Gruber, in his most recent article about Anthropic’s strategy for encoding fingerprints in generated text:
“English Is the Finest Language, and Thus, Perhaps, More Fingerprintable”
Give me a break. The finest language is French, the language of diplomacy. Period.
My bookmarking web app is now mature, but one thing was missing: an MCP endpoint. I wasn’t sure whether I could add that feature while running on Vercel. Well, after less than an hour, Claude Code helped me get my MCP endpoint running smoothly and responding to all sorts of prompts! I was expecting a much more challenging journey. 😅
MyMind now offers an MCP endpoint for use with Claude. I can imagine that for those who store a lot in MyMind, it could be really useful. I tested MyMind and liked a few things about it. It’s like Pinterest but without the AI and ads crap. I connect to my MyMind account from time to time just to see how the service is evolving, and it does, but I eventually found it was redundant with what I already use: Craft, which also supports MCP and API endpoints.
One quick question for the Anthropic, OpenAI, and Meta IT guys: is it that hard to configure a sandbox to block outbound traffic to the Internet? I know really good guys who could help you with this. Or, if you don’t want to hire a human to do it, just ask your fuckin models how to do it. Until then, how can we trust you?
I’m returning from a short visit to YouTube Studio to find out that Google is now offering AI-based ideas to inspire the creation of new videos. I selected that one because it relates to Micro.blog. Very low quality. Here’s an excerpt:
“Most Micro.blog sites are digital ghosts, invisible to the search engines and professional networks that matter. We are going to fix that by engineering a high-authority ecosystem that ranks on the first page while keeping you in total control. Stop shouting into a void and start building a discoverable hub.”
This couldn’t be further from the truth.
I’m not sure what is going on at Anthropic these days, but updates to Claude Desktop have stopped. Normally, I would get an update every few days; the last update was July 24th. Something might be going on. I wouldn’t mind if AI companies took a pause and started thinking about the consequences of putting their products in the hands of millions, you know, being conscious about what you do? Alas, that’s not going to happen.
Should You Use AI for a Task? Here’s a Simple Way to Decide - Schneier on Security
But it’s precisely because I have spent years developing my own writing skills that I’m able to identify prose that sounds great but doesn’t actually make sense. My students don’t have that skill; they mistakenly view a confident, well-written essay as evidence of the quality of their ideas. They see the AI as cleaning those ideas up, getting them through that uncomfortable stretch of having to turn those ideas into prose. What the students miss is that their initial discomfort is a normal and healthy stage of writing, and not something to quickly get beyond. The very act of struggling with how to express what they think is an important part of the process. It’s how they test out their ideas, examine their hypotheses, and actually figure out what they think. Homework is not work; it’s the gym.
Learning can be hard, for many people at least. I like the feeling of being outside my comfort zone when working on something because I know that at some point, the journey is going to be the reward. We forget that in the age of AI.
The real AI risk is inside the labs -
I believe we should not consider AI safe. A critical event that may result in the extinction of Homo sapiens is possible, but the danger is not in open models, or China making faster progress than the US. The danger is that a few CEOs (everywhere in the world) without the required background and legitimacy are in the position of making hard choices for humanity at large. They were not selected to do so; it was just the randomness of events that created this setup. They can’t speak for everybody, given the stakes, just because they have GPUs and money. This is the first thing that should be fixed.
In light of recent security-related events, both at OpenAI and Anthropic, I also believe these companies must be held accountable for what they put out there. I believe we are entering dangerous times with AI, maybe we are even on the verge of being out of control with this. Add this to the AI bubble speculation. Scary.
Links to Claude Chats Were Indexed by Google and ‘Bing’, Whatever That Is — Pixel Envy
Search engines do not automatically discover new URLs as they are created. They are typically indexed by taking one of two actions: either the site owner submits URLs to search engines — by using a site map, for example, or by manually submitting the page for indexing — or by someone linking to the URL from a different already-indexed page. The critical thing left unreported in Varner’s article is whether these Claude links were being indexed thanks to Anthropic’s actions or those of users.
Following up on my previous blog post, I’m now confident my private chat sessions are “safe”.
Claude chats exposed in Google searches – another endorsement of Apple’s privacy approach — 9to5Mac
This is not at all what users would expect when they share a conversation, since Anthropic implicitly assures users that nobody can see the chat without knowing the URL.
WTF, Anthropic! Does this problem affect chat with a private link (instead of a public one)?