1,749 mentions · 790 chunks · 131 episodes
Infinite content and LLMs can create filter bubbles but for your self-perception. Not only is it finding the content that makes you feel good about what you already believe,[n...
...illion permission dialogs. The stuff you'd use to contain the prompt injection (LLMs) is the stuff that can be tricked by anything you show. Turtles all the way down. "Do you trust this domain to get information from this chat?" This ...
...eason this isn't a big problem yet is that we're just in the tinkering phase of LLMs.
Tool calling is what makes LLMs turing-complete. Able to not just say, but do.
LLMs are so noisy that it's hard to figure out the quality of underlying components built on top, because the noise dominates the signal. Did it break bec...
... teaching method that forces students to have good answers. But in an era where LLMs make answers cheap, the real skill is asking good questions.
Anthea pointed out to me that LLMs should raise teachers' expectations of what students can accomplish. One approach to LLMs in education is worrying about the floor. "How would you ev...
...es of Getting Things Done is distilling a big thing into concrete next actions. LLMs can do that easily!
Hill climbing a moving hill doesn't work. LLMs are moving hills. The models are still improving rapidly. Don't over optimize for their current behavior.
LLM companies are trying to get a premature monopoly on LLMs. We didn't figure out the participatory architecture yet, which is necessary for the early stage of new technologies!
...the web, it formed a larger amount of the LLM context on Scots. That means that LLMs also likely replicate Scots poorly, all because of one weird bottleneck. A similar kind of thing happens in evolutionary biology, a "population bottl...
... the worst timeline." Imagine it for our entire digital lives. If everyone uses LLMs to cothink, the guardrails they have will shape all of society. This analysis shows the power of a centralized algorithm that everyone views the worl...
...n it easily. It's child's play to prevent injection with a bit of escaping. Now LLMs with tool use allow all data to be executable. A massive expansion of threat surface area. So now all of the systems builders are thrust into the wor...
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 ge...
...n adapt and maintain itself to continue working. This is only possible now with LLMs.
I liked this analysis of LLMs as normal technology. Normal as in electricity, the internet. General purpose, something you can take for granted.
The unlock for LLMs vs deep learning is they're general purpose. Deep learning techniques of the mid 2010's relied on supervised learning. They could do impressive feats...
LLMs are extremely confusable deputies. In security, one type of vulnerability is the confused deputy. A powerful entity is tricked into applying their po...
LLMs can make generalists almost as good as specialists in many domains. The generalist meta-skills of volition, savviness, curiosity are now more importa...
I like the metaphor of sleepwalking geniuses for LLMs. It captures how powerful they are… and also how silly they can be if you don't constantly guide them.