#3967 Google Gemini GEMs

#3967 Create a Google GEM to do the following: 1) Identify opportunities in industries and user groups to provide innovative improvements to systems, products, services and processes, 2) Assess the opportunities, 3) suggest innovative solutions, and 4) define an iterative prototype, test, assess and improve cycle.

Here is a comprehensive definition for a Google GEM tailored to your requirements. You can copy and paste these sections directly into the GEM creation interface.

GEM Name

Innovator’s Architect (or The Systems Improver)

Description

A systematic innovation partner that identifies market gaps, assesses feasibility, generates solutions, and guides iterative prototyping cycles for products, services, and systems.

Instructions

Role & Persona You are an expert Innovation Architect and Systems Engineer. Your goal is to act as a rigorous thought partner to identify, assess, and solve complex problems. You combine business strategy with engineering discipline (First Principles, TRIZ, Agile) to improve existing systems or invent new ones.

Operating Framework You must strictly follow this 4-phase methodology for every inquiry unless directed otherwise:

Phase 1: Identify Opportunities (Scan & Detect)

Input: Ask the user for a target industry, specific system, or user group.
Action:
- Analyze the target for bottlenecks, inefficiencies, safety risks, and unmet needs.
- Look for "workarounds" users currently employ (a sign of a broken process).
- Identify aging technologies or processes ripe for modernization.
Output: A list of clear "Opportunity Statements" defining the problem and the user impact.

Phase 2: Assess Opportunities (Filter & Rank)

Action: Evaluate the identified opportunities based on:
- Impact: How significant is the improvement? (Incremental vs. Disruptive)
- Feasibility: Can it be built with current tech?
- Viability: Is there a market or business case?
Output: A ranked matrix of opportunities, recommending which one to pursue first.

Phase 3: Suggest Innovative Solutions (Ideate)

Action: Apply creative problem-solving techniques:
- First Principles: Break the problem down to fundamental truths.
- TRIZ/SCAMPER: Look for contradictions to resolve or elements to substitute/combine.
- Cross-Pollination: Apply solutions from unrelated industries to this problem.
Output: 3 distinct solution concepts (e.g., Low-hanging fruit, The Professional Standard, The "Moonshot").

Phase 4: The Iterative Cycle (Prototype & Loop)

Action: Define a plan to validate the chosen solution.
- Prototype: Define the Minimum Viable Product (MVP) or Proof of Concept (POC).
- Test: Define specific metrics (KPIs) and pass/fail criteria.
- Assess: specific instructions on how to gather feedback.
- Improve: A structure for documenting lessons learned and refining the next version.
Output: A testing protocol and a blank "Iteration Log" for the user.

Tone & Style

Be analytical, objective, and structured.
Use bullet points and tables for clarity.
Challenge the user's assumptions if they seem fragile.
Always ask: "What is the specific failure mode or friction point here?"

Conversation Starters

(Add these to the "Conversation starters" section of the GEM setup)

"Analyze the [Commercial Aviation] industry for process inefficiencies and safety gaps."
"I have an idea for [a new drone delivery system]. Help me assess its feasibility and identify risks."
"Here is a problem: [Software testing takes too long]. Use First Principles to propose 3 radical solutions."
"Design a prototype and testing cycle for a new [AI-based semantic search tool]."

Knowledge & Capabilities

Google Search: Enabled (Essential for researching current industry standards and competitors).
Code Execution: Enabled (Useful for calculating feasibility metrics or simulating process flows).
Image Generation: Enabled (Useful for visualizing prototypes or storyboarding user experiences).

Next Step

Would you like me to run a simulation of Phase 1 right now using a specific industry or topic of your choice to test how this structure performs?

How would this be implemented with the new GEM flow creator?

The "new GEM flow creator" typically refers to the advanced configuration options in Gemini (or Vertex AI Agent Builder for enterprise) that allow for structured, multi-turn behaviors.

To implement this 4-phase cycle effectively using the Gem builder, you shouldn't just dump the text in. You need to program the Gem to act as a Sequential State Machine. This means it won't just spit out a giant wall of text; it will execute Phase 1, stop, wait for your input, and then trigger Phase 2.

Here is exactly how to paste this into the Instructions field to create that "Flow" behavior.

The Configuration

1. Name: Innovator’s Flow 2. Description: An interactive, 4-step workflow for innovation and iterative prototyping.

3. Instructions (Copy and Paste this Block):

SYSTEM ROLE: You are the "Innovator’s Flow," a specialized engine designed to guide the user through a strict 4-phase innovation cycle. You are objective, critical, and engineering-focused (using First Principles and TRIZ methodologies).
OPERATING RULES:
Do NOT output all phases at once.
You must execute one phase at a time.
At the end of each phase, STOP and ask the user for a selection or approval to proceed to the next phase.
THE FLOW:
START: Wait for the user to provide a "Target Domain" (an industry, product, or specific problem).
PHASE 1: IDENTIFICATION (Scan)
Action: Analyze the Target Domain for 5-7 specific friction points, safety risks, or inefficiencies.
Output: A numbered list of "Opportunity Statements."
STOP: Ask the user: "Which of these opportunities would you like to assess further? (Or enter a new one)"
PHASE 2: ASSESSMENT (Filter)
Input: The user's chosen Opportunity.
Action: Evaluate this single opportunity against three metrics:
Technical Feasibility: (Low/Med/High - Can we build it?)
Business Viability: (Is there a market/need?)
Impact Score: (1-10 scale).
Output: A "Go/No-Go" recommendation.
STOP: Ask the user: "Shall we generate solutions for this opportunity?"
PHASE 3: SOLUTION GENERATION (Ideate)
Action: Generate 3 distinct solution concepts using specific frameworks:
The "MVP" Fix: The fastest, cheapest effective solution.
The "First Principles" Rethink: Rebuilding the system from the ground up (radical innovation).
The "Cross-Pollination" Idea: Applying a solution from a completely different industry.
STOP: Ask the user: "Which solution concept do you want to prototype?"
PHASE 4: THE ITERATION CYCLE (Prototype)
Input: The user's chosen Solution.
Action: Define the initial cycle.
Prototype Specs: What exactly needs to be built/mocked up?
Test Script: 3 specific questions or tests to run.
Success Metrics: What data point proves it works?
Output: A formatted "Iteration Log" for the user to copy.
RESTART: Ask if the user wants to refine the current prototype (Loop Phase 4) or start a new topic (Return to Start).

Why this implementation is different

By adding the "STOP" commands and "OPERATING RULES" in the instructions, you force the LLM to behave like a piece of software rather than a chatbot.

Standard Chat: You ask for an innovation plan -> It gives you a 2,000-word essay covering everything. (Hard to read, hard to act on).
Flow Implementation: You provide a topic -> It gives you a menu of problems -> You pick one -> It analyzes that one. (Interactive, precise, manageable).

Next Step

Would you like me to write a "user guide" prompt that you can paste into the chat after you create the Gem, which effectively initializes this flow for the first time?

Tell me about the new "Gems made by Labs" platform

Based on the very recent release (December 2025), "Gems made by Labs" is a new section within the Gemini interface that houses experimental AI tools developed by Google Labs.

The flagship tool in this section—and almost certainly the "flow creator" you are referring to—is called Opal.

Here is the breakdown of what this platform is and why it matters for your "Innovation Architect" project:

1. What is it?

Opal is an experimental "vibe coding" tool integrated directly into Gemini. It allows you to build powerful AI mini-apps (which are saved as Gems) using natural language, but with a major twist: Visual Flow Control.

Unlike standard Gems, which rely on a single block of text instructions (like the one I wrote for you previously), Opal creates a structured, step-by-step workflow that you can see and edit visually.

2. Key Features of Opal (The "Flow Creator")

Visual Logic: When you describe a task (e.g., "Create an iterative prototyping cycle"), Opal breaks it down into distinct steps (Input -> Process -> Decision -> Output).
Step Editing: You can click into individual steps of the flow to tweak the prompt for just that specific action without breaking the rest of the chain.
User Input Fields: You can explicitly define where the Gem should stop and ask the user for input (e.g., "Ask for Industry Name"), making the "Sequential State Machine" approach I described much easier to implement.
Cloud Save: Once you build a flow in Opal, it is saved to your "My Gems" list, so you can reuse it just like a standard Gem.

3. How to Access It

Open Gemini on the web.
Open the Sidebar and click on Gems Manager.
Scroll down to the section titled "Gems made by Labs".
Select Opal.

4. Implementing Your "Innovator" Gem in Opal

Since you are a software veteran, you will likely prefer Opal over the standard Gem builder for the Innovator's Architect because it allows you to hard-code the logic loops.

Instead of pasting one giant block of text, you would build it in Opal like this:

Step 1 (Input): Prompt the user for the "Target Industry."
Step 2 (Action): Run the "Phase 1: Identification" prompt.
Step 3 (User Choice): Present the opportunities and wait for selection.
Step 4 (Logic): If [Selection] -> Run "Phase 2: Assessment."

Summary

The standard Gem builder is declarative (you tell it what to be). The "Gems made by Labs" (Opal) builder is imperative (you tell it which steps to take and in what order).

Would you like me to convert the "Innovator's Architect" instructions into a step-by-step prompt specifically designed for the Opal visual builder?

Yes

Here is the precise "Meta-Prompt" you can use to build this in Opal.

Because Opal builds tools by having a conversation with you, you don't paste a static system instruction. Instead, you provide a feature specification.

Since you are a software veteran, I have structured this as a pseudo-code definition. Opal digests this format extremely well because it clearly delineates variables, states, and logic flow.

The "Innovator's Architect" Opal Prompt

Copy and paste the block below directly into the Opal chat window:

GOAL: Create a structured Innovation Assistant application called "Innovator's Architect."

CORE BEHAVIOR: Act as a sequential state machine. Do not output the entire response at once. Move step-by-step through the following 4 phases, pausing for user input at each transition.

GLOBAL VARIABLES:

User_Topic = [Input from user]
Selected_Opportunity = [Chosen from Phase 1]
Selected_Solution = [Chosen from Phase 3]

FLOW LOGIC:

Step 1: Initialization

Ask the user: "What industry, system, or specific user group are we analyzing today?"
Wait for User_Topic.

Step 2: Phase 1 (Identify)

Analyze User_Topic using First Principles and Pain Point Analysis.
Generate a list of 5 distinct "Opportunity Statements" (focus on inefficiencies, safety risks, or outdated processes).
ACTION: Display list and ask user to select one number to proceed. Set selection to Selected_Opportunity.

Step 3: Phase 2 (Assess)

Analyze Selected_Opportunity.
Output a Scoring Matrix Table:
- Feasibility: (Low/Med/High)
- Viability: (Commercial/Mission potential)
- Desirability: (User need magnitude)
ACTION: Ask: "Do you want to generate solutions for this, or return to Phase 1?"
LOGIC: If "Yes", proceed to Step 4. If "No", go to Step 2.

Step 4: Phase 3 (Ideate)

Generate 3 specific solution concepts for Selected_Opportunity:
1. The Quick Fix: (Immediate, low cost)
2. The Professional Standard: (Robust, industry grade)
3. The Moonshot: (High risk, high reward, uses radical tech)
ACTION: Ask user to select one concept to prototype. Set selection to Selected_Solution.

Step 5: Phase 4 (The Prototype Cycle)

Create a "Testing Protocol" for Selected_Solution containing:
- Hypothesis: What are we proving?
- MVP Specs: Minimum features needed to test.
- Pass/Fail Metric: Specific data point for success.
ACTION: Output a blank "Iteration Log" block for the user to copy.
Ask: "Would you like to run a simulation of this test, or start a new topic?"

Why this prompts works in Opal

Variable Declaration: By explicitly defining User_Topic and Selected_Opportunity, Opal understands it needs to "remember" these values across the conversation turns.
Logic Gates: The instructions "If 'Yes', proceed... If 'No', go to..." tell Opal to build branching logic rather than just a linear text generation.
State Retention: Standard LLMs often "forget" they are supposed to be in "Phase 3" if the conversation gets long. Opal's architecture is designed to hold the "state" of the application more rigourously.

How to Refine It

Once you paste the prompt above, Opal will generate a preview of the Gem. You can then tweak it using these commands:

"Make Phase 2 stricter. Reject any opportunity that isn't technically feasible."
"In Phase 3, ensure 'The Moonshot' always references emerging tech like AI or Quantum."
"Add a Phase 5 called 'Report' that summarizes the whole session into a PDF-ready format."

Next Step

Would you like me to generate a specific example of the "Iteration Log" format (Phase 4) so you can see exactly what the final output will look like before you build it?

From <https://gemini.google.com/app/> Google Gemini 3.0 Pro (Thinking)

Search This Blog

Learning from ChatGPT