Category: Blog
-
System Prompts – Claude Docs
See updates to the core system prompts on [Claude.ai](https://www.claude.ai) and the Claude [iOS](http://anthropic.com/ios) and [Android](http://anthropic.com/android) apps. — Read on platform.claude.com/docs/en/release-notes/system-prompts
-
Claude’s new constitution \ Anthropic
A new approach to a foundational document that expresses and shapes who Claude is — Read on www.anthropic.com/news/claude-new-constitution
-
10 things I learned from burning myself out with AI coding agents – Ars Technica
with LLMs, context is everything, and in language, context changes meaning. Take the word “bank” and add the words “river” or “central” in front of it, and see how the meaning changes. In a way, words act as addresses that unlock the semantic relationships encoded in a neural network. So if you put “checkerboard” and…
-
Trump airing Macron’s private message was designed to hurt and intimidate | Donald Trump | The Guardian
No president had ever achieved such omnipresence. In this fractured country, it gives him a singular power. It wasn’t money that propelled him but conversation. The creation of chaos is not an accidental byproduct of Trump, it is the method. Every void is filled with a provocation. He lights more fires than can be put…
-
Claude Code and Graphic Design
I asked Claude to tell me what it was good and bad at: What I’m Bad At (Don’t Ask Me) 1. Graphic Design / Visual Aesthetics My SVG logo attempt was a perfect example – technically valid SVG, but visually uninspiring and generic. The “wood grain” is just a gradient that looks nothing like actual…
-
A Visualization of the Evolution of Paris, 300 BCE to 2025
A Visualization of the Evolution of Paris, 300 BCE to 2025 — Read on kottke.org/26/01/a-visualization-of-the-evolution-of-paris-300-bce-to-2025
-
A quote from Boaz Barak, Gabriel Wu, Jeremy Chen and Manas Joglekar
One way to think of confessions is that we are giving the model access to an “anonymous tip line” where it can turn itself in by presenting incriminating evidence of misbehavior. But unlike real-world tip lines, if the model acted badly in the original task, it can collect the reward for turning itself in while…