Hey!

Feeling lost in the sea of AI models?

GPT-this, Claude-that, Gemini-something-else… it’s easy to get analysis paralysis that prevents us from moving forward.

I still see a lot of confusion about which AI model to use and how to use it effectively.

Black-and-white thinking (“AI is amazing!” or “AI is useless!”) doesn’t help.

We’ve discussed this before. The right tool depends entirely on the job.

So, forget the hype for a second. I want to share my personal, battle-tested AI workflow — the models I actually rely on daily and why.

My Go-To AI Arsenal

Here are the heavy hitters in my current rotation:

OpenAI:

• GPT-4o: My all-around workhorse for general questions and quick research. Fast, reliable, and versatile. It has actually replaced Google search for me.
• o1: The strategist. When I need to go deep, structure complex ideas, and develop actionable plans based on research.

Anthropic:

• Claude 3.7: The writer. My first choice when I need to write any text. Unlike GPT models, it forms text more readably and less “robotically”. A model capable of “reasoning”. One of the best models for programming, but it needs good guidance.
• Claude 3.5: The predecessor to 3.7. Since this isn’t a “reasoning” model, it’s good for precise instructions and sticks to them better than 3.7, which can “think” a bit on its own - not ideal for coding.

Google:

• Gemini 2.5 Pro: A new contender that accepts and understands text, images, or video in the chat window. Excellent at generating code. In short: I feel this is the best and most useful general-purpose model. Brainstorming, writing, programming.

On My Radar (For Coding):

• DeepSeek V3.1: An open-source model showing impressive coding potential. Watching this one closely!

How I Actually Use Them (My Workflow)

It’s not just about having tools, but knowing when and how to use them. Here’s a peek into my typical process:

✅ General Q&A / Quick Info: GPT-4o in ChatGPT app. Need a quick fact or summary? This is my first stop. Via the ChatGPT mobile or web app.

✅ Quick OCR (Text from Images):GPT-4o in ChatGPT app. Since I watch a lot of educational YouTube videos, I often pause the video, take a screenshot, paste it into ChatGPT, ask “What’s the text in this image?” -> Bam! Then save the text to Notion (note-taking app). Super handy for quickly extracting text from images.

✅ Deep Research & Initial Idea Structuring:GPT-4o in ChatGPT app or gpt-4.5. This function isn’t available in the free version. You specify what the model should research, and it starts massive information gathering from the internet, finally condensing it into a coherent whole. This usually takes about 20 minutes!

Deep research

✅ Strategy Development: ChatGPT-o1. If I need strategic thinking from the above information, I switch to the o1 model after completing the previous step and trigger reasoning.

✅ Real-Time Web Info: GPT-4o with Search. When I need current events or information that goes beyond the model’s training data.

✅ Coding: Claude 3.5 or Claude 3.7 (usually via Cursor IDE). Fantastic for generating, explaining, and refactoring code. If I get stuck or need a different perspective (especially with more complex agent tasks), I switch to Windsurf AI IDE, which often helps me move forward with a problem.

✅ Image Generation: GPT-4o in ChatGPT app. Quick and easy for generating images right within the chat interface.

Create image

✅ This is my winning combination:

1. Start with ChatGPT-4o or ChatGPT-4.5 (deep research) to develop an idea “at the desk”.
2. Then go for a walk and switch to Voice Mode ChatGPT-4o in the mobile app within the same conversation.
3. Talk hands-free about the idea and AI helps me think. Sometimes I say: “Critique all the information so far and highlight potential problems.” Then continue the conversation about how we can avoid these. Finally, I add: “Please summarize the entire conversation into main points.”

Voice mode

This process is the main reason why I currently won’t switch from the mobile ChatGPT app to any other.

Claude lacks internet access and voice mode.

However, Google with the Gemini model and mobile app is on the right track - if you have an Android phone, you can test it via “OK Google” (hold the power on/off button) or by directly installing the “Gemini” app.

Paid vs. Free:

Is paying for AI worth it? For me, absolutely — for the right tools.

Here’s my current paid stack:

YES - Worth Every Cent:

• ChatGPT Mobile App (~€20/month): As mentioned, this is currently an essential app for me, and I don’t want it switching to inferior models like it does in the free version + missing the deep research function.
• Cursor AI IDE (~€20/month): Essential tool if you plan to code.
• Windsurf AI IDE (~€10/month): Right behind Cursor IDE in popularity and also an excellent tool.

NO (doesn’t justify the cost for me right now) - Cancelling Next Month:

• Anthropic Paid Version (Claude models) (~€20/month): Although Claude itself is fantastic, I can access it effectively through Cursor. I can also access MCP servers through Cursor or Windsurf IDE.

---

The Takeaway?

There’s no single “best” AI. The magic happens when you build your own workflow, leveraging the strengths of different models for specific tasks.

Stop worrying about picking the “perfect” tool. Start experimenting! Try the free versions, see what clicks for your needs, and don’t be afraid to mix and match.

Until next time.
Primož

P.S. What’s your go-to AI tool right now, and why? Hit reply and let me know — I’m always curious to hear about other workflows!