From AI Talk to AI Walk: How 24 People Harnessed 4 AI Superpowers in One Night

Let’s be honest. Out of 100 people talking about AI, only about 10 are able to use it effectively every single day—matching the right tool to the right task.

The rest? They’re stuck. Overwhelmed by options. Frustrated by vague outputs. Or just copying prompts that don’t quite work.

Then there’s the other reality no one talks about: in the tech event space, without a giant role model in the room, getting 20+ attendees feels like winning a lottery. Cross your fingers. Say a prayer.

But last night? 24 attendees. One Google Meet. And a whole lot of AI firepower.

We ran the AI Sprint I’d posted about earlier—and the energy was electric.

Here’s exactly what we learned, what we built, and how you can start using AI like the top 10%.

The 4 AI Superpowers That Change How You Work

We started by unpacking four core capabilities every professional needs. Think of these as your AI utility belt:

• Summarise – Turn long documents into sharp, actionable insights.
• Transform – Change tone, format, or even language in seconds.
• Brainstorm – Generate dozens of ideas, then filter to the best ones fast.
• Execute – Write, code, or build—faster than ever before.

These aren’t tricks. They’re workflows.

The Prompt Formula That Separates Magic from Meh

You’ve heard “prompt engineering” thrown around. Here’s the actual formula we used:

Role + Task + Context + Format

That’s it. Get these four right, and your outputs go from generic to genuinely useful.

But we didn’t stop there. We went straight into advanced moves:

• Chain-of-Thought – forcing the AI to show its reasoning step by step.
• Red-Team Critique – asking the AI to tear its own answer apart.
• Structured Output – demanding JSON, tables, or templates, not paragraphs.

These techniques turned hallucinations into helpful answers.

Guardrails First: Because AI Is a Tool, Not a Boss

Before touching any tool, we set non-negotiable rules:

🔒 No sensitive data – ever.

✅ Always verify facts – especially numbers and names.

🧠 Human decides, AI suggests – you stay in the driver’s seat.

With those in place, we got hands-on.

Tools We Used (And What They Did for Us)

Each tool served a different purpose. Here’s the breakdown:

Claude & Claude Code

• Set up project instructions, artifacts, and connectors
• Used “skills” to maintain context across long conversations
• Perfect for coding and structured technical work

ChatON.ai

• Acted as a study coach for personal and academic life
• Helped plan study sessions and break down complex topics

Google Gemini

• Created an Excel sheet from a presentation estimate
• Exported it to Google Sheets in 3 seconds
• Real-world spreadsheet automation, no formulas needed

NotebookLM (by Google)

• Uploaded our session notes
• Generated:
  • An audio podcast discussion
  • An interactive video like a real lecturer
  • Flashcards and auto-generated study notes
• One upload, four learning formats

ChatGPT

• Applied the prompt formula + chain-of-thought
• Drastically reduced hallucinations and confident wrong answers
• Showed that even the most popular model needs technique

Kimi (Moonshot AI)

• Created presentation slides collaboratively
• Checked docs, sheets, and websites simultaneously
• Demonstrated visible agent control – you see and guide what the AI does

The Bottom Line

In one 150-minute session, 24 people went from “curious about AI” to confidently using four superpowers, a repeatable prompt formula, and seven different tools—each for the right job.

That’s the difference between talking about AI and actually using it.

Best Practices for Using AI—and Using It Responsibly

The sprint wasn’t just about speed. It was about building the right habits from day one. Here’s what we agreed every AI user should live by:

1. Always Start with a Clear Goal

Don’t open an AI tool and “see what happens.” Know what you need before you type. A focused prompt with a clear objective will always outperform a vague, open-ended one. The Role + Task + Context + Format formula exists for exactly this reason.

2. Treat AI Output as a First Draft, Never a Final Answer

AI is brilliant at generating ideas, drafting content, and automating repetitive work. But it can also hallucinate facts, invent citations, and present wrong information with total confidence. Always review, verify, and refine before you ship anything.

3. Protect Sensitive Information

Never paste passwords, API keys, personal data, or confidential business information into any AI tool. Most cloud-based models process your input on remote servers. If you wouldn’t post it publicly, don’t paste it into a prompt.

4. Use the Right Tool for the Right Job

No single AI tool does everything well. We used seven tools in one session because each had a strength: Claude for coding, Gemini for spreadsheets, NotebookLM for learning, Kimi for slides. Matching the tool to the task is what separates power users from everyone else.

5. Stay in the Driver’s Seat

AI suggests. You decide. Never let a model make final calls on important decisions—whether it’s publishing content, sending an email, or deploying code. The human in the loop isn’t optional; it’s the whole point.

6. Learn the Technique, Not Just the Tool

Tools change every month. Techniques last for years. Chain-of-thought prompting, red-team critique, and structured output work across ChatGPT, Claude, Gemini, and whatever comes next. Invest in the skill, not the subscription.

7. Share What You Learn

AI literacy shouldn’t be gatekept. If you discover a workflow that saves you hours, teach someone else. That’s exactly what this sprint was about—24 people levelling up together, not in isolation.

Let’s keep building. Faster. Smarter. Responsibly.

Special thanks to the 24 attendees who showed up with energy, sharp questions, and a willingness to sprint: Alex Nyambura, Erick Mwangi (outgoing GDG On Campus University of Embu Lead), Charles Muthui, Philip G, Faith Jebet, and incoming GDG Lead Martin Muchiri.