A full-screen learning feed for data scientists and solution architects. Live news from the site's feeds with a thinking prompt on every story, plus flashcards, built quizzes, terms and deep reads from the library, each in its own look and sound. Every card you engage with grows a petal.
The brief it was built from
A place where data scientists and solution architects can scroll as if it were Instagram or LinkedIn, with learning as the objective. Byte-sized news from the AI and architecture feeds, with flashcards, topics and concepts dropped in. React, like, skip and share cards (images and links) to social media. Fully dynamic, never repetitive in content, look or sound, with no login. Lives on the feed pages and on the Design landing page.
How it works, honestly
What it runs on
- News cards are the site's own feed entries (Google Alerts on AI, agents, LLMs, data science, solution architecture, system design and more), newest first, as the poller stored them; links go to the publisher
- Library cards are the site's own content, live: concepts and flashcards from the AI & ML library, quiz questions, flashcards and glossary terms from the architecture practice, editorials from both desks, architecture field guides, anatomy dissections and the other Design pieces
- Concept and term quizzes are built from the library, never written by hand: one entry's summary with its name blanked out, and three other entries from the same track or topic as the choices, so the right answer is right by construction
- The pulse card counts the words in this week's headlines against the week before; the thread card pairs two stories from different publishers that share a phrase in the same ten days
- The background link on a story matches its keywords against the library with the same keyword index the knowledge graph uses
- Today's picks are the same daily picks the /learn and /architecture hubs show every visitor
Limits & choices
- "Never the same card twice" holds per browser: the shuffle seed, the shelf bookmarks and the seen-news ranges live in this browser's storage, so another browser or a cleared one starts afresh. A private window forgets on close
- Near-duplicate headlines are judged by word overlap (62% of the distinctive words in common), so the odd syndicated story with a reworded title can slip through, and the odd distinct story can be dropped
- The lenses are general habits of mind written for this piece, not a reading of each article; a lens can fit one story better than another
- Petals measure engagement (a story read for six seconds or opened, a flashcard graded, a quiz answered, a concept stayed with for seven seconds), not understanding
- Your mix learns from dwell time, reactions and skips in this browser only; the thresholds (under 1.3 seconds is a skip, over 7 seconds is interest) are design choices, not research
- Like and mind-blown counts are per connection (a salted hash of the address), not per person, and show only once a card has three
- The breather card's eye rest follows the 20-20-20 rule of thumb that eye doctors recommend; the evidence behind its exact numbers is thin
- When the deck API is out of reach the page falls back to a saved set of library cards and says so; the news never comes from that saved set
What it teaches
- The newest stories in your field, each read through a question a careful data scientist or architect would ask
- Retrieval practice in passing: flashcards and quizzes between the news, the kind of testing that improves long-term retention
- Spaced second looks at the flashcards you missed
- A map of the site's library, one card at a time: concepts, terms, editorials, field guides and 3D dissections you can open in full
Sources
Stories belong to their publishers; each card links to the original · Fonts: Inter, Fraunces, Unbounded, JetBrains Mono, Caveat (SIL Open Font License)
Learning paths
Original
Learning paths
Original
Learning paths
Original