Pouch, Newsletter Reader

Newsletters arrive in a regular inbox. Mixed with promotions, notifications and spam, they get flagged as unwanted, forgotten, or lost in thousands of unread messages.

Pouch gives you a dedicated address for newsletters only. Everything arrives in one place, in an interface designed for reading.

Active newsletter readers: creators, developers, designers, founders. Someone subscribed to 10, 15, 20 different newsletters who actually wants to read them, not just accumulate them.

Before touching Figma, I mapped the onboarding as a flowchart. Every step, every decision point, every screen transition. Once the structure was clear, I used Claude to generate the screens based on what I had already defined. From there I reviewed, refined and iterated to get to a v1.

I also used Claude for the token system and prototype interactions.

The design decisions came from me. Claude handled the execution.

Pouch is a text app. Its core content is long-form reading. Every decorative element that doesn't serve that is visual noise. I kept the interface dense and minimal to leave room for what actually matters: text and typographic hierarchy.

I treated both modes as equal, with a slight preference toward light mode as the default.

Dark mode works in specific contexts: dim rooms, occasional use, OLED screens. For a reading app used throughout the day, light mode performs better. Dark text on a light background reduces reading effort and eye strain over time. Dark mode in Pouch is a valid option, not the primary direction.

The onboarding uses a split-screen layout: form on the left, app preview on the right. Inspired by Sana. The user sees where they're going before giving their email.

I evaluated three options:

AI-generated image. Not viable at scale. It would require reading, summarizing and generating a banner consistent with each newsletter's brand. Massive cost, marginal benefit.

Category-based visual. Too generic. Users lose each sender's visual identity.

Sender logo + initials/unique color fallback. Fast, recognizable, scalable. Zero server cost.

I went with option 3. Familiar pattern, zero server cost, the user identifies their newsletter in a fraction of a second.

I added a search bar and 3 filters: All, Unread, Recent. These felt like the most essential ways to cover what most users actually need without overloading the interface.

I kept the filter selection deliberately subtle. It's not a feature that deserves high visual hierarchy. The sidebar layout already changes enough between the 3 modes that users know where they are without even looking at the active filter. Pushing more emphasis on those buttons would create noise for very little gain.

What the user actually wants is to read their newsletters. Everything else should get out of the way.

Sana for onboarding. Cloaked for menus and overall density spacing and component simplicity. Discord for sidebar logic and information density.

Onboarding (7 steps, light + dark)

Home

Inbox

Reader

Clickable Figma prototype

The most interesting decisions on this project weren't visual. Choosing the logo over AI, defaulting to light mode for functional reasons, designing an onboarding that shows before it explains. All of these came from thinking about the product problem first.