Valley Speak

Silicon Valley idioms, one short at a time

Swipe up for the next short. Keys: J/K to move, 1–3 to answer.

The playbook

Six hooks from the feed, pointed at your memory

Shorts and Reels are tuned to keep you watching. Valley Speak borrows the same mechanics and aims them at one outcome: phrases you can still pull up in a meeting a month from now. Each pair shows the original hook and what the app does with it.

In the feed

No bottom to the page

Infinite scroll removes the moment where you would decide to stop. Its inventor, Aza Raskin, has said he regrets it.

In Valley Speak

Rounds that end

Six new phrases plus recall, then a stop card. Swiping feels the same, but the round finishes, because the gap before the next review is part of how memory works.

In the feed

A slot machine under your thumb

Every swipe might pay off, and you never know which one will. Tristan Harris compared pull-to-refresh to a slot machine. Dopamine neurons fire hardest for rewards you did not predict (Schultz, 1997).

In Valley Speak

Every card is a small bet

You guess the meaning before it is shown, so each reveal is a win or a surprise. Rare origin-story cards turn up at random.

In the feed

Motion before words

An internal TikTok document, described by The New York Times in 2021, scored videos on predicted likes, comments, plays and watch time. The opening second decides whether you stay.

In Valley Speak

Open on a moving picture

Each card starts with an animated scene of the phrase taken literally, like a man holding a lighter to the ocean. You see it before you read a word.

In the feed

The loop

Shorts replay on their own. Creator guides report that rewatches are one of the strongest ranking signals.

In Valley Speak

Every replay is a rehearsal

Scenes loop, and the Replay button counts your rewatches. Each pass strengthens the link between the picture and the phrase.

In the feed

A feed that learns you

The ranking picks whatever is most likely to hold your attention next.

In Valley Speak

A feed that ranks by forgetting

Phrases return just before you would lose them: after 1, 3, 7, 21, then 60 days. Cepeda and colleagues (2008) found the best gap is roughly 10–20% of how long you need to remember.

In the feed

One thing fills the screen

Full-screen video leaves nothing else competing for your eyes.

In Valley Speak

One phrase, one picture, one line

Richard Mayer's coherence principle: people learn better when extra words, pictures and sounds are cut. Each card carries one scene and one plain definition.

What the app leaves out

Like counts, autoplay into another round, streaks that punish a missed day, and late-night notifications. In 2023 a bipartisan coalition of US state attorneys general sued Meta, arguing that features like these were built to keep young users hooked. The mechanics are powerful, so the app keeps the parts that serve memory and drops the rest.

The memory science

Your hunch about pictures is right. It's half the answer.

Pictures are remembered far better than words. But recognizing a picture is easier than producing the phrase on demand in a meeting. So the app pairs every picture with a test.

83%

of 10,000 photos were recognized after one 5-second look at each.

Standing, 1973→ every phrase gets a scene
61% vs 40%

of a passage recalled a week later after practice tests, compared with rereading. Rereading won on an immediate test, which is why it feels like it works.

Roediger & Karpicke, 2006→ recall cards
Wrong guesses help

Trying to answer before studying improved later learning, even when the attempt failed.

Kornell, Hays & Bjork, 2009→ guess before the reveal
Sure and wrong

Errors made with high confidence were the most likely to be corrected after feedback.

Butterfield & Metcalfe, 2001→ the “I'm sure” toggle
Weird and funny

Bizarre images improved delayed recall, and humor carried the effect. It works when the odd item stands out from ordinary ones.

Worthen & Deschamps, 2008→ absurd scenes, plain definitions
Draw it

Drawing a word beat writing it out for later recall.

Wammes, Meade & Fernandes, 2016→ planned “sketch it” mode

About being a “visual learner”

Pictures help nearly everyone, so this is not about having a visual learning style. Pashler and colleagues (2008) reviewed the learning-styles research and found almost no solid evidence that matching teaching to a person's preferred style improves results. Use pictures because human memory favors them across the board.

Blueprint

From this prototype to real shorts

This page is the prototype: 21 phrases, animated drawings, a guess, reveal and recall loop, and a review schedule saved in your browser. The real app swaps the drawings for 15 to 20 second vertical videos and keeps everything else.

Anatomy of one 20-second short

0–2sLiteral scene, no text
2–7sThe twist: narrated meaning, caption shows only the phrase
7–13sHeard in a meeting: one real sentence, acted out
13–17sPause for your guess
17–20sBack to frame one so it loops

Narrate the explanation and caption only the phrase itself. Mayer's redundancy principle found that graphics with narration beat graphics with narration plus the same words on screen.

The pipeline

  1. Pick phrases you can picture. Each needs a literal image: a yak, a moat, a runway. Abstract jargon gets a bizarre stand-in.
  2. Draft scripts with an LLM, check origins by hand. Origin stories are where invented facts creep in. Every card should trace to a source, like the ones below.
  3. Animate the literal scene. Rive or Lottie for looping vector scenes like these, or an AI video model for richer ones, with text-to-speech narration.
  4. Schedule with FSRS. The open-source scheduler available in Anki. Feed it guess results, confidence, replays and time on card.
  5. Ship on the web first. This page already works on a phone. Expo (React Native) is the route to the app stores once the loop proves itself.

Guardrails

  • A daily cap and a stop card at the end of every round.
  • Reminders in the morning only, never at night.
  • Streaks that forgive a missed day.
  • One success metric: phrases recalled after 30 days, not minutes spent in the app.

Sources

  1. Aza Raskin on infinite scroll — Gulf News report
  2. Harris, “How Technology Hijacks People's Minds” (2016) — HuffPost
  3. Schultz, dopamine reward prediction error — review, PMC
  4. “TikTok Algo 101,” as described by The New York Times (2021) — summary; rewatch signal per Hootsuite
  5. States sue Meta (2023) — NPR
  6. Standing, “Learning 10,000 pictures” (1973) — journal
  7. Roediger & Karpicke, “Test-Enhanced Learning” (2006) — journal
  8. Kornell, Hays & Bjork, “Unsuccessful Retrieval Attempts Enhance Subsequent Learning” (2009) — paper
  9. Butterfield & Metcalfe, “Errors Committed With High Confidence Are Hypercorrected” (2001) — paper
  10. Worthen & Deschamps, humor and bizarreness (2008) — PubMed
  11. Wammes, Meade & Fernandes, the drawing effect (2016) — ScienceDaily
  12. Cepeda et al., “Spacing Effects in Learning” (2008) — PDF
  13. Mayer, multimedia learning principles — chapter
  14. Pashler et al., “Learning Styles: Concepts and Evidence” (2008) — PDF
  15. Idiom origins: dogfooding, yak shaving, disagree and commit

Your progress is saved in this browser only.