The leader track 🤝 · for C-level, boards & managers · 6 x 45 min · no tech required
Prompting as delegation, not magic words. The thinking mode: ask the right question, think with the machine, judge what comes back, and work with your data and tech people to get ideas realised. Also a friendly on-ramp for LinkedIn readers.
Prompting is delegation, not magic words
The new-hire mental model, the golden rule that catches weak asks, the 4D skill map, and a live makeover of a real request.
🎯 Leader session 2Ask the right question
The five jobs you can give AI, the two principles behind every strong ask, and how to diagnose a weak answer by its missing ingredient.
🧠 Leader session 3AI as thought partner
Brainstorm properly, make AI attack your plans, dodge the sycophancy trap, and know when it knows vs searches vs reasons.
🔍 Leader session 4Judge the output
Why it guesses so confidently, the discernment habits that catch inventions, and matching your checking effort to the stakes.
🚀 Leader session 5From idea to shipped
Specify like a PM, understand evals as acceptance tests, and the five questions to ask any team that demos you an AI feature.
🛡️ Leader session 6Risk, safety, strategy
Injection vs jailbreak in plain words, the lethal trifecta test for any AI system, and who owns your prompts as they become assets.
The builder track 🛠️ · for practitioners · 10 x 45 min · any model or API
For DA, DE, DS, and engineers. One running build - TriageBot, a support-ticket triage prompt system - hardened session by session: techniques, structured output, reasoning models, long context, agents, vendor deltas, evals, and security.
The 2026 map: where prompting fits now
Success criteria before prompts, prompting inside context engineering, and the model-class decision tree that rewrites 2023 advice.
🧱 Builder session 2Core techniques
Clear and direct, delimiters and structure, templates with variables, few-shot examples, and the honest 2026 role of roles.
📦 Builder session 3Output control
Format matching, schema-enforced structured output, verbosity control, and the prefill deprecation with its five migration patterns.
🤔 Builder session 4Reasoning: CoT to thinking models
The 2022 papers as lineage, the great inversion (no more "think step by step"), and effort parameters as the new dial.
📚 Builder session 5Long context, grounding, caching
The placement wars between vendors, document tags and quote-grounding, RAG vs grounding, and cache-shaped prompt layout.
🤖 Builder session 6Chaining + agentic prompting
Prompt chains and self-correction, the three agentic reminders, eagerness calibration, and tool descriptions as the forgotten prompt.
⚖️ Builder session 7Vendor deltas masterclass
Claude, GPT, and Gemini side by side: terminology, model-specific pages, the sampling-params history, and the portability checklist.
📏 Builder session 8Evals I: from vibes to numbers
Why looks-good-to-me fails, the grading taxonomy from exact match to LLM-judged, and your first eval harness with promptfoo.
🧪 Builder session 9Evals II: judge, CI, agent evals
LLM-as-judge done properly, regression gates on every prompt change, and the router-skill-trajectory taxonomy for agents.
🔐 Builder session 10Security + production
Injection vs jailbreak, the lethal trifecta, why defense is architecture not wording, and prompts as versioned, gated code.
Choose your path 🗺️
Leaders start at Leader session 1; practitioners start at Builder session 1. The fast-tracks fit a specific gap.
The knowledge map 🧠
The builder track at a glance - hover a session to spotlight its concepts, click any node to jump in.