Episode 29:
The Art of Stacking AI Agents

Hey friends,

Here’s something I learned the hard way: trying to use one AI model for everything is like using a hammer for every job.

Sure, it can work — but you’ll dent a lot of screws along the way.

After thousands of prompts, dozens of experiments, and a lot of wasted time, I’ve landed on a simple truth:

I’m not talking about chasing the newest model every week.

It’s more about understanding what each AI does best — and building a workflow where they connect with each other.

Let me show you exactly what I mean.

📌 What Is “Stacking” AI Agents?

Stacking is using multiple AI tools together, where the output of one becomes the input for another.

Here’s why it makes all the difference from just “using AI”:

• Better outputs: You get strategy from thinkers, execution from doers, polish from editors
• Specialization: Each model handles what it’s actually good at, not what it can do
• Less frustration: You stop forcing a writing model to think strategically (and vice versa)
• Compound results: Small wins from each tool stack into massive time savings

Think of it like hiring a team instead of asking one person to do everything.

A researcher, a writer, an editor and each plays their role.

✨ How I Stack AI Models (My Real Workflow)

I’m going to walk you through exactly how I combine tools. just what actually works for me

Feel free to steal it or inspire your own AI stacks.

✅ 1. Planning with ChatGPT, Coding with Claude

This one was a turning point for me.

When I started building full stack apps with Claude Cowork and Claude Code, I noticed something:

Claude is incredible at executing code, but ChatGPT is the better planner.

ChatGPT’s “planning” I have felt is more thorough.

So now my workflow looks like this:

• GPT-5.2 Thinking: I dump my messy idea and ask it to build a Product Requirements Doc (PRD). It asks clarifying questions, identifies edge cases, and structures the whole project before a single line of code is written.
• Claude Code: Then I hand the plan to Claude, and it builds. Properly. With structure, error handling, and clean code.

This burns less tokens when you’re coding because there’s less errors.

Pro tip: Don’t skip this step if you are serious about building!

✅ 2. Chaining Agents for Complex Tasks

The real power comes when agents work together in a chain:

Single agent: You → Agent → Output → You review

Chained agents: You → Agent 1 → Agent 2 → Agent 3 → Final output

Here’s an example content creation chain I’ve used:

1. Research Agent (Perplexity): Finds trends, gathers examples → outputs research brief
2. Outline Agent (ChatGPT): Creates structured outline with hooks → outputs 7-point outline
3. Writing Agent (Claude): Writes full draft in my voice → outputs complete post
4. Editor Agent (Claude): Reviews for clarity and engagement → polished final version

Your role: Review the final output, make tweaks, approve.

Time: 20 minutes instead of 2 hours.

Pro tip: Start small and where you’re most comfortable

✅ 3. Reasoning Models for Strategy, Regular Models for Speed

Not all AI models are built to think with you.

Here’s the split I use:

Best reasoning models: GPT-5.2 Thinking, Gemini 3 Pro, Claude Opus 4.5 with extended thinking

Best speed models: GPT-5.2 Instant, Claude Sonnet 4.5, Gemini 3 Flash

I don’t ask GPT-5.2 Thinking to draft my newsletter, it’s not great at that.

But for thinking through a messy strategy and turning it into a clear roadmap? That’s where it shines.

Pro tip: Traditionally thinking models aren’t the best at writing, but you could get good output from it still. Just not my style

✅ 4. Perplexity for AI Search, NotebookLM for Learning on the Go

This one’s been SUPER helpful for absorbing complex topics without sitting at my desk.

Here’s the stack:

• Perplexity AI:Perplexity AI: I use it as my go-to AI search engine for finding the latest information, researching topics with citations, and getting quick answers to complex questions without digging through multiple sources

• NotebookLM: I drop Perplexity’s output (and my original sources) into NotebookLM and generate an Audio Overview or Video Overview.

Now I’ve got a podcast-style deep dive I can listen to while walking, commuting, or doing chores. Or a video slide deck to review on my phone.

The magic: Perplexity AI does the heavy research.

NotebookLM synthesizes it and transforms it into something I can absorb without screens (more on this in future episodes).

Pro tip: Use NotebookLM’s “Customize” option to focus the Audio Overview on specific sections or reinforce the style of audio you want.

🛠 How to Get Started (3 Steps)

1. Map your current workflow first: Before you automate anything, write down what you actually do. Agents come last, not first.
2. Pick ONE pain point to solve: Research, content, planning, or documentation. Don’t spin up five agents on day one.
3. Match the model to the job: Use this quick cheat sheet:

📈 Why This Matters

Here’s the thing: your bosses don’t care about models.

They care about outcomes.

When you learn to stack AI tools properly, you translate that into:

• Time saved: “I cut research time from 3 hours to 20 minutes.”
• Quality improved: “I catch inconsistencies I used to miss.”
• Capacity unlocked: “I can take on two more projects this quarter.”

The people who master this aren’t just “using AI”.

They’re building AI systems.

And systems compound 💪🏾

🫡 Final Thoughts

I used to chase the newest model every week.

Now I just ask: what’s each tool actually good at?

ChatGPT thinks.

Claude codes.

Perplexity researches.

Notion organizes.

NotebookLM teaches me on the go.

Stack them together, and suddenly you’re not working harder.

You’re working smarter, with a team that never sleeps.

Try this today: Take one task you do regularly. Break it into stages. Ask yourself: which AI is best at each stage?

That’s the beginning of your stack.

If this helped you see AI differently, forward it to someone still trying to do everything with one tool.

They’ll thank you 😉

To smarter stacks and sharper workflows,
Nahid
​

​​​​​

600 1st Ave, Ste 330 PMB 92768, Seattle, WA 98104-2246
​Unsubscribe · Preferences​