The Operator Stack

My Current AI Stack

4 min read

image This is not a recommendation guide. This is not a list of “must-haves.” This is a log of my current operational reality.

The market shifts weekly. My setup shifts with it. But the principle remains constant: I do not hire “AI.” I hire specific capabilities.

Here is the current roster.

The Coding Forge

Tools: Claude Code | Gemini Antigravity IDE (Gemini 3)

The Reality: I have never coded in my life. I am a zero-code builder. This means I cannot optimize for speed. I optimize for survival.

The Protocol:

  1. The Pre-Work: Before a single line of code is generated, I forage. YouTube, Reddit, expert forums. I define exactly what I am building and the tools required.
  2. The Blueprint: I compile this research into a single brief.
  3. The Negotiation: I debate the brief with Claude first to spot architectural flaws. I refine it with Gemini or ChatGPT.
  4. The Build: Only then do I code.
    • I use Claude Code as the primary builder.
    • I am testing Gemini Antigravity IDE for its Gemini 3 capabilities.
  6. The Kill Rule: My goal is simple—it must compile, and it must work. If a tool fails a task twice, it is fired.

The Thinking Engine (Pressure Testing)

Tools: Gemini (Lead) | ChatGPT | Claude (Audit)

The Protocol: I don’t use these to “get answers.” I use them to break my arguments.

Hard Rule: “Do not be friendly.” (I explicitly instruct models to drop the customer service polite-mode).

The Workflow:

  1. Initiate in Gemini: Its long context window handles the initial “brain dump” best.
  2. The Adversarial Prompt: I feed the premise to all three models:
    • “Where is the logic gap?”
    • “What assumption is false?”
    • “Simulate the counter-argument.”
  4. The Cross-Fire: If a line of reasoning survives the critique of three different models, it is ready for the real world.

The Drafting Smith

Tools: ChatGPT | Claude | Gemini

The Protocol: Structure, compression, and clarity.

Hard Rule: AI never writes the final draft. It is the editor, not the author.

The Workflow:

  1. Context Load: I provide the goal, the audience, and the structure.
  2. The “Anti-Slop” Guardrail: I apply strict style rules to prevent generic “AI Slop” output.
  3. The Grind: I go through 4-5 iterative edits with AI.
  4. The “Day 2” Audit: I mark the draft as “Final,” then walk away. I revisit it the next day. Errors always appear with fresh eyes.
  5. Final Polish: Only after my manual review do I use AI for a final clarity check.

Deep Research & Search

Tools: Gemini

The Protocol: For synthesis and scanning the horizon. When I need market data, sourcing, or technical docs, Gemini wins because of its Google Search integration. It reduces the “search friction” of opening 20 tabs.

The Caveat: No AI tool builds a great report on its own. They are too shallow.

  • Step 1: Let AI build the base.
  • Step 2: Read the report to find the gaps.
  • Step 3: Create a targeted question list for the missing pieces.
  • Step 4: Re-investigate those specific areas manually or with focused prompts.

The Architecture (The Real Secret)

Tool: Google Docs

The specific models listed above matter 10% of the time. How I manage context matters 90% of the time.

Most of us burn out from “Context Fatigue”—starting every new chat by re-explaining their business, their tone, and their constraints.

My Solution: The Single Source of Truth.

  • Every project has one Master Doc.
  • This Doc contains the current state, the goal, and the constraints.
  • I update this Doc continuously.
  • Since all AI tools connect to Google Drive, every new session starts by reading this Doc.

The Result: The models always see the latest version of reality. I never repeat myself. I swap models instantly without losing momentum.

Tools are commodities. Systems are assets.

Graph View

Backlinks