Creator: Enrico Tartarotti · Published: 2026-02-20 · Length: 17:02 · Watch on YouTube

The most satisfying interface in the world isn't satisfying by accident. (from the video description)

Enrico argues that the short-form video feed works because of three layers, and only the first one is visible: the interface, the content it shapes, and the recommendation algorithm. Features short-form creator Alasdair (channel Alementary). Sponsored by UX Pilot. Written by Enrico Tartarotti; produced by Enrico Tartarotti and Bartek Malinowski; edited by Seequence.

Summary written from the auto-generated captions and sampled frames; the video has no chapter markers, so section times below come from the captions. The captions garble some names (Enrico's own name appears as "Ammar" and "Mariko"), so names below follow the description where possible.

1) Detailed outline

[0:00] The most successful interface in the world

• Every day we collectively watch about 616,572 years of content through the short-form feed, one of the most hyper-optimized pieces of technology there is. Every button and tiny decision is engineered to make you spend more time.
• Custom scroll physics: most apps use the iOS or Android UI kit to get a scrollable list for free. Instagram built its own UI kit that simulates velocity and acceleration, so in the Reels feed a tiny flick moves you to the next video, making it easier to scroll in bed at 3 a.m.
• There are three layers that make this work, and only the first is what you can see.

[1:15] Layer 1: The interface

• Speed over smoothness: Apple and Google designers spent hundreds of hours tuning home-screen animation curves so the phone feels like real-world physics. TikTok, Shorts and Reels animate much faster, optimized purely to get the next video to you sooner.
• Full-screen video, floating UI: no containers or bars; the interface sits on the video with a slight shadow for just enough contrast.
• Thumb reach zone: on Instagram, TikTok and YouTube Shorts the like, comment and share buttons sit in the same spot on the right, where your thumb rests. They want you to press them.
• When you swipe, the buttons move with the video and fade slightly, signalling that likes and comments belong to that video rather than the overall interface.
• Controls you're not meant to use: options like tweaking your algorithm or switching feed type on Instagram sit at the top, deliberately cumbersome to reach. There's no grid of thumbnails and no menu, only the next video the algorithm picked.
• Unit bias: one unit of something feels like the right amount, so you get one video and one thing to focus on.
• No endings: videos loop with no signal that they've ended. Clips under 15 seconds have no play bar at all, and longer ones get a tiny one, because seeking is friction; if you want to skim, the intended action is to swipe.
• Chronoception: like casinos that remove clocks and windows and are built as mazes, the feed removes every cue that content is starting or ending so the session blurs and you lose track of time.
• No spinners: YouTube shows a loading wheel; short-form feeds preload the first frame of the next video and show it while the rest loads, so you never notice you're waiting.
• Sign-in walls: TikToks are indexed on Google, but clicking one hits a sign-in wall and a hard push to download the app. Regular YouTube still lets you watch. For short form the single video isn't the valuable part (there's no ad on it); the value is you getting hooked on the algorithm, which needs you signed in.

[5:05] Why the feed became a design pattern

• Enrico shows a concept from an older video where swiping diagonally lets you dig deeper into a topic or explore something different. Platforms haven't built anything like it because short-form UI is about extreme simplicity: no carousels, no grids, one gesture.
• Hick's Law: the more options you have, the harder it is to decide. YouTube gives you a grid, Netflix gives you rows, short form gives you one video and one action, swipe.
• The feed was popularized by TikTok, then copied by Instagram and YouTube Shorts; Spotify tried it and Amazon launched an Inspire feed. It may be the most replicated interface ever.
• It went from "how TikTok shows videos" to a design pattern, like a right-click menu with submenus or "the Tinder of music," which people understand instantly without instructions.
• Each platform adds a twist: Instagram shows friends' profile bubbles on videos they liked, wrapping the feed in a social layer; YouTube Shorts links a related long-form video at the bottom as a reminder you're still on YouTube.

[7:15] Layer 2: The content

• Thought experiment: as an Instagram product manager on Reels, how many characters do you give creators for captions? It seems inconsequential.
• Alasdair, a friend who makes short-form videos with hundreds of millions of views, explains how the interface shapes the content itself. Another decision product managers had to make was whether the video pauses or keeps playing when you open comments.
• On YouTube long-form, Enrico says the interface barely affects his videos (beyond keeping the bottom-right thumbnail corner clear for the duration badge). On short form, interface and content are deeply tied.
• Creators made videos where you had to open the comments to get the full value, and kept them very short so people would loop and rewatch while reading, pushing retention to something like 500%.
• Platforms handle comments differently: some cover the video, others squeeze it smaller. Squeezing keeps the video easy to follow; covering it focuses you on the comments.
• First-frame optimization: the next video used to load its first frame but not start playing until it was fully on screen, so creators optimized the top half of the first frame, which viewers were already taking in mid-swipe.
• Caption length, autoplay, overlay versus shrinking comments: these innocuous decisions change how millions of people create content.

[10:02] Sponsor: UX Pilot

• Enrico says he spent a whole day at the start of last year recreating Instagram's UI by hand for his free-scroll prototype. In UX Pilot he prompts the same flow and gets screens with interface and imagery.
• The Figma plugin generates wireframes and designs directly in Figma as fully editable layers, and you can import your design system from Figma into UX Pilot.
• Pitch: it isn't replacing designers but giving founders, engineers and product managers a way to communicate ideas visually. Code BUILD15 for 15% off.

[11:14] The medium has always shaped the media

• In 1949, 7-inch records held about 3 minutes per side and AM radio wanted short songs, so 1950s and '60s pop songs ran about 2½ minutes. LPs and CDs let songs get longer.
• TV writers built tension around ad breaks; streaming removed the need, which is why Friends feels oddly paced when you binge it.
• Short form takes this further because creators get a second-by-second retention report.
• Alasdair on analytics: on every platform you want a high share of people choosing to watch, then a retention line as flat as possible. The gap between viral and not is tiny: going from about 75% to 80% average watch on YouTube Shorts can be the difference between 100,000 and 10 million views, so small details like which caption words show first add up.

[13:18] Not evil, just optimized

• No designer sat down to build something manipulative. Platforms run hundreds of experiments at once (which is why your friend's app may look different from yours), keep what raises engagement and kill what doesn't.
• Over time the winning designs exploit psychology, not by choice, but because "the math just converges there."
• Creators converge the same way: identical retention graphs push everyone toward the same hooks, editing style and tricks. The platform doesn't just host the content, it designs it.

[14:06] Layer 3: The algorithm

• An ex-TikTok engineer called it the most valuable piece of software in existence: the secretive algorithm that picks what we watch among billions of videos.
• ByteDance's research paper (the Monolith real-time recommendation system; the video says 2023, and the paper shown on screen was presented at RecSys 2022 in Seattle) explains how it differs from YouTube- or Netflix-style systems.
• Traditional systems train on historical data, then serve recommendations, and by the time training finishes your interests have moved on. The paper calls this concept drift: visual and linguistic patterns barely change over centuries, while one user's interest "could shift their zeal every next minute."
• TikTok's pipeline: every like, skip, loop and swipe is logged into a Kafka queue streaming from everyone's phones. A joiner matches each action with your profile, history and context (time of day, etc.) to create a training example fed straight into the model.
• The model is updated every minute, not weekly or daily, faster than you can finish the TikTok you're watching.
• A chart from the paper compares accuracy over 50 hours: a model frozen at the start versus one updated every 30 minutes. The gap is about 13–14%, and Enrico calls it the secret sauce that makes short-form feeds more addictive than anything before.

[16:13] Outro

• When something becomes this ubiquitous, it's usually because there's a huge amount of hidden design, engineering and psychology underneath.
• Teaser for his video on how tech sounds disappeared (Why Our Tech Got Silent).

2) Things mentioned

Platforms, apps and companies

• TikTok (and parent company ByteDance)
• Instagram Reels (Instagram's custom UI kit, friend bubbles, feed-type and algorithm controls)
• YouTube Shorts (related long-form video link) and regular YouTube
• Netflix (row-based browsing)
• Spotify (tried a short-form-style feed)
• Amazon Inspire (Amazon's short-form shopping feed)
• Google Search (indexes TikToks behind a sign-in wall)
• Tinder (the swipe pattern, as in "the Tinder of music")
• iOS and Android UI kits and home-screen animation curves
• Figma
• UX Pilot (sponsor; prompt-to-screens and wireframe generation, Figma plugin, design-system import)
• Flask (Enrico's video-collaboration app)

Engineering and algorithm concepts

• Custom scroll physics (simulated velocity and acceleration)
• Preloading the next video's first frame instead of a spinner
• Monolith, ByteDance's real-time recommendation system paper (RecSys 2022)
• Concept drift
• Apache Kafka message queue for user actions
• The joiner (a streaming job, shown as a Flink job in the paper's diagram) that turns actions plus profile data into training examples
• Online training with model updates every minute; frozen vs. 30-minute-updated model comparison (about 13–14% better)
• A/B testing at scale (hundreds of concurrent experiments)
• Retention graphs and average percentage watched

Design patterns and psychology

• Thumb reach zone for engagement buttons
• Floating UI with subtle shadows over full-screen video
• Deliberately hard-to-reach feed controls
• Looping videos with no end signal; no play bar under 15 seconds
• Sign-in walls and app-download prompts
• Unit bias
• Chronoception (and casino design without clocks or windows)
• Hick's Law
• Design patterns that need no instructions (context menus, swipe-to-like)
• Enrico's diagonal-swipe feed concept from an earlier video
• Comment overlays vs. shrinking the video
• First-frame and top-half-of-frame optimization by creators

Media history

• 7-inch records (1949, about 3 minutes per side) and AM radio's preference for short songs
• LPs and CDs allowing longer songs
• TV writing around ad breaks; Friends pacing on streaming

Other references

• Alasdair's channel Alementary
• Enrico's related video Why Our Tech Got Silent

3) Biographies

Enrico Tartarotti

Italian-born YouTuber, designer and entrepreneur whose channel goes "behind the scenes of the design and psychology and stories behind the tech you use everyday." He builds Flask, a video-collaboration tool, offers consulting calls, and produces his videos with Bartek Malinowski, with editing by Seequence.

Alasdair (Alementary)

A short-form video creator and friend of Enrico's whose videos have drawn hundreds of millions of views; his YouTube channel is @alementary. In the video he explains how platform interfaces shape short-form content and how thin the retention margins are between ordinary and viral videos. The video doesn't give more background on him.

The ex-TikTok engineer

Enrico quotes a former TikTok engineer who called the recommendation algorithm "the most valuable piece of software in existence." The auto-captions render his name as "Armand Conqueer," and the video doesn't spell it out, so his identity isn't confirmed here.