I also think it's interesting that @karpathy 's style preference seems quite different from my own! I actually prefer complex lines a lot of the time, because the intermediate variables introduce more free choices and spread things out more. I have to look to see if the variable is reused later. Obviously there's a limit and dense lines are often pretty bad in ML code, but I definitely wouldn't have a "one op per line" rule in my style guide.
AI
-
AI Projects: Common Failures and Success Strategies
By
–
The Expensive Hobby Mistake: Why #AI Projects Fail and How to Succeed
by @Khulood_Almani #MachineLearning #ArtificialIntelligence #ML -
Iterative Improvement Over Single-Pass Instructions for AI Agents
By
–
You can't expect instructions like "Every line should do one thing" in an AGENTS.md to work. You need multiple passes. Trying to front-load all the style advice is like setting a really high learning rate. The space of possible code solutions is very large. Rather than trying to get there in one step, it's better to make lots of little steps that reliably iteratively improve the solution. Andrej Karpathy (@karpathy) I'm not very happy with the code quality and I think agents bloat abstractions, have poor code aesthetics, are very prone to copy pasting code blocks and it's a mess, but at this point I stopped fighting it too hard and just moved on. The agents do not listen to my instructions in the AGENTS.md files. E.g. just as one example, no matter how many times I say something like: "Every line of code should do exactly one thing and use intermediate variables as a form of documentation" They will still "multitask" and create complex constructs where one line of code calls 2 functions and then indexes an array with the result. I think in principle I could use hooks or slash commands to clean this up but at some point just a shrug is easier. Yes I think LLM as a judge for soft rewards is in principle and long term slightly problematic (due to goodharting concerns), but in practice and for now I don't think we've picked the low hanging fruit yet here. — https://nitter.net/karpathy/status/2035173492447224237#m
-
Live speech translation with Gemini eliminates language barriers
By
–
This is live speech translation, powered by Gemini, rolling out inside the Google Translate app on Android. The takeaway for leaders is simple.
The language barrier is no longer a strategy constraint. Watch the full video to see how it actually works, then tell me. Don't miss -
Real-time conversation translation with speech preserving tone and rhythm
By
–
ere’s what stopped me mid-scroll. This isn’t text translation.
This is live conversation. → Speech to text in milliseconds
→ Translation in real time
→ Spoken back to you, with tone and rhythm intact It feels less like software and more like a second brain. -
Real-time multilingual conversation mode eliminates translation barriers
By
–
The breakthrough is conversation mode. → You speak your language
→ They speak theirs
→ Everyone hears replies in their own language No pauses.
No awkward repeats.
No “lost in translation” moments. That’s a massive shift for travel, negotiations, and crisis scenarios. -
Google removes language barrier in global business meetings
By
–
We’ve spent decades optimizing global business, except for the biggest bottleneck of all.
— Ronald van Loon (@Ronald_vanLoon) 21 mars 2026
Language.
Now imagine walking into a meeting where no one switches languages, and no one slows down.@Google just removed that friction. pic.twitter.com/fs57SdHs59We’ve spent decades optimizing global business, except for the biggest bottleneck of all.
Language. Now imagine walking into a meeting where no one switches languages, and no one slows down. @Google just removed that friction. -
Proving AI Security: The Challenge Beyond Attack Prevention
By
–
"Solved" implies that it's not possible to craft an attack that works. I'm not convinced that's true, even as attacks get harder The big challenge is proving you can't get an attack through – I'm not sure how you'd do that
-
Robot Performs Spider-Man’s Spectacular Stunts in Film
By
–
Spider-Man’s Jaw-Dropping Stunts Are Pulled Off by a #Robot
— Ronald van Loon (@Ronald_vanLoon) 21 mars 2026
by @lukas_m_ziegler#Robotics #Engineering #ArtificialIntelligence #Innovation #Technology pic.twitter.com/m4KUg14c9NSpider-Man’s Jaw-Dropping Stunts Are Pulled Off by a #Robot
by @lukas_m_ziegler #Robotics #Engineering #ArtificialIntelligence #Innovation #Technology -
First Criminal Case of AI Streaming Fraud in 2026
By
–
-> First criminal case of streaming fraud by AI. And it's completely insane. A North Carolina man used AI to generate hundreds of thousands of songs. He put them on Spotify, Apple Music, Amazon. Then he botted billions of streams on his own tracks. -> 660,000 fake streams per day. Distributed across thousands of titles so nobody noticed anything. -> $1.2 million per year. For music that no human has ever listened to. -> Result: $8 million pocketed. Meanwhile, real artists struggle at $0.003 per stream, do promo on TikTok, beg for playlist placements. Him? AI made both the music AND the audience. He got caught. He has to repay the $8 million. But this problem goes beyond him: The music industry spent 10 years fighting piracy. Now it has to fight songs that don't exist, listened to by people who don't exist. And since this case, AI has only gotten better. The playbook is public. Tomorrow's fraud will be even harder to detect. AI doesn't just change creation. It also changes cheating. [Translated from EN to English]
→ View original post on X — @flashtweet, 2026-03-21 09:03 UTC