Hey everyone,

Last week, I revealed my AI Operating System—the mental framework I’ve developed for using AI in my work.

(If you missed last week’s post, click here.)

It is the biggest strategic lever I’ve ever employed for productivity. I really can’t imagine working without these tools anymore.

Changing one’s mindset toward AI is the first crucial step to going full AI-native. We covered that last week.

Now, to show you what going full AI-native looks like, we’re getting down to the nitty gritties of my personal AI tool stack.

I’ll reveal every AI tool I use in every domain, from email to coding. No stone will be left unturned.

I highly recommend you steal my stack or adopt a new tool if you haven’t used it before. And if you think someone could benefit from this, share the post with them.

Let’s get into it.

My AI Operating System Pt. II

Rocket ship for the mind

Jobs' timeless quote from the '90s was hyper-prescient for its time.

With hindsight, it's obvious how incredibly revolutionary computers have been, even when we look at the iterations that Jobs worked on.

But today we have far more efficient modes of transport than the bicycle. In the same way, computers and software have evolved far beyond what they were in 1990.

If the computers of the '90s were bicycles for the mind, then AI tools powered by large language models are like rocket ships.

We’ve never had a technology that accelerates our thinking as much as AI does.

As humans, we've never been able to write emails this quickly, take notes this quickly, or code apps this quickly.

With such powerful technology at our disposal, you'd be doing yourself a disservice by not using it in all facets of your work.

It's no joke when I say that there is an AI tool for every single thing you do in your day.

The list below should prove that.

The tools

My co-pilots

ChatGPT and Claude form the foundation. They are always open in my browser.

After using them for thousands of hours, I’ve learned the models’ strengths and weakenesses.

ChatGPT tends to be better at structured analysis and maintaining context across long conversations.

Claude, on the other hand, dominates creative ideation, writing, coding and conversational flow. The suite of Claude models is truly underrated and potentially underappreciated.

ChatGPT rules the Zeitgeist but Claude just has that secret sauce a lot of the time.

If you haven’t tried Claude before, I’d highly recommend giving it a whirl. Especially for coding and creative writing.

ChatGPT and Claude are my thinking partners and executive virtual co-founders. I’ll constantly bounce between them throughout the day as I fine-tune my ideas and brainstorm new ones.

To me, this is what it truly means to use AI as an operating system.

The benchwarmers

Gemini and Grok don’t start the match, but when they come on, they make a hell of an impact. Fundamentally, they serve as my second opinions.

Just like you’ll go to different doctors for unique perspectives on an MRI, when I need a second opinion on a problem, I'll run the prompt I gave to Claude through Gemini and/or Grok.

I’ll even post Claude’s exact answer into Gemini or Grok to stress-test the response and poke holes in the original thesis.

My core principle is that I don’t just want the first answer the AI comes up with. I want the best answer. Getting that inherently requires more time, effort, planning, and scrutiny of the AI's answers.

A pro tip from Greg Isenberg I found hilariously helpful was to pit the AIs against each other.

When cross-checking ChatGPT with Gemini, don't just paste an answer. Explicitly tell Gemini that this answer came from ChatGPT and motivate it to do even better.

Somehow this tends to tease out better answers. And it’s super fun.

At a high level, the benchwarmers principle gives you a board of advisors in your pocket.

Each AI has its own personality, reasoning process and answering style, which means you get genuinely different approaches and perspectives to the same challenge.

Google replacement

Perplexity has become my default search engine. Essentially, anything I used to search on Google, now goes into Perplextiy.

Perplexity does an excellent job at aggregating internet sources (just like Google) but the cherry on top is that it gives me synthesized AI-based answers from those sources.

This saves me from the old ritual of opening thirteen different blue links and trying to piece together information myself.

It's particularly powerful for social listening. I can ask, “What are people saying about ChatGPT on Reddit?” and get actual insights.

Search aside, the Perplexity team is prolific, shipping tools and features around their core search service that make the product a pure joy to use.

Deep Research

All the usual suspects have a ‘deep research’ mode.

Imagine this as what Perplexity normally does but on steroids.

Deep research instructs the AI to spend a lot more time searching the internet for what the user is asking for. It’s perfect for in-depth research reports of any kind.

So recently, I ran a fascinating experiment.

I needed to understand exactly how Robinhood makes money, so I pitted all four major deep research tools against each other:

• Perplexity Research

• ChatGPT Deep Research

• Gemini Deep Research

• Grok DeeperSearch

The differences were striking. Each tool approached the research differently, with outputs of varying lengths, styles and levels of depth.

Researching Robinhood this way probably saved me days in manual data gathering, synthesis and writing.

Using these tools is like having a team of personal research assistants in your pocket.

Context Kings

Writing personalized reports on a specific topic that requires a lot of thought is hard.

You have to gather a tonne of information, make sense of it all, reason through the problems, synthesize it and structure it in a way that makes sense. All before physically writing it out.

Now, when I need to process massive amounts of information and reason through multi-step problems (whether I’m building detailed client proposals, technical specs or comprehensive reports), OpenAI’s reasoning models are my go-to.

These models excel at tasks requiring extensive context and careful reasoning. I'll feed them everything: client background, project requirements, return format, market research, competitive analysis and most importantly, a tonne of context.

They synthesize it all into a coherent, well-written and actionable report in a matter of minutes.

Writing

My writing process is now a multi-tool orchestra.

Grammarly handles the basics. It catches things I miss when it comes to spelling and grammar.

Claude dominates content ideation and first drafts, especially when I need writing that matches my voice and style.

ChatGPT's Canvas tool is brilliant for targeted edits. I can highlight specific sections and ask for improvements without rewriting the entire piece.

And for visual content (like post thumbnails), I have a custom GPT trained specifically for creating thumbnail prompts, which I then feed into Midjourney or GPT-4o for image generation.

Yes, AI writing is not at a level that surpasses human writing on its own… yet. But AI tools for writing make my process 10x more efficient.

Most of all, writer’s block has never existed and will never exist for me because of how capable AI is at generating a decent first draft.

Coding

My coding workflow starts with Bolt and Lovable for rapid prototyping. I can go from concept to working demo in minutes, not days.

For full-stack apps, Replit has it all—frontend, backend, database, all in one environment.

When projects get complex, I switch to Cursor with Claude 3.7 Sonnet as my coding assistant.

The speed is intoxicating. I'm building things I could never have built before, and I'm doing it in a fraction of the time.

I’ll be writing more on my vibe coding process in the coming weeks.

Email

I don’t recommend software tools lightly. It takes a lot for something to blow my mind.

As I mentioned last week, the first time I used ChatGPT was a good example. Another was the first time I used Superhuman.

I never realised how much time I wasted on email before using this product. Superhuman literally saves me 2-4 hours a day in getting through my emails.

One thing it does exceptionally well is how it categorizes emails automatically with AI, filtering out the noise.

But the real magic happens when I need to write responses. It generates personalized, context-rich first drafts that actually sound like me.

The best part is that I get through my emails in a tenth of the time.

Meetings

Notion AI and Microsoft Copilot handle all my meeting transcriptions and summaries.

I can focus entirely on the conversation, knowing everything important will be captured and organized.

The AI summaries save hours of manual note-taking and often catch nuances I missed in real-time. It's the equivalent of having a perfect memory of every conversation from your work day.

Voice

Aqua Voice has changed everything. No exaggeration.

This tool is the most accurate speech-to-text transcription tool I've ever used, and it's made me realize that speaking is just fundamentally faster than typing.

There’s just too much latency with typing. We all know that we can speak faster than our fingers can move.

The limiting factor until today has always been a voice tool that actually works instead of transcribing garbled text that is completely unrelated to what you said.

Ahem, Siri…

I now dictate emails, brainstorm ideas and prompt LLMs with my voice.

Anything that involves getting thoughts out of my head and into digital form goes through Aqua.

As you can imagine I am saving loads of time using this tool.

Automation

Zapier's AI agents are the newest addition to my stack.

They sit within my workflows and make intelligent decisions based on context.

They can categorize leads, prioritize tasks and route information to the right team members—all without my intervention.

It's business automation that actually thinks rather than being stuck in rigid logic.

Yes, agents still have a long way to go, but it’s clear that the rate of improvement is exponential.

Horse & Carriage

Speaking of exponential improvement, let’s talk about the first-ever motor vehicles.

In 1900, cars were noisy, unreliable and disregarded by the public. People literally threw stones at them and spread nails on the roads to puncture their tires.

The consensus at the time was that horses worked just fine—why fix what wasn't broken?

Sound familiar?

Today, incumbents dismiss AI as ‘unreliable’ or ‘just a novelty.’ They're emotionally attached to their manual workflows and their way of doing things.

They’ll see an occasional hallucination and proclaim from the hills, “See? I told you this stuff doesn't work.”

But here's what I know after thousands of hours working with these tools:

The question around AI replacing traditional workflows has changed from if to when. The real debate revolves around how quickly it'll happen.

In 1920, the skeptics weren't arguing about cars anymore. They were buying them.

The people who adopted cars early were accessing opportunities that horse and carriage advocates couldn't even see coming.

The same can be said for AI’s early adopters.

But the transition from 99% manual workflows to 99% AI is not going to take as long as 20 years, like it did for the car. The infrastructure is evolving rapidly. The tools are getting better daily. And the adoption curve is going exponential.

You can't afford to get left behind. Not because I say so, but because the future has a way of arriving whether you're ready for it or not.

Like I said last week, AI won’t take your job. But someone using AI will.

Was this newsletter forwarded to you? If you liked it, sign up here.

And, if you loved it, forward it to your friends and family.

Thanks for reading,

— Luca

P.S. Want to work together?

1. Get your free AI audit: If you’re still manually doing work that AI could handle in minutes, I'll show you exactly where you're bleeding time and money. In 25 minutes, I'll audit your current workflow and map out your personalized AI roadmap. This is the same process I used to save 15+ hours per week in my own business. 🤖

2. See my complete tool stack: Beyond what I've shared today, I maintain a living document of every AI tool I test, why I use it, and which ones actually move the needle. Access my complete AI toolkit by reaching out here. 💼

That’s it from me. See you next week, Luca 👋

P.P.S Let’s connect on LinkedIn and X.