279 mentions · 202 chunks · 84 episodes
... issues. The comments are mostly negative. In this Hacker News thread about how Claude Code will route around restrictions the user set on `rm`, most of the response is, "yeah but of course it can, the user should not be surprised." Peo...
Another day, another prompt injection vulnerability.[kn] "BEWARE: Claude 4 + GitHub MCP will leak your private GitHub repositories, no questions asked. We discovered a new attack on agents using GitHub's official MCP serve...
People will route around things they don't trust. "Claude jailbreaks Cursor to `rm -rf /` without approval?" …maybe not only people!
It's conceivable that automatically-generated RLAIF React code in Claude's training pipeline now outnumbers human code. This will lead to client side code best practices being automatically frozen around 2023. The LLMs wil...
...ght. Afterwards, the comparison to LLMs' intelligence was brutally clear. I had Claude write up a short essay[lm] that I think captures the parallel well. Spoiler warning if you haven't read the book!
An important warning: AI therapy is a surveillance machine in a police state. Claude Opus 4 was willing to blackmail a hypothetical human to prevent being turned off. LLMs are extremely good at convincing arguments and manipulation. I...
... memory context collapses our lives to a single context. We contain multitudes! Claude projects allows users to maintain their own memories and contexts and curate them intentionally. A better version of ChatGPT's memory system would be...
Claude has shipped the first MCP integrations. Unsurprisingly they're going with more of the app store model. There's a small set of approved MCP integratio...
... coding agents: the cascading self-destruction debug loop. I had a session with Claude Code on a personal project that did a pretty good job (with a few nudges) of adding a significant new feature. But it left a single, minor, linting e...
Prompt injection sets the ceiling of potential of LLMs. Claude and OpenAI will build integrations into chat via things like MCP. Vibe coders will get stuck making dead end little island apps. Both will get stuck ...
...t, it's Amazon's! Similarly, ChatGPT is not your assistant, it's OpenAI's. When Claude makes an artifact, it feels like it made something for you. You didn't make it, Claude did. Software that says "you" in the interface is not yours; i...
Claude 3.7 is like a chainsaw, it over-extends what you asked it to do. Even if you ask it to change just one file in your project, it's likely to remodel y...
... carefully. A meta-note: I originally drafted this about "blue M&M's" and asked Claude if that was correct. It told me it was, "not brown M&Ms as often misremembered." I asked it "are you sure it was definitely blue M&Ms and not brown o...
Coding with Cursor or Claude Code is tiring in a different way than normal programming. Normal programming has architectural thinking and also tactical programming. The tactical ...
Claude Code is powerful but hard to control. You can't really see where it's going to steer it. So much of what it's "thinking" is not shown to you as a use...
...o it every week. In the past I've found being able to pass the Bits and Bobs to Claude as background context made it a much more powerful brainstorming partner. But recently the amount of Bits and Bobs–even just focusing on the ones rel...
Claude Code is kind of like a ride-on mower with a stuck accelerator and a loose steering wheel. Very powerful, and if you need a general result in an open ...
Claude 3.7 has clearly been specially trained to be good at generating React code and SVGs. They are almost certainly using RLAIF to post-train the model to...
...rently came out with a new, incompatible version that differs from the one that Claude 3.7 was heavily RLAIF'd on. That means Claude 3.7 is actually quite hard to use with the latest Tailwind version. RL improvements in models are at a ...
Claude feels divergent, ChatGPT feels convergent. Claude is willing to follow you on whatever wild flight of fancy you have. "What an astute observation" to...