A librarian that connects you with your people

Open a thread and write. Spark answers like a chat assistant — the unusual part is optional matching to other people writing about the same thing.

A new thread starts Hidden 🔒 — just you and Spark. Nothing is shared, nothing is searchable. You can stay here forever if you want; many threads do.

While you write, something else is happening underneath — Spark compares your topic to other users’ topics to suggest overlaps. Step 3 explains this in detail.

Spark has three modes for who else can be in the room. Pick the one that matches what you want to do right now — you can always open up later (but not close back down, so start small).

Visibility only opens up — 🔒 → 👁️ → 🔥, never back. Someone may already have a copy of the more-open version. To go dark again, delete the thread.

Here’s the part that doesn’t exist anywhere else.

Whatever mode you’re in, Spark is reading the meaning of what you’re writing and continuously checking it against everyone else’s threads. When two threads start circling the same idea, they spark.

Most tools connect you to people who match your profile. Spark connects you to people who match your current question — the one you’re in the middle of, right now. Over time, several of them — not one — because any question worth carrying tends to be circled by more than one mind. Spark’s job is to notice those overlaps, quietly, as they happen.

A few very different first lines that could become sparks if two people wrote them in the same season (labels are who you’re “being” in that half-line, not your profile in the product):

• A PARENT
  “How do I help my teenager who's anxious about climate change without dismissing it?”
• A FIRST-TIME FOUNDER
  “I'm bootstrapping a small product and can't decide whether to charge from day one or build an audience first. How do you make that call?”
• A CURIOUS LEARNER
  “Teaching myself music theory at 40. How do I tell whether I'm making real progress or just noodling?”

Two strangers writing those out in parallel could find each other through Spark without either one searching, following, or posting a feed.

How a spark behaves depends on the mode of the thread that sparked:

How sparks behave:

Your spark bio — the optional one-liner from your profile — rides alongside any anonymous spark you’re in, but only if you’ve opted in via Also show on anonymous sparks in your profile. Visibility is asymmetric on purpose — if you share and they don’t, they still see your bio and you still see nothing of theirs. Toggle it off and the bio disappears from any open sparks immediately.

When you’re signed in, Spark knows your alias. In a solo private thread it can also use your optional spark bio, your city/country, and subscription state so a “what do I pay?” answer can be specific. In a group, other people see Spark’s reply too, so it is shown only your alias — bio and billing details are withheld on purpose. In an anonymous spark it sees nothing personal at all.

Everything — the app, your data, our email, our monitoring, and the AI itself — runs in the EU. Your conversations are processed by Mistral AI SAS, a French company headquartered in Paris — both Spark's replies and the semantic embeddings used for matching. All processing and storage happens inside the EU, on EU-owned infrastructure.

Spark is in closed beta — invite-only. One paid plan at €19 a month, VAT included, with the first month free. Sold as a personal subscription; multi-seat and company billing isn’t available during beta. That subscription is the whole business model — no ads, no tracking for sale, no free tier propped up by selling your attention to a third party. You’re the customer here, not the inventory.