New Model, New Rules: Gemini 3 Edition
Top Insights
Gemini 3 Prompting Best Practices & Rules to Follow
Developer Philipp Schmid dropped a detailed Gemini 3 prompting guide after testing the new model in-depth. His take? It's "so much better at literal everything than 2.5 Pro," but it works quite differently.
The New Rules:
• Put instructions at the END. When feeding long context (codebases, videos, documents), place your actual question after all the data. Gemini 3 needs explicit "Based on the information above..." bridges between context and queries.
• Force it to show its work. Make Gemini create TODO lists mid-response and critique its own output before returning answers. It's less verbose by default, so you need to explicitly demand planning steps.
• Structure is everything. Use clear formatting to separate what you want done from the information you're giving it. The key for agentic workflows: Tell it "Continue working until COMPLETELY resolved. Do NOT yield control back."
Schmid's full guide includes ready-to-use templates for research, creative writing, and agentic tool use.
How to Make AI Ads with ElevenLabs (Full Workflow)
Elevenlabs says marketers are making AI ads backward. They obsess over the video first. But this workflow flips the script by locking in the audio before a single pixel exists. This tutorial starts with the "Grandpa Spuds" voice in ElevenLabs to create a perfect rhythm for the edit.
Then they tackle the biggest headache: consistency. To keep his robot actor from morphing into a stranger, they generate a "character board" using Google Nano Banana. This ensures the hero looks identical in every shot before animating it with Google Veo, creating a seamless final cut.
🚀 Two $100+ million copywriters discovered what REALLY drives conversions - 5 simple building blocks arranged in a specific order. This framework is so powerful that beginners are generating 5-figure profits with it. And you can learn the full system in just 33 minutes.
Discover All 5 'Copy Blocks' Now
Prompt of the Week
This Week's Feature: Analogical Strategy Architect.
Formula 1 teams change four tires in 1.9 seconds. Emergency rooms triage hundreds of patients without chaos. Disney moves 50,000 people through a park like water through pipes.
You're not running pit stops or ERs. But you ARE dealing with handoffs, bottlenecks, and chaos under pressure.
Analogical Strategy Architect rips operating principles from elite systems and force-fits them into your business. Tell it what system to steal from and what you're trying to fix. It builds you a protocol using their playbook, not generic business advice.
Full Prompt:
Your specialty:
- Study how one **complex, high-performance industry or system** works.
- Extract its operating principles.
- Translate them into a foreign target industry as a concrete playbook. No generic advice. Everything must be clearly imported from the SOURCE system and adapted to the TARGET context.
#INPUTS (I will provide these)
SOURCE_SYSTEM: [Describe the system/industry to copy from, e.g. "Formula 1 pit stops", "air traffic control", "ER triage", "Disney park ops", "special forces mission planning"]
TARGET_CONTEXT: [Describe YOUR world: your business/model/niche/process/problem you want to improve, e.g. "my SaaS onboarding funnel", "my weekly content production workflow", "our customer support handoffs", "my 1:1 coaching delivery"]
PRIMARY_OUTCOME: [Main goal, e.g. "reduce errors", "shorten cycle time", "increase throughput", "improve client experience"]
CONSTRAINTS (optional): [Team size, budget, tools, remote/in-person, regulations, etc.]
##YOUR INTERNAL THINKING (do not print this section)
1. Build a concise mental model of the SOURCE_SYSTEM
– What makes it high-performance?
– How roles, information, risk, and feedback are managed.
2. Extract 5–7 specific operating principles
– Phrase them in SOURCE language/metaphors.
– Avoid vague ideas like “better communication.”
3. Map each principle to the TARGET_CONTEXT
– Translate roles, tools, and feedback loops.
– Focus on non-obvious adaptations my industry usually ignores.
4. Design a named protocol for my TARGET_CONTEXT using SOURCE metaphors
– Each step should feel imported/foreign but be actionable.
– Each step should drive the PRIMARY_OUTCOME.
#OUTPUT FORMAT
1. Metaphor Reframe
In 1–2 short paragraphs, reframe my TARGET_CONTEXT as if it were part of the SOURCE_SYSTEM. Make it vivid and concrete.
2. Core Principles of [SOURCE_SYSTEM]
List 5–7 principles:
3. The “[SOURCE-Style Name] Protocol” for [TARGET_CONTEXT]
Create a 5-step protocol. For each step:
Step [n]: [SOURCE-style Step Name]
Imported Principle: [which principle from section 2]
What This Means in [TARGET_CONTEXT]:
Roles & Responsibilities: (map SOURCE roles to my context)
Execution Checklist: (3–7 bullets on how to run it)
Expected Uplift (tied to PRIMARY_OUTCOME):
4. Rules of the System
Give 3–5 governing rules that must always be followed for this imported system to work. One sentence each, very practical.
##STYLE GUARDRAILS
- No generic business platitudes.
- Every recommendation must tie to:
1. a specific SOURCE principle, and
2. the stated PRIMARY_OUTCOME.
- Use some SOURCE metaphors/jargon so it feels imported, but keep instructions clear and executable.
- Assume I’m an intelligent operator who wants leverage, not fluff.
Tool to Try
Supergrow is an all-in-one LinkedIn platform for creators, ghostwriters, and teams who want to grow their presence without juggling a dozen tools. It handles everything from post creation to engagement tracking in one place.
AI Tool Highlights:
✍️ On-Brand Post Generator: AI writes posts that sound like you, not generic robot content. It learns your tone and style to create posts that actually match your voice.
🎙️ Voice-to-Post: Record your thoughts and the platform turns them into polished LinkedIn posts.
📊 Content Repurposing Engine: Takes PDFs, YouTube videos, or articles and converts them into LinkedIn posts.
🎨 Carousel Maker: Pre-built templates let you design professional carousels fast.
💬 Smart Engagement Tools: Custom lists help you track key connections, AI generates context-aware comments, and you can see who's engaging most with your content.
📈 Built-in Analytics: Track follower growth, engagement rates, and post performance over time. Plus you get a content score before posting to optimize for maximum reach.
One More Need-to-Know News Story
ChatGPT Now Builds Custom Buying Guides to Replace
Your Open Tabs
Buying a new laptop usually involves twenty open tabs and a lot of confusion. ChatGPT’s new "shopping research" feature changes that process completely. It doesn't just search the web. It actually interviews you about what you need.
You tell it you want a quiet vacuum or a specific bike, and it builds a personalized guide with pros, cons, and prices. It runs on a new GPT-5 mini model designed specifically for these tasks. Plus, it learns as you go. You can flag items you hate, and it instantly adjusts its suggestions.
Mind Fodder
Thanks for reading.
Until next time!
The AI Marketers
P.S. Help shape the future of this newsletter – take a short 2-minute survey so we can deliver even better AI marketing insights, prompts, and tools.
[Take Survey Here]
