The Novalis Almanac

The Open Beta Edition

no·va·lis — Latin: newly cultivated land.

Everyone's AI has memory.
Novalis has a mind.

It doesn't just remember what you said, it learns who you are, and it always tells you how.

Join the Waitlist See how it works

I. The Problem

You keep re-teaching the machine.

You have explained your startup to a chatbot before. Then again, a week later. Then again, tonight. Same context, same corrections, same “as I mentioned earlier” — except you never get to finish that sentence, because it never remembers you said it at all.

Some assistants claim to remember now. What they keep is a scattering of notes you never asked for and can’t see, can’t organize, can’t correct — a junk drawer, not a memory. Novalis was built to be read, not just used.

II. Three Fields

What we cultivate.

It Compounds.

Every person you mention, decision you make, and project you’re building gets filed and connected — never dumped in a pile. Talk to it once; it remembers for good. Talk to it again; it already knows where to start.

It Shows Its Work.

Watch it think, in plain view: what it recalled, what it pulled into the conversation, why it answered the way it did. Ask “why?” and get a straight answer back. No black box, no guessing.

You Hold the Keys.

It never rewrites what it knows about you behind your back. It proposes, you approve. Correct anything with one click, any time. A plain-English digest arrives weekly, so you always know exactly what it’s learned.

III. It Meets You

It meets you, not the other way around.

  • Push it hard, move fast, ask a lot: it’ll give you more grounding, not more hype.
  • Come in scattered or overwhelmed: it gets sharper and more direct, not softer.
  • Explain your worst idea and your best one back to back: it explains things back to you the way you actually think, not the way a manual would.

It never does this quietly. Open Personalization and it tells you exactly what it noticed, and when. Don’t like the read? Correct it in one click, same as any memory.

IV. The Almanac

A record that ripens.

  1. Week One

    The first entry is thin, like all first entries.

    Novalis knows your projects, your names for things, and how you like an answer delivered — short, long, blunt, gentle. Nothing more. Nothing has had time to grow yet.

  2. Month One

    By now, the pages start talking to each other.

    It connects a decision you made early in the month to a deadline you mentioned weeks later. It notices when you contradict something you said before — gently, and only when it matters. Patterns surface that you never wrote down on purpose.

  3. Year One

    This is the entry no other AI can write.

    A year of context, cross-referenced and correct. It knows what you tried and abandoned, who you trust and why, and what you actually meant the third time you rephrased the question. Not just an agent that answers — the only one that actually knows you.

The longer you stay, the harder it is to leave. That’s not a trap. That’s the first AI actually worth investing in.

Every AI you talk to forgets what it’s like to be you.

Not the facts. Plenty of them remember facts now. What they forget is the shape of you: how hard you push, what you actually meant last time, why a decision you made mattered.

So you re-explain. Every time. And even when it does remember, you have no way of knowing if what it remembers is right, or how it decided to use it, or whether it quietly changed its mind about you overnight.

Novalis is what happens when an AI is finally willing to be held to something. It tells you what it remembers and why. It admits when it’s wrong, keeps the record, and lets you fix it. It adjusts to you openly instead of guessing at you in the dark.

That’s the whole value, underneath every feature: you’re not just being used by something smart anymore. You’re finally being known by something honest.

V. Subscribe

Be first to the open beta.

No spam, no noise — just the real story as it's built. A few updates on the way, one letter the moment it's your turn.