1.12 — Workshop through-line: "The two brains" 🔥

Goal: feel, on your through-line track, the difference between rigid logic and generative reasoning. No code required — pen and paper are enough to get started.

🟢 In plain terms — you’ll play the expert system (write rules), see where they break, then turn the task into a natural-language instruction: your very first prompt.

🎯 The block below matches the track you chose on the home page. (You can change it there; with no choice, you see the “your own project” version.)

🧠 Track “Your documents copilot”

1. The task, in one sentence: “answer a question from my documents”.
2. 3 fixed rules FAQ-style: IF the question contains "hours" THEN answer "9am–6pm", etc.
3. 3 questions that break them: an unexpected rephrasing, a question straddling two documents, one whose answer isn’t in the FAQ.
4. Rephrase in natural language: “Answer this question relying only on the documents below; if the info isn’t there, say so.” 👉 Deliverable: the draft of your grounding prompt (picked up at Level 2).

📥 Track “Email sorter”

1. The task, in one sentence: “sort an incoming email into 3 categories (urgent / normal / junk)”.
2. 3 fixed rules: IF the subject contains "invoice" THEN → accounting, etc.
3. 3 cases that break them: an ironic phrasing, a typo, a message straddling two categories.
4. Rephrase in natural language: “Read this message and tell me whether it’s urgent, normal, or junk, and why.” 👉 Deliverable: the draft of your classification prompt (picked up at Level 2).

🎯 Track “Lead machine”

1. The task, in one sentence: “tell whether a prospect is qualified for my offer”.
2. 3 fixed rules: IF headcount > 10 AND sector = "construction" THEN qualified, etc.
3. 3 cases that break them: a weak signal (good fit but off-criteria), an atypical profile, an edge case.
4. Rephrase in natural language: “Here’s my offer and this prospect: are they relevant, and why?” 👉 Deliverable: the draft of your qualification prompt (picked up at Level 2).

📄 Track “Document extractor”

1. The task, in one sentence: “extract the amount, the date, and the vendor from an invoice”.
2. 3 fixed rules: IF the line contains "Total incl. tax" THEN it's the amount, etc.
3. 3 cases that break them: a vendor that changes format, a skewed scan, an unusual label.
4. Rephrase in natural language: “Read this document and give me the amount, the date, and the vendor.” 👉 Deliverable: the draft of your extraction prompt (picked up at Level 2).

✍️ Track “Content factory”

1. The task, in one sentence: “turn an idea into a post for a given channel”.
2. 3 fixed rules: IF channel = "LinkedIn" THEN 800 characters + 3 bullets, etc.
3. 3 cases that break them: a topic that doesn’t fit the template, another channel, a different tone.
4. Rephrase in natural language: “From this idea, write a LinkedIn post, professional and accessible tone.” 👉 Deliverable: the draft of your repurposing prompt (picked up at Level 2).

🎯 Track “Your own project”

1. Define your project’s task in one sentence. (e.g., “sort my incoming emails into 3 categories”.)
2. Write 3 IF…THEN rules as an expert system from the 80s would.
3. Find 3 cases that break your rules: an unexpected phrasing, a typo, irony… That’s exactly where the generative approach wins.
4. Rephrase the task in natural language (“Read this message and tell me whether it’s urgent, normal, or junk”). 👉 Deliverable: the draft of your first prompt, which we’ll work on at Level 2.

For techies (optional): compare your rule list to a real LLM call on the same 3 tricky cases. Note where each one wins and loses.