So true!
Let me look at your implementation of GiantClass
[read 100 lines]
Perfect! I see the core issue now. Let me now take the next 3 minutes to re-write the entire file from scratch with this totally not hallucinated fix…
@karpathy
-
Code Review and Implementation Fix Strategy
By
–
-
Why Is Chess Engine App Size So Large Neural Networks
By
–
How could chess engine be that giant? Very large local neural nets? Electron shennanigans? Too highres assets? What do you think @chesscom ? :p. Fwiw I couldn’t find other chess apps that were dramatically smaller so I ended up playing a bit just on the website.
-
2MB Chess App Shows Efficient Software Design Innovation
By
–
Very interesting. Love that it’s 2MB – I tried to download a chess app yesterday and it was 400MB.
-
Hallucinations in RNNs: Unreasonable Effectiveness Revisited
By
–
I believe this is true, I used the word in my “Unreasonable Effectiveness of RNNs” post from 2015, and as far as I can remember I also hallucinated it.
-
Model Convergence Through Distillation and Subliminal Learning Transfer
By
–
This model convergence is quite perplexing. Possibly related to recent results on subliminal learning? Basically deeper knowledge correlations transfer when training via distillation. As the amount of data online from LLMs increases, it’s possible this makes them converge to some
-
LLMs Beyond Chatbots: Rethinking AI Model Perception
By
–
Like! My personal top tweets are those that most people find confusing/wrong first and then obvious/self-evident a few years later. LLMs were at one point seen (and still are) as chatbot, like a better Eliza or so, hence the chat bubbles, which lacks imagination by about 1000X.
-
Using Cursor and Claude for AI-Assisted Development Workflow
By
–
I use CC from Cursor and I assumed most do as well (?).
I end up with a mixed thing where Cursor is the UI layer for reading the code, manual edits, tab completion and chunk edits, and CC for larger changes, architecting, Q&A. Still rapidly evolving though… -
Better Model for Analysis Claude Haiku Update
By
–
"Using a better model for analysis" I didn't realize I was using haiku all this time, no idea when claude code snuck this one in rofl.
-
Real-Time Video Diffusion Model MirageLSD Transforms Video Generation
By
–
Diffusion video models but now – **realtime**!
— Andrej Karpathy (@karpathy) 17 juillet 2025
Simple video filters are real-time but can only do basic re-coloring and styles. Video diffusion models (Veo and friends) are magic, but they take many seconds/minutes to generate. MirageLSD is real-time magic. Unlike simple video… https://t.co/34nBZ7Q1zRDiffusion video models but now – **realtime**! Simple video filters are real-time but can only do basic re-coloring and styles. diffusion models (Veo and friends) are magic, but they take many seconds/minutes to generate. MirageLSD is real-time magic. Unlike simple video