon matters here, because offloading the open loops onto the page is how you protect the judgment you have left.

Privacy that is structural, not a promise. Your entries are encrypted on your own device with AES-256-GCM before anything syncs. The key is generated in your browser and never leaves it; the server only ever holds ciphertext that no one — including the founder — can decrypt. You own the recovery key. This is not a stricter policy, it is a different architecture, and it matters more than it sounds: privacy changes how honestly you write, and on the decisions you would least want exposed, honesty is exactly where the value is. It is the premise behind whether an AI can be a confidant at all — pattern memory plus real privacy is what makes the difference.

Trilingual throughout: English, Turkish, and Spanish, across the interface, the mentor's responses, and voice in and out. There is also PIN lock, full export of everything you have written, and TTS playback if you would rather hear a response than read it. It is a web app today; an iOS app is in development.

A good decision is not the one that feels best in the moment. It is the one you can still defend after you have written down the part you were trying not to say.
When this fits — and when it doesn't

This is for you if

You are weighing something real — a job, a relationship, a move, a fork in the road — and you keep circling without landing. You want continuity across the whole decision, not a fresh chat each time. You want to write the honest version somewhere mathematically private, and be pushed instead of soothed.

This is not for you if

You want the tool to make the call for you, or to tell you that you are right. You are in crisis or need clinical care — this is honest self-reflection for when you are not in crisis, with no medical claims; please talk to a human professional. Or you just need a general assistant to draft and summarize, where ChatGPT wins.

Three kinds of big decision

Whether to leave your job

You write out the case for staying and the case for leaving. The mentor reads it against the resentment you wrote about months ago, the conversation with your manager you mentioned in passing, the version of yourself you said you wanted to become. Then it asks the question underneath: what part of staying is about the work, and what part is about the story you have built around it? The pattern is visible because the journal remembers it — not because you summarized it well today.

A relationship at a crossroads

The hardest part is admitting what you actually feel, which you will only do if you are certain no one else can read it. Encrypted on your device, you can write the unguarded version. The mentor asks what conversation you are not letting yourself have, and surfaces the same dynamic showing up in a new shape if you have been writing for a while. For where the line sits between reflection and a human, how to talk through relationship issues with an AI — and when to see a human instead.

A move, a leap, a reinvention

Relocating, starting the thing, betting on a different life. These decisions stall because the fear and the excitement use the same words, and at night they get louder — which is its own open-loop problem, the subject of how to stop overthinking at night. Writing closes the loop enough to think. The mentor separates the fear that is information from the fear that is just noise, and keeps you honest about which is which across however many entries it takes.

FAQ
Can AI actually help me make a big decision?
Not by deciding for you. A high-stakes decision is rarely blocked by a lack of information — it is blocked by a lack of clarity about what you actually want and what you are avoiding. The Architect helps by giving you a private place to write the decision out in full, then reading what you wrote against everything you have written before. It surfaces the contradictions, names the thing you keep circling, and asks the question behind the question. You still decide. It just makes sure you are deciding with your eyes open.
How is this different from asking ChatGPT for help making a decision?
ChatGPT is a brilliant general assistant, but it starts most conversations fresh and is tuned to be agreeable — so when you propose a course of action, the default response is supportive. For a real decision you need the opposite: continuity and friction. The Architect keeps every entry as context for every future response, runs pattern detection across your history after three or more entries, and uses mentor personas built to push back rather than placate. It is also zero-knowledge: your entries are encrypted on your device before they sync, so even the founder cannot read them. The full breakdown is in The Architect vs ChatGPT.
How does the mentor push back instead of just agreeing with me?
You choose a mentor persona, and each one has a consistent philosophy rather than a single people-pleasing voice. The Stoic does not console; it asks what is actually in your control. The Sage asks the question behind the question you thought you were asking. The Billionaire strips the situation down to where the real leverage is. Because the mentor can see your full journal, it can also do something a fresh chatbot cannot: notice when this week's framing of the decision contradicts what you wrote a month ago, and say so. A single-voice AI mentor handles this differently — see The Architect vs Purpose.
Is what I write about my decision private?
Yes. Your entries are encrypted on your own device with AES-256-GCM before anything syncs, using a zero-knowledge architecture. The encryption key is generated in your browser and never transmitted, so the server only ever stores ciphertext that no one — including the founder — can decrypt. You own the recovery key. The decisions that matter most are usually the ones you are least willing to write down somewhere a breach or policy change could expose, which is exactly the case for writing them somewhere mathematically private.
Is this therapy, or a substitute for it?
No. The Architect is a tool for honest self-reflection when you are not in crisis — thinking through a decision, a pattern, a fork in the road. It makes no medical or clinical claims and is not a replacement for therapy or professional care. If you are in crisis or dealing with something that needs a clinician, please talk to a human professional. For the lower-stakes work of getting clear on a hard choice, a private space and a mentor that remembers is exactly the right tool.
What does it cost to try?
There is a permanent free tier called Seeker with no card required — four entries on your first day, one per day after that, and three of the five mentor personas (The Stoic, The Sage, The Mystic). That is enough to write out a real decision and see how it responds. Paid plans are Builder at $15/month and The Architect at $25/month or $199/year, which works out to $16.50/month billed annually, for the full persona library, a custom persona you design, pattern detection, voice in and out, PIN lock, and full export.
Keep reading

If you are in the middle of a hard call right now, these go deeper on the parts that get stuck.

Think it through →
Permanent free tier · No card required