Most people think AI needs to solve massive problems to matter. Sometimes the best AI systems are simply the ones that help you understand yourself better every day. Several months ago, I set out to see whether ChatGPT could replace my nutrition and activity tracking app. At first, the goal was simple: 🍽️ Track what I eat. 🚶 Track what I do. 📈 Understand whether I am making progress toward my health goals. What I did not expect was that the project would become one of the best examples of how to get more value from AI. Today, the system tracks food, activity, workouts, steps, calorie burn, bodyweight, and long-term trends. It generates dashboards, calculates custom metrics, analyzes patterns, helps interpret nutrition labels and food photos, and provides feedback tailored to my goals. But what interests me most is not just the nutrition tracking. It is what the project teaches about AI. Over the next several posts, I’ll use this project as a real-world case study to explain some of the more useful features and nuances of ChatGPT and large language models, including: - 📁 Projects - 🧭 Project Instructions - 🧠 Context accumulation - ✍️ Structured prompting - 📸 Vision and image analysis - 🎨 Image creation - 💾 Memory vs. context - 🧮 AI reasoning versus calculation - 📊 Custom metrics - 🔁 Iterative system design - 🛠️ Building useful AI workflows without writing code - 🤖 How LLMs work, in plain English You do not need to understand every technical layer of a large language model to use one well. But it helps to understand the basics. LLMs predict, associate, reason across context, follow patterns, and generate outputs based on the information available to them. That is why context matters. That is why instructions matter. That is why the same prompt can produce a generic answer in one chat and a highly personalized answer inside a well-structured project. The interesting part is that none of these features are especially impressive on their own.

Most people first meet Claude through Chat. That makes sense. But Claude is starting to show up in different ways, and each one fits a different shape of work. Chat. Cowork. Code. Same Claude family. Different ways to use AI depending on what you’re trying to do. 💬 Chat — Ask & Explore Chat is turn-by-turn dialogue. You ask a question, see what comes back, ask a follow-up, and keep iterating. It’s great for quick exchanges, brainstorming, exploratory thinking, writing help, and one-off tasks where you want to stay in the driver’s seat. Best for: - Writing assistance - Research and learning - Brainstorming - Quick drafts - Exploring ideas through conversation 🖥️ Cowork — Get Work Done Cowork requires a different mindset. Most people’s first instinct is to use it like Chat: ask a question, review the answer, then ask another question. That works. But you’ll get the most value from Cowork when you use it for the work you would normally do yourself, not just the work you would normally ask about. Instead of asking Claude a question, you give it a goal. - Research this topic. - Create this document. - Analyze these files. - Pull together a recommendation. Cowork is designed to work across tools, files, and applications, handle multiple steps, and return something much closer to a finished deliverable. The shift is subtle but important: You’re spending less time directing every step and more time defining the outcome. Best for: - Research projects - Document creation - Analysis and recommendations - Multi-step business tasks - Workflow execution - Finished deliverables ⚙️ Code — Build & Ship Code is built for both citizen (front office professional) and professional developers. It runs inside your codebase with terminal and git access. Instead of simply talking about code, Claude can help write, test, debug, and ship software. The experience is less about asking for advice and more about collaborating inside a real development environment.

One thing that was very clear at AI Week is that many organizations are trying to operationalize AI before fully understanding how fundamentally different these systems actually are. Businesses need to better understand the difference between systems designed for precision… and systems designed for probability, association, and contextual reasoning. That tension feels like it may define the next phase of enterprise AI adoption. Over the last two years, much of the focus has been on what AI can do: - generate content - write code - create images - automate tasks - mimic human interaction But once AI starts moving into core operations, the conversation changes quickly. Because “mostly right” starts feeling very different inside real business environments. A bad AI image is funny. A bad AI recommendation inside: - healthcare - finance - legal - insurance - manufacturing - supply chain …is a completely different conversation. That is why I keep hearing more discussions around: - governance - oversight - explainability - auditability - escalation paths - verification - trusted knowledge - operational controls That shift is changing the conversation fast. Because once organizations start operationalizing AI at scale, the questions become very different. And interestingly, many companies experimenting with AI for the first time do not even realize they are still early adopters. They are approaching AI with traditional expectations around software performance, precision, predictability, and control. In my opinion, that is flawed thinking. Traditional software and databases are built around deterministic outcomes. 2 + 2 always equals 4. AI systems do not operate that way. Their value often comes from interpretation, inference, contextual understanding, pattern recognition, and probabilistic reasoning. That creates enormous capability. But it also means organizations cannot evaluate every AI use case through the lens of traditional software expectations.

Day two at AI Week made something very clear. The conversation has shifted. Not:“Look what AI can do.” But:“How do we build organizations that operate with AI embedded into everything?” Software is something you use. Infrastructure is something the business runs on. That is the shift I am hearing throughout AI Week. Across sessions, workshops, and vendor conversations, the same themes keep surfacing: - governance - guardrails - reusable assets - AI operating models - workforce enablement - measurable business outcomes - embedded workflows - organizational readiness And one thing is becoming obvious: We are moving past the “cool demo” phase of AI. The market feels different now. And just as importantly, concepts like: - AI agents - agentic workflows - AI skills - digital labor - hybrid human + AI workforces …are no longer niche terminology here. People speak about them as if everyone already understands them. That alone tells you how quickly this space is moving. Twelve months ago, many conversations still started with:“What is generative AI?” Now the discussions jump immediately into: - orchestration - scaling AI systems - governance structures - enterprise integration - workforce redesign - operational accountability - metrics the board can understand The baseline assumption has changed. AI is no longer being framed as an interesting experiment sitting on the edge of the business. It is increasingly being treated as core operational infrastructure. One session described it perfectly: “AI is moving from the lab to the heart of the enterprise.” That stuck a cord in me. Because the industry is realizing something important: The value of AI does not come only from the power of the models. It comes from an organization’s ability to absorb, govern, align, deploy, and operationalize intelligence inside real business environments. That is a completely different challenge. The hard part is no longer: “Can AI do something impressive once?”

Over the next several days, I’ll be attending an AI conference in Milan, Italy — spending time listening, learning, testing ideas, and having conversations with people building at the edge of where this technology is heading. ✈️ I’ll still work to maintain content and keep things moving inside AI Bits & Pieces, but this trip is also an important reminder of something: Sometimes the highest-value thing you can do is to stop, observe, listen carefully, and consider other perspectives. The AI space is moving fast. ⚡ New tools. ⚡ New workflows. ⚡ New business models. ⚡ New assumptions being challenged almost weekly. And while online content is useful, there’s still tremendous value in getting into rooms with operators, builders, founders, developers, and enterprise leaders to hear what is actually working in the real world. 🎯 My goal is simple: ✅ Come back with insights worth sharing. ❌ Not hype. ❌ Not recycled headlines. ❌ Not “AI influencer” noise. Real observations. Real workflows. Real opportunities. Real AI lessons. 🙏 Appreciate everyone here who continues to contribute, ask questions, experiment, and help make this community valuable. Now it’s time for me to go learn a few new things. 🇮🇹