YOU ARE VIEWING ONE ITEM FROM THE AICRIER FEED

Karpathy's MenuGen collapses apps into features

AICrier tracks AI developer news across Product Hunt, GitHub, Hacker News, YouTube, X, arXiv, and more. This page keeps the article you opened front and center while giving you a path into the live feed.

// WHAT AICRIER DOES

7+

TRACKED FEEDS

24/7

SCRAPED FEED

Short summaries, external links, screenshots, relevance scoring, tags, and featured picks for AI builders.

Karpathy's MenuGen collapses apps into features
OPEN LINK ↗
// 51d agoNEWS

Karpathy's MenuGen collapses apps into features

In a fireside chat at Sequoia Ascent 2026, Andrej Karpathy used MenuGen as a concrete example of why LLMs are more than just a coding productivity boost. MenuGen is a vibe-coded app that turns a restaurant menu photo into generated dish images, but the larger point from the talk is that model-native multimodal experiences can make some standalone apps feel unnecessary once the base model can do the same job directly.

// ANALYSIS

Hot take: MenuGen is a neat demo, but the real story is that the moat for many “LLM wrapper” apps is getting thinner as foundation models absorb the core workflow.

  • The product is memorable because it sits at the intersection of vision, generation, and utility, not just text completion.
  • Karpathy's example supports the thesis that LLMs create new product categories, but also erase weak ones fast.
  • If a base model can read, reason about, and visually transform the input itself, the app layer needs deeper workflow value or distribution to survive.
  • The talk is more important as a signal about where multimodal software is heading than as a launch announcement for MenuGen itself.
// TAGS
llmsmultimodalvibe-codingai-appssequoiakarpathy

DISCOVERED

51d ago

2026-04-30

PUBLISHED

51d ago

2026-04-30

RELEVANCE

7/ 10

AUTHOR

karpathy