Based on the experience I have gained through testing, trial and error over the past few months, I have discovered that there are two types of Skills in Claude—or two types of GPTs in ChatGPT—that interfere with or undermine each other. What I was trying to do was build Skills or GPTs that would first explore certain topics with me and then generate documents, artifacts, analyses, or research. I now understand that the two tasks I was asking them to perform were antagonistic, and that their instructions were contaminating one another within the AI’s context. I have given them names to make the distinction clearer: - Dialogue Skills — discovery, facilitation, exploration, Socratic dialogue. - Production Skills — document creation, content tools, research, and analysis. The first type should not make assumptions, search the internet, or propose actions. It should only listen, summarize, and clarify. I see this as useful when I want to clarify what I am doing or thinking. In such cases, it is counterproductive for the AI to search online, make assumptions, or ask me to choose from options that it has generated itself. This type is also useful when I am developing my brand voice. The second type needs to search online for figures, statistics, research, and factual data in order to provide solutions. It is useful, for example, when I want to calculate TAM or SAM. What I discovered is that when I include instructions from both categories in the same Skill or GPT, they interfere with one another. I end up with neither the best dialogue and exploration nor the best research and analysis, precisely because the model is operating with contradictory instructions in its context. I tested splitting the process into two separate Skills or GPTs that run sequentially. The first, focused on dialogue, produces a document that then serves as the source for the second. The results were significantly better: a genuinely Socratic dialogue, unaffected by unsolicited proposals and recommendations—which are often tempting—followed by a much more factual analysis that required far fewer input requests from me.

New AI obligations come into force on August 2 (including for marketing) I’d like to share my own interpretation (as I currently understand it) of the new legislation, for those who may be interested. If we use ChatGPT, Claude, Gemini, or other AI tools to create text, images, videos, or marketing materials, we may fall under a new set of legal obligations starting August 2. A new phase of the European Union’s AI Act comes into effect, introducing transparency requirements for organizations that develop or use AI systems in their professional activities. What matters for us? - If you use AI to generate or modify content, in certain situations you will be required to inform users that the content was generated or modified using artificial intelligence. - Deepfake content must be clearly labeled. - If AI-generated text about matters of public interest is published without meaningful human editorial review and oversight, additional transparency obligations may apply. - Responsibility does not lie only with AI developers (OpenAI, Google, Anthropic, etc.), but also with organizations that use these systems professionally. In other words, with us. At first, I thought: “That’s fine. It only applies to deepfakes. I don’t create those.” But after looking into it more closely, I realized that if I use AI to generate a video in which a customer appears to speak or appear in a way that never actually happened, that content is very likely to fall within the definition of a deepfake and therefore be subject to transparency requirements. The AI Act requires AI providers and online platforms to implement mechanisms for identifying AI-generated content. As a result, we may soon see authentic-looking content on social media being labeled as AI-generated based on various detection methods. Companies such as Meta have never been particularly transparent about how their algorithms classify, label, or curate content, so it will be interesting to see how these mechanisms evolve.

I don't know if you've noticed, but both Anthropic and OpenAI seem to be launching new programs every couple of weeks. The difference is that these programs are no longer what we used to expect: new AI models or new capabilities. Instead, they're packaging existing AI models into commercial offerings. To me, this is yet another sign that the industry's focus is shifting from technology development toward marketing and sales. As I mentioned in another post in this group, I believe we're moving—gradually, yet very quickly—from the "romantic" era of AI, where access was almost free, into an era where AI is increasingly priced based on usage. That's exactly why I think what Olivia and Ada are doing for us here couldn't come at a better time. It's much better to understand the landscape before spending significant amounts of money building agents, applications, or AI artifacts. Personally, I'm putting a lot of energy into moving in this direction. For those who may not be subscribed, I'm sharing OpenAI's latest offer below. You might find it interesting. As a side note, I've also attached a screenshot of their disclaimer—just so you can see how tiny and practically unreadable the text is in a document that can't even be zoomed unless you copy it somewhere else. That's exactly the kind of marketing practice I don't like. The full disclaimer text is included below. Introducing the ChatGPT for Small Businesses Program A new program bringing together AI tools, education, events, and resources for small businesses. Sign up for updates

I'm so glad you're all here! This is a space for experts and solo business owners on the journey to building a complete AI-powered marketing department: one that runs without you being the bottleneck every single day. Here's how to get started: 👉 Step 1: Introduce yourself below: tell us who you are, where are you from, what you do, and what is your favorite AI platform or tool. BONUS POINTS for sharing a picture of your workspace! We want to see all that messy clutter on your desk, do not hide it! 😊 👉 Step 2: Head to the Classroom and start The Strategy Foundation module -> five free SUPER IMPORTANT lessons, five concrete deliverables, the groundwork everything else is built on 👉 Step 3: Explore the community: check out AI Tools + Resources, ask questions, share your wins, however small I'll go first. I'm Olivia Angelescu. I've spent 20 years in marketing, sales, and business and I've watched too many brilliant experts stay invisible because they had no system behind them. Not because they lacked talent. Because they lacked infrastructure. Heck, I was one of them! I built an entire methodology and community first in a smaller, local market - proved it worked - and now I'm rebuilding it here, for the English-speaking market. I'm not "teaching" from a comfortable distance. I'm doing exactly what you're doing, right alongside you. What I'm building here is bigger than a community. It's a movement of experts who refused to become a statistic: the 35% who made it because they built the (unsexy) systems that most people skip. That starts with strategy. Then we build the rest together. Can't wait to see what you build here. Olivia Founder, AI Entrepreneur Institute P.S. You'll find my workspace image in the first comment below! 😂

Hello, When we work in marketing, sometimes we work with precision, and other times we work creatively. For example, we work rigorously when we talk about our values, our policies, terms and conditions, pricing offers, or budget calculations. But we also work creatively when we come up with campaign names, write marketing copy, look for ways to differentiate our content from competitors, or create marketing materials. When we do this with AI, we face a temperature problem. I'm not referring to the temperature outside, but to the temperature of the AI model we're working with. Temperature is the level of rigor or creativity with which an AI model responds. It is a technical parameter set by the provider (Anthropic, OpenAI, Google), typically somewhere around 0.7–1 on a scale from 0 to 2—that is, somewhere in the middle. In other words, neither too creative nor too rigorous. If we're using a chat interface, we usually don't have the option to adjust this parameter because the provider doesn't expose it as a configurable setting. However, when using an API to connect an application and a database to the model, this parameter can be set. That's valuable, but it's generally intended for advanced users. A second, much simpler approach is to try to express the desired temperature in the prompt itself. This won't actually change the parameter, but it can influence the model's behavior in the desired direction. For example, if I want a legal recommendation, I prefer to ask for it "at a temperature of 0.2." If I want a social media post, I ask for it "at a temperature of 1.5." The results can be remarkable. I'm attaching two screenshots where I experimented with ChatGPT using the same model and the exact same prompt. @Olivia Angelescu and @Ada Munteanu , I'm curious whether you've found a solution for this, whether you've experimented with it yourselves, and whether, when working with prompts and AI applications, you think it's worth paying attention to temperature as well.