18:42
J
I Tried Every Coffee Shop in My City
James Does Stuff
2.1M views · 3 days ago
YouTube as-is
My feed did not know I was learning German.
Useful German videos existed, but they competed with every other recommendation and disappeared as soon as our normal viewing habits took over.
YouTubeDE
AllDocumentariesScienceTravelTechnologyNews
12:05
T
World's Strongest Magnet vs iPhone
Techsperiments
8.7M views · 1 week ago
24:18
M
We Built a Tiny House in 48 Hours
Maker Duo
3.4M views · 5 days ago
31:56
H
The Real Story Behind the Roman Empire
History Bites
1.2M views · 2 weeks ago
22:47
T
$1 vs $10,000 Vacation!
TravelMaxx
15M views · 3 weeks ago
14:22
C
Why Planes Don't Fly Over Antarctica
CuriousMind
4.9M views · 6 days ago
16:33
S
Ranking Every Fast Food Fries
SnackAttack
980K views · 2 days ago
09:57
D
My Morning Routine as a Programmer
DevDiaries
410K views · 1 day ago
20:11
S
Extreme Hide and Seek in a Mall
Squad Games
6.2M views · 4 days ago
Project frame
What we built, and what I owned.
- Team
- Two international founders in Germany
- My role
- Shared product direction; built the current extension codebase and deterministic demo
- Build window
- May–July 2026
- Program
- MM startup program
- Status
- Internal testing; launch planned after MM
- Product
- LAD (Learning All Day), a YouTube-native German learning extension
Build story
From a language filter to a learning layer.
- 1May 2026
Make the feed language-aware
The first build filtered YouTube surfaces by language and nudged locale settings without replacing the platform.
- 2May–June
Fix the cold start
Onboarding, starter channels, scope controls, original-audio handling, and local tracking made an empty filtered feed useful from day one.
- 3June–July
Add the learning layer
CEFR and Goethe-based cues, the curated German feed, progress views, and a catalog of real videos turned filtering into a repeatable habit.
- 4Current
Test the story before launch
We are using deterministic states for internal testing and launch preparation, with a Chrome Web Store release planned after MM.
Architecture
LAD stays inside the existing YouTube habit.
YouTube UI
The learner stays in the product they already use.
Content scripts
SPA-aware filtering and feed injection react to navigation.
Locale + audio
MV3 rules nudge German context and prefer original audio.
Local learning state
Goals, sessions, history, and vocabulary stay in the browser.
Dashboard + tests
Progress becomes visible; watched content can later become practice.
Chrome Manifest V3No OAuth, API key, or separate accountSPA-aware DOM observation + local browser state
Product decisions
The trade-offs are part of the product.
DecisionWhyTrade-off
Chrome extension, not a separate learning app
YouTube already contains the content and the habit.
The product depends on YouTube’s DOM and platform behavior.
Local-first, with no account or backend
Installation stays lightweight and viewing history does not need another server.
Progress does not yet sync across browsers or devices.
Curated starter feed before live recommendations
Strict filtering can leave a new learner with an empty homepage.
The team must maintain a high-quality channel and video catalog.
Deterministic launch demo
Live recommendations, captions, RSS, and watch history are timing-dependent.
The walkthrough proves product behavior, not live recommendation quality.
Evidence
What is validated, and what is not.
80
passing tests
Across 10 Vitest files covering feed data, difficulty, vocabulary, tracking, whitelists, and demo state.
177
captured videos
A deterministic snapshot from 53 real German channels, with original IDs and metadata.
Internal
testing stage
We can claim product testing, but we are not presenting adoption, retention, or revenue numbers yet.
Current limits
- YouTube can change its DOM and require selector maintenance.
- Normal YouTube watch history can still influence recommendations outside the extension layer.
- CEFR difficulty is an estimate based on available signals, not a certified assessment.
- Test Mode is a hardcoded premium-feature prototype, not launched functionality.
Business + launch
Free immersion first; structured practice later.
The core LAD extension is planned to remain free. The future paid layer is Test Mode: comprehension and exam-style practice based on videos the learner has already watched. The demo shows a hardcoded prototype, not launched functionality. No pricing or revenue claim is being made yet.
Launch plan
Chrome Web Store release planned after MM
- Package the installable release
- Complete permissions and privacy review
- Polish onboarding and first-feed recovery
- Expand testing beyond the founding team
- Prepare the Chrome Web Store launch after MM