SimpleSolutions.Technology

Recovery is messy. Healing isn’t linear. But showing up for 90 days? That’s power.

I just finished building a free web-based companion to the 90 Days Strong Journal — no sign-ups, no tracking, just you, your thoughts, and a quiet place to get honest.

🔗 90daysstrong.simplesolutions.technology
📘 Print version available on Amazon

If you’re just starting out, starting over, or starting again — this is for you.

90 Days Strong – Sobriety Journal

🧪 I accidentally taught myself data science trying to figure out which LLMs are worth keeping on my machine.

A few weeks ago, I downloaded LM Studio…

Then downloaded 30+ local LLMs…

Then asked myself, “Okay, now how do I know which one’s actually good?”

So I:

Created a test rubric with zero-shot and one-shot prompts

Scored 31 different models on logic, math, creativity, code, ambiguity, and more

Wrote the results into a CSV

Built a Python + pandas notebook to analyze and visualize the scores

And published it all to GitHub (with Colab and nbviewer links to explore it live)

🧠 The goal? To find the best LLM for my usage habits — not just generic benchmarks.

🔥 The result? I ended up diving headfirst into Jupyter, matplotlib, and real-world evaluation design.

This is part of a much bigger project I’m launching soon. But for now, I’m proud to say:

📊 This is my first real data science + LLM experiment.

✅ And it’s live, transparent, and open-source.

→ Blog: https://chicanoinparis.github.io/Prompt-Engineer-Portfolio/2025/06/19/my-first-dive.html

→ GitHub: https://github.com/ChicanoInParis/LLM-Prompt-Evaluation-Lab

🙋🏽‍♂️ Got suggestions? Favorite open LLMs to try?

💬 Drop me a comment — I’m learning in public and building every day.

GitHub - ChicanoInParis/LLM-Prompt-Evaluation-Lab: A collection of test prompts run through 30+ local models, with performance notes, screenshots, and a rubric developed across multiple use cases. A collection of test prompts run through 30+ local models, with performance notes, screenshots, and a rubric developed across multiple use cases. - ChicanoInParis/LLM-Prompt-Evaluation-Lab

Like most people, I have dabbled with AI using ChatGPT, Gemini, and a few others.

I found myself wondering, since DeepSeek is open source, shouldn’t I be able to use my own AI on my own computer without it having to go out onto the interwebs? Being curious by nature I did some googling and with help from my AI friends I was able to set up LM Studio and download a few LLMs.

There are SO MANY. And I don’t know enough to know them apart so…

I decided to test them out.

No, I have never worked with LLMs prior to the last almost two weeks. Yes, I am more than a little bit intimidated and struggling with a touch of imposter syndrome.

But I am nothing if not intrepid.

Also, it has been a lot of fun so far.

So, come along if you want. Maybe I’ll give you a good laugh. Maybe I’ll inspire you to try something you think you have no business taking on. For me, the worst case scenario is I learn something. Best case scenario? There really is no limit to that.

R

Stay Curious, Be Intrepid | Prompt Engineer Portfolio I found myself wondering, since DeepSeek is open source, shouldn’t I be able to use my own AI on my own computer without it having to go out onto the interwebs? Being curious by nature I did some googling and with help from my AI friends I was able to set up LM Studio and download a few LLMs.

Cyber Monday Sale coming!