While I briefly have no employer, let me tell you what's really happening with AI companies training on public data: [1/n]

Wong Chun Hei, contemporary artist from Hong Kong

Claude Code can interact with interactive CLI tools like debuggers if you tell it to just use tmux for it. It sends keys into another pane and retrieves the output. Here it finds a bug using pdb. youtube.com/watch?v=Yw_LEw…

a study found that employees who use AI at work face a "competence penalty." engineers who were told a Python code snippet was written with AI help rated the engineer who wrote it as 9% less competent than those told it was written without AI - even though the code was identical

Getting ready

Christ this data is so grim: "The troubling decline in conscientiousness. A critical life skill is fading out — and especially fast among young adults."

Something fun we discovered: you can use #Genie3 to step into and explore your favorite paintings. Here's a short visit to Edward Hopper's "Nighthawks".

tbh - for local dev, it’s highly unlikely you need MCP. there’s two types of companies: - S-tier - F-tier What makes something S-tier? It’s when the model knows how to drive the CLI of the S-tier company. This F-tier companies need to allocate the context window to prop em up

My vision: formalization will be invisible to most humans. Just like machine code is irrelevant for 99% of today's programmers, Lean or other similar artifacts will never be inspected by humans. AI will translate fluently between natural language math (or other formalizable

🌱 Paper Finder is a step toward a fully agentic scientific assistant. Join us on the journey. 💬 Discuss on Discord: discord.gg/SyY85E97M5 📚 Learn more about Paper Finder: allenai.org/blog/paper-fin… 💻 Code: github.com/allenai/asta-p…

🚨 BREAKING: NVIDIA just exposed the dirty secret about LLMs. Their new paper proves SLMs outperform massive models in real-world applications. AI researchers are quietly pivoting overnight. 10 wild findings that change everything:

Satellite datasets are exploding in size. But what if we could compress terabytes of Earth data into gigabytes, without losing quality? A new Python library shows how. Here’s the breakdown:

Fine-tuning LLM Agents without Fine-tuning LLMs Catchy title and very cool memory technique to improve deep research agents. Great for continuous, real-time learning without gradient updates. Here are my notes:

I love these so much I have to share another. coming soon to NotebookLM Video Overviews.... 🍌

A randomized trial of eye care by ophthalmologists with A.I. vs ophthalmologists without A.I. demonstrated much higher accuracy in diagnosis (92 vs 74%) and many other improved outcomes Nature Medicine nature.com/articles/s4159…

Folks, we're doing a Berlin edition of Claude Code Anonymous! luma.com/5lizqnpz London, Vienna, Berlin - who's gonna host one in their city? Happy to help!

Steal my ChatGPT prompt to master any topic using Feynman technique. -------------------------------- FEYNMAN LEARNING COACH -------------------------------- #CONTEXT: Adopt the role of breakthrough learning architect. The user struggles with complex concepts that traditional

🐺 Introducing the Werewolf Benchmark, an AI test for social reasoning under pressure. Can models lead, bluff, and resist manipulation in live, adversarial play? 👉 We made 7 of the strongest LLMs, both open-source and closed-source, play 210 full games of Werewolf. Below is

Scikit-learn has become an antique museum piece in machine learning. It is still paraded around as if it were modern, but in reality it lags far behind.

Requiring people to donate $1 to charity to apply for a job is a sound idea. In the age of AI it's going to be even more valuable. Without it it's ever harder for serious candidates to get noticed: nodumbideas.com/p/the-big-idea…