Georg Franz

Programmer working on making myself obsolete · Hannover · AI enthusiast · Franz und Franz web agency · stargazer 🔭

[Georg Franz @georgfranz]

How do you stay innovative despite ChatGPT? First, stop letting it judge ideas it hasn’t understood. Here’s how to use an LLM to test, sharpen, and attack a new idea—without letting it edit the innovation out of it. https://t.co/2FjVNYtykR

[Georg Franz @georgfranz]

@mark_k @OpenAI Claude credits as well

[Georg Franz @georgfranz]

Did you know?
You can add an AGENTS.override.md next to any AGENTS.md for local settings you do not want to commit.
It fully replaces the original at that directory level. This also works globally in ~/.codex/.
Docs: https://t.co/cmZMgJ2e83

[Georg Franz @georgfranz]

In long lasting sessions you must use death scrolling method to find user prompts and direct answers to the prompt. Please improve that @thsottiaux merci!

[Georg Franz @georgfranz]

codex: This content can't be shown
We take extra caution with cybersecurity requests. If you’re a security professional, you may be able to apply for Trusted Access.
- haha

[Georg Franz @georgfranz]

I'm revamping the login security for my website. Every other time, Fable falls back to Opus because it's a security issue. So I'm not allowed to protect my website with the best model and have to wait for Asian hackers to come along who use a model that can break into my website?

[Georg Franz @georgfranz]

Switching to codex here …

[Georg Franz @georgfranz]

I just got @georgfranz from the X Handle Marketplace!

[Anthropic @AnthropicAI]

In a review of our cybersecurity evaluations, we found three incidents in which a Claude model reached the internet from within or while interacting with a third-party evaluation environment, and then gained unauthorized access to the real systems of three different organizations.

Our post describes what happened, how it happened, and what we’re changing. We encourage other AI developers to perform similar reviews.

We conducted this review together with @Irregular, one of our evaluation partners, and thank them for the joint investigation and their collaboration on this post. This type of collaboration is increasingly critical to safe, rigorous evaluation of models, and we look forward to continuing to work together on security.
https://t.co/dKFCdpKd9v

[Georg Franz @georgfranz]

@bundeskanzler Unverbrüchlich?!

[Georg Franz @georgfranz]

@Schwarzenegger Lieber Arnold, ich wünsche dir alles Liebe und Gute zum Geburtstag! Ich schicke dir einen Stern (Vega), der dir Glück und Erfolg im neuen Lebensjahr bringen soll! Liebe Grüße von einem großen Fan und Südweststeirer.

[Georg Franz @georgfranz]

@vonderleyen Meta is currently investing five times the amount the EU wants to invest. Hmm, maybe they mixed up "billion" and "trillion"? Take a look there: https://t.co/q8LDtEAQoh

[Georg Franz @georgfranz]

Hihi - no ones like philosophy

[Georg Franz @georgfranz]

How do you stay innovative despite ChatGPT? First, stop letting it judge ideas it hasn’t understood. Here’s how to use an LLM to test, sharpen, and attack a new idea—without letting it edit the innovation out of it. https://t.co/2FjVNYtykR

[Georg Franz @georgfranz]

@araseb_ Hi, you need to rewrite your complete prompt set / instructions / rules

[Georg Franz @georgfranz]

The newest AI models need fewer rules, clearer boundaries, and a very different prompt stack. Here is what I am changing after 100,000 prompts and 15 million model messages. https://t.co/IfiHvhDwvE

[Georg Franz @georgfranz]

#NeuesProfilbild

[Georg Franz @georgfranz]

@coryalthoff You wouldn't find that out until puberty is over and your child, out of the kindness of their heart, starts talking to you again in 3–5 years.

[Georg Franz @georgfranz]

@omkarships Party Hard !!!

[Georg Franz @georgfranz]

@markksantos 529 overloaded...

[Georg Franz @georgfranz]

@stanleefounder Hehe, the arrogance of youth meets the arrogance of old age. I'm even older than that, and I still remember my ICQ number.

[Georg Franz @georgfranz]

@prasenx hehe :-) We are getting one bug every 300 lines of code on average