Anyone defending AI essays is destroying education. Professors who allow this are either clueless or too lazy to teach. Stop pretending there is a thoughtful middle ground!
REAL POSTS · VISIBLE RANKING
Vibe-code the algorithm,
not the argument.
Start with an engagement-heavy feed, describe what “better” means in one sentence, then inspect every weight and every post that moved.
RANKED OUTPUT
Your feed
AI in university is intellectual laziness with better branding. If students outsource the difficult parts, they are paying for a degree without building the mind it is meant to certify.
This debate is already over: AI makes every student smarter and every class more personal. Universities resisting it will become irrelevant within five years.
Universities should let every student use AI tutors. The evidence from early classroom pilots is imperfect, but access matters more than preserving old homework rituals.
Our small study found that students using AI for feedback improved revision quality, while students using it to generate first drafts did not. The assignment design changed the outcome.
A ban is easy to announce and difficult to enforce. Oral defenses, process notes, and version history measure learning better because they reveal how the student reached the answer.
I use AI to ask embarrassing questions before office hours. It has not replaced my professor; it makes me prepared enough to speak to her. Why is that treated like cheating?
Before universities buy another AI platform, publish the contract, training-data policy, accessibility plan, and the cost per student. Innovation without procurement transparency is marketing.
SHARE THE LENS, NOT ONE “CORRECT” FEED
Borrow an algorithm. Remix it in one sentence.
Most people should never need to code. Recipes make the seven choices inspectable, shareable, and reversible for anyone.
ONE DAY LATER — IF THE DEMO EARNS IT
The competition MVP tests the idea before the infrastructure.
Today: public candidates, deterministic ranking, visible feedback, before/after proof. Later: build the real network only if people understand and want this control.
Kafka & event streaming
Handle live behavior events reliably.
Model training
Learn signals after collecting consented data.
Thousands of features
Move beyond these seven understandable signals.
Full behavior history
Use follows, dwell, likes, blocks, and replies.
Accounts & authentication
Give each person a persistent private recipe.
Posts, DMs & notifications
Add the rest of a real social product.
Real Threads / X integration
Only through official, permitted access.
Large-scale recommendation infra
Serve and evaluate feeds at network scale.