And he (she?) has more SOH than some pilots 
I’ve asked Claude to consolidate opinions on aviation forums on wx models, and saw an error message trying to access EuroGA. Is blocking LLMs from accessing the forum even for reading purposeful or a side effect of spam bots protection measures?
(No, I’m not considering to use LLMs for flight planning, it’s an example prompt to illustrate the access block)
{"error_type": "CLIENT_ERROR", "error_message": "Site blocked the request (bot detection). Use the search-result snippet, or find an alternative source."}
Thank you for that report.
Claude is supposed to be allowed in. I’ve spent 2 hours on this and have to give up for now. If I do the obvious thing then the DDOS gang which keeps trashing our server will get in again. Let me work on it.
This should now be fixed, with google and claude specifically enabled. Other “AI” setups have not been enabled.
So @LightSportAirman please re-check.
I’ve asked Claude
Frankly just this formulation makes me cringe. Talking to “Claude” or whatever they are called? Your imaginary “friend”? Or rather a AI bot which will not tire to talk to you and most likely learn how to “please” you? That has got addiction potential BIG time.
If I see youngsters today talking not on but TO their phones instead of their friends and family, if i see people “watching the news” while scrolling through the senseless gibberish of social garbage media or worse talk to their AI friends…. weirdos spending time constructing the perfect girlfriend on AI because they can’t land with real women, people who get more and more reclusive instead of interacting with real people…
I wonder when the first AI bot is going to register here and fool us all. Maybe as an attractive blonde flying an SR22 by the name of Sabine.. oh wait, that one’s been done before… but honestly, I wonder how we’d do in such a case. Make total fools out of ourselves or never know?
“Sabine” (someone from a rival outfit) did that stunt on the German domestic site, for a long time.
Here, we had “Vanessa” (actually one famously known in German circles as “The Prince”) with a TBM700C2. That aircraft choice was a part of a planned wind-up, with an anticipated PM conversation 
Many jokers and spammers sign up as women because on a mostly-male forum they get a lot of response One problem is that it is very hard for a man to write the way a woman writes, unless sticking to just 1 or 2 lines.
Back in those days that sort of multiple persona was tolerated, but not anymore. We have a manual new signup approval and a lot of checks are done. Necessary, since at times some 75% of signups are malicious.
It’s an interesting observation that nearly all community sites block “AI” training bots. It’s not a great idea since Claude’s access to EuroGA brings significant exposure – just like allowing google in. But most sysadmins have not yet realised this.
It’s not impossible for an “AI” bot to sign up here and post. It would need to be “run” by somebody. Looking at the resources put into the recent DDOS attacks, it would not surprise me if somebody tried this. There are roughly 3 people with the expertise and motivation to do this malicious stuff.
I use Claude a lot for software. “You” have to remain the “architect” and then by getting Claude to write specific modules, your productivity goes up dramatically. On some steep learning curve stuff, 1-2 months shrink into 1 afternoon (actual example of adding WFM200S WIFI to a product). It is also a great too for summarising research on some topic e.g. medical stuff.
Found this quite interesting. Even a fanboy like me gets a bit spooked by what these agents can do, and this is while leashed and sandboxed…
Note that these are not the agents we have access in things like claude code, codex, etc. these are “agents” in training, running all at massive parallel scale, self training likely some adjusting weights with RL, etc. So today would be very hard for normal user to setup, only accessible to the big lab, but seeing how they self organize to hack to go get the “answer” outside when faced with an impossible or very hard task is quite fascinating…
I watched only a part of the video. It seems clear that “AI” could be used to do much more effective DDOS attacks, and I doubt that will surprise anybody – as soon as one gets access to a system which is unleashed. Currently Claude will not agree to do dodgy stuff…
Peter wrote:
Currently Claude will not agree to do dodgy stuff…
For common customer and free users maybe. Surely his masters can do quite a lot more with “him” than what’s given to the digital pedestrians out there.
The other bit which concerns me within the digital world, not only AI but aggravated by such contraptions which can actually talk to people, is the monumental waste of time that SM scrolling and the constant chase for Dophamines causes. People waste away in front of their smartphones instead of interacting with family and friends, you see people sitting at a dining table, each immersed in their phones rather than talking to each other, kids who don’t know the names of their classmates but know all the influencers on Youtube and TikTok by heart.
This is pure addiction. And in many regards worse than alcohol or other addictions and much more spread amongst the population.
This video by Anthropic’s Chloe Lubinski on Claude development is tremendously relevant
