Every product needs a one-liner, and the failed attempts are often more informative than the one that sticks. Pouch kept getting pulled toward three defaults — "note app," "cloud drive," "email for myself" — and each of them was the wrong shape for what we actually wanted to build. The exercise of naming what pouch isn't produced the sharpest version of what it is.
The recurring frustration that started it
I need this file on that machine, in the next five minutes, and every option is bad. Email it to myself; the attachment hasn't arrived yet and I look like an idiot in front of a meeting-room projector. Drop it in Google Drive; log in on the demo machine; fight the 2FA prompt; still can't find the right folder. AirDrop; only works if both are Apple, both on the same LAN, both signed in to the same iCloud. USB stick; I own three and none is in the bag today.
None of these are notes. None are files I want to organize into a tree. They're things in transit. Ephemeral, short-lived, between-places. The tools I reach for are optimized for storage or collaboration — the wrong optimization for what I'm doing.
Three framings that were wrong
Not a note-taking app
Notes apps optimize for forever-storage and rich editing. Outlines, backlinks, collaborative cursors, long-form capture. Great for a second brain; wrong for "I need this in five minutes on the demo machine." Notes apps also implicitly suggest you should keep everything, which is the opposite of the behavior we wanted to encourage.
Not a cloud drive
Drive products optimize for mirroring folder trees across devices. You organize into hierarchies, sync, share folders with teammates. Also great for its job; also wrong for ours. A drop in pouch is not a file in a folder. It's a thing with a lifespan — possibly very short — that should be reachable by any client you have, then gone.
Not email-for-myself
The "mail a copy to yourself" pattern has a lot of what we want: any-device retrieval, low ceremony, forgettable. But email is threads and attachments and contacts and filters and rules — and somebody else's infrastructure deciding what's spam. We wanted the retrieval pattern without the conversational overhead.
What we chose: personal data relay
"Relay" is load-bearing. A relay connects two points. It doesn't optimize for keeping things; it optimizes for moving things through. Most of what a relay handles passes quickly; a small fraction gets pinned and kept. The pinning is explicit.
That gave us the one-liner:
Pouch is a personal data relay. Ephemeral by default, durable by opt-in, yours to own.
Each phrase earns its place. Ephemeral by default: drops have a TTL you set, and most expire quietly. The default is not "keep." Durable by opt-in: the small things that deserve to last move explicitly into the "kept" stream — a deliberate act, not passive drift. Yours to own: self-hostable (eventually), exportable always, with encryption options for the most sensitive material. You know where your bytes live and you can take them with you.
What cascaded from this framing
Framing decisions compound. Once we said "relay, not note app," several product decisions became obvious:
- TTL is a first-class field on every drop. Not a hidden flag. You see the expiry on the card.
- Streams are policy lanes, not folders. "Inbox" is the ephemeral default; "kept" is durable. More on this in the next post.
- Delivery is equally important as capture. A relay needs both ends. We spent significant design time on the "how does the drop get onto the target machine" problem, with a multi-tier trust model (intra-pouch push, short URLs, QR codes, pickup codes, LAN peers).
- Paid from day one, no free tier. Tools that handle your personal data should be funded by you, not by monetizing it.
Why this frame and not another
The test for a positioning is whether it forecloses things. "Personal data relay" forecloses conversation threads, social features, a public content layer, AI-summarization-first workflows, and dozens of adjacent product directions that sounded interesting but aren't this. That foreclosure is not a bug; it's the whole point. Every feature we don't build is one we don't have to maintain, defend, or migrate away from later.
The positioning also gives us something to say "no" against. When someone proposes — very reasonably — that pouch should do X, the answer is "not if X makes us into a different category." This has saved us twice already, and the product is still in pre-launch.
Open questions we're holding
Some things we haven't fully decided:
- Where the vault (client-side encrypted, zero-knowledge storage) slots in. Core promise or opt-in tier? Tied to streams or orthogonal?
- How "smart" enrichment (LLM-driven auto-tagging, summarization) integrates without drifting into note-app shape.
- Whether the connector ecosystem should be monorepo-owned or open to third parties earlier. Current posture: monorepo through Phase 2, open in Phase 3.
We'll write each of these up as the answer clarifies.
Next post: Streams, not folders — why pouch chose policy lanes over hierarchy.