... never the plan. The plan was research. This journey is 1.8 years old, and for the first 1.2 of those there was no product at all. Just me and my four friends: Claude Code, ChatGPT, Gemini, and Google Colab. Training runs. Benchmarks. Experiments that failed by Tuesday. Experimental products that will never see the light of day there's a whole graveyard of them on my hard drive. I wasn't trying to build a startup. I was obsessed with ...

... Fine-tuning AI models is still way harder than it should be. Every time I tried, I had to: - find datasets - clean messy data - write training code - debug notebooks - deal with GPUs Even with tools like Hugging Face or Colab, it s still frustrating. So I built something for myself: Miniclay AI. The idea is simple: You describe the model you want, and it handles everything. finds datasets cleans data (this is the hardest part) runs the training pipeline gives ...

... very easy) trick, let me clarify one thing: the goal here is to build a website not for ranking, but for testing . With that in mind, here s how to create any demo site in minutes using Google Colab. 1 Go to Google Colab

... like Next.js, Supabase, headless CMS , and act like that alone makes a product great. Me ? I used what worked. What was fast. What was already in my hands. Here s my entire setup: Odoo to run the platform Google Colab to make up for where Odoo falls short Canva for design (yes, Canva ) And a brain that wanted to build instead of scroll That s it. No fancy frontend framework. No cloud-native, serverless, containerized unicorn magic. Just tools ... ... modular, open, and already has a structure I could hijack and flip into something new. Is it perfect? Nope. Does it fight back sometimes? 100%. But it gave me just enough to build what I needed. Google Colab