The best interfaces are the ones you stop noticing. Friction is a design decision, not an accident.
Bonside
DemoSample data — last synced Aug 14, 6:48 PM. Connect X and Postgres to see your own bookmarks.
Folder
Author
All authors
Sort
Media
Hold a bookmark, then drag it here
Shipped a side project every month for a year. Here's everything I learned: 1. Scope is the enemy. Cut it in half, then cut it again. 2. Deploy on day one, even if it's a blank page. 3. Momentum beats motivation. A 20-minute session counts. 4. Nobody cares about your stack. They care if it loads fast. 5. The projects that stuck were the ones I built for myself first. The twel…
CSS masonry is finally here and nobody is talking about it. display: masonry just shipped behind a flag. The columns hack era is ending.
Grid systems of the 1960s, redrawn. Müller-Brockmann's poster grids as interactive overlays.
Rate limits are a gift. They force you to build the caching layer you should have built anyway.
Reminder that 'we'll add search later' is how every internal tool dies. Search IS the product for anything that stores more than 50 items.
This post on how Figma scaled their Postgres cluster is the best infra writeup of the year. No magic, just relentless sharding discipline.
figma.comHow Figma's databases team lived to tell the scaleView transitions + server components is the most underrated combo in modern web dev. Native-app feel, zero client state.
New specimen day. Antique grotesques with their quirks intact — the slightly awkward g is the whole point.
Benchmarked cursor-based vs offset pagination on 10M rows so you don't have to. Cursor wins by 40x at page 1000. Offset pagination is a demo-scale illusion.
Winter market, Ljubljana. The light here between 4 and 5pm is unreal.
Suisse Int'l vs. Helvetica Neue — a thread on why the Swiss revival keeps winning in 2026. Look at the terminals on the lowercase a.
Postgres tip that saved us at scale: a partial index on the 2% of rows you actually query beats a full index every time. CREATE INDEX ON orders (created_at) WHERE status = 'pending';
We interviewed 40 people who abandoned their bookmarks folders. The pattern was identical: saving feels like progress, so retrieval never gets designed. Save-time optimism, read-time debt.
Two minutes of surf at Uluwatu, shot on the FX3. Sound on for this one.
Every onboarding flow should be reversible. If users can't undo the setup choices they made on day one, you've built a trap, not a product.
Wrote up our research process for anyone starting out: We don't ask users what they want. We ask them what they did last Tuesday. Specific recall beats hypothetical preference every single time. The full method: recruit 8 people, ask each to walk through their actual last session with the product — screen recording if they'll share it. You're not listening for feature request…
Design portfolios in 2026: less case study theater, more 'here's the thing, click it.' Live work beats process decks.
Your database schema is a love letter to the engineer who maintains it in three years. Usually that engineer is you.
The gap between 'saved it' and 'found it again' is where most knowledge tools quietly fail. Retrieval is the product. Capture is just the pitch.