Inspiration
What it does
How we built it
Challenges we ran into
Accomplishments that we're proud of
What we learned
What's next for In Between
Inspiration
Every event, coffee chat, and intro call leaves you with something valuable: a person, a need they mentioned, an offer they made, a promise you made back. And almost all of it ends up as scattered notes in your phone, a stack of business cards, or nowhere at all.
Months later, a project changes and suddenly that one conversation matters, but the context is gone. You don't remember what they needed, what you promised, or whether now is even the right time to reach out. We kept seeing the same gap: relationship history stays disconnected from what you're working on right now.
InBetween is our answer: a private professional memory that helps you remember the connection, recognize the opportunity, and act at the right time.
What it does
Capture with review. Speak or type a note, scan a business card, or import a document. InBetween extracts people, conversations, interests, needs, and offers, plus your goals, projects, events, trips, and capacity. It asks clarifying questions you can answer, skip, or come back to later, and you edit everything it extracted before anything is saved.
Recommendations with evidence. InBetween connects your saved relationships with your current needs. Every recommendation shows why now, the supporting evidence, what's still unknown, and your capacity limits. It prepares an editable next step and never assumes the other person is interested.
Follow-through with control. Mark any suggestion as Pursue, Later, Not relevant, or Already contacted. Track tasks, commitments, promises, and opportunity outcomes. Paused opportunities can carry reactivation conditions, so they come back when something actually changes. Dismissed suggestions don't keep coming back unless something meaningful changes.
Privacy with clear boundaries. Private workspaces, plus a separate read-only fictional demo so nobody has to share real data to try it. Retention controls, data export, and full workspace deletion. Drafts stay editable, and outreach is never sent automatically.
How we built it
We built InBetween around one rule: the AI suggests, the user decides.
Voice, text, card scans, and document imports go through AI extraction that turns unstructured input into structured records like people, needs, offers, and commitments. Before saving, we added a clarifying-question step and a review screen so nothing gets stored without the user checking it.
Saved relationships are matched against the user's current goals and projects. Every recommendation has to come with evidence, unknowns, and a reason for why now, and it can't just be a bare match.
User decisions (Pursue, Later, Not relevant, Already contacted) feed back into the system so it learns what not to repeat. Workspaces are isolated, a separate read-only demo workspace runs on fictional data, and users get export and deletion controls.
Challenges we ran into
Trust vs. automation. Making AI extraction useful without letting it quietly save wrong information. We fixed this by making review mandatory and clarifying questions skippable, so it helps without forcing anything.
Recommendations that aren't creepy. A suggestion like "reach out to this person" feels wrong without context. Showing evidence and unknowns, and never assuming interest, made recommendations feel like advice instead of pressure.
Not nagging. Early on, dismissed suggestions kept coming back. We had to define what counts as a meaningful change before a suggestion is allowed to come back.
Privacy for demos. People wanted to try it without uploading their real network, so we built a fully separate fictional demo workspace.
What we learned
In relationship tools, timing matters as much as the connection itself. The right person at
Built With
- llm
- node.js
- ocr
- openai
- postgresql
- react
- speech-to-text
- tailwindcss
- typescript
