…after my last post.
A lot has happened in my life: I changed my home country, worked in the travel, trading and renewable energies industries, played Skyrim and Fallout for days, contributed to Open Source and the Ruby ecosystems even more, managed development teams and taught lots of people how to be better in their development career, and shipped stuff in production as they say.
I should have kept updating this blog. But, I have one big issue with time management: I often dedicate 200% of my time on one task only. Bad thing. I’ll improve from there, because there is so much to share, learn and teach with everybody.
I am also sharing tech thoughts again in this blog because an old growing passion is taking a turn: AI. I want to deeply understand it, use it, and go beyond the buzz about it. I see a lot of new opportunities with the recent progress in AI. I hope what I learn here will be useful to others.
So now I got into AI, for whatever that means
… and it means a lot of different things for different people!😅
For now I truly think that AI is revolutionizing the way we build applications, and also the applications themselves. As a senior developer and tech lead, I must catch up on this. It is time for me to master whatever I can of it. I decided to step in it fully 1 year ago, and my journey so far has been full of surprises that can be summarized like this.
I wanted to understand and build AI
I love how AI has been built in many domains. In fact I was always amazed by neural nets and genetic/evolution algorithms since the 90s. Now I see a whole new scale of it. I want to learn about it so much.
I started by learning Deep Learning in depth (Andrew Ng, Andrej Karpathy and Yann LeCun) being great mentors. Neural nets, computer vision, GPTs, GANs, RNNs….
👉My Ruby playground for this learning is ruby-neural-nets.
I wanted to use AI
I realized that AI can help me a lot improve my development skills, when used properly. And that’s the trick. Using it correctly is challenging: you need a lot of money to spend on expensive models. It will become easier for sure in the years to come, but right now is just the mere beginning.
So I decided to improve my developer life with AI (daily tools, workflows…) that make me produce new products and features. This is when I discovered prompts engineering, skills and agentic systems. I played a lot with Cline, agents skills, and then multi-agents systems.
👉My Ruby playground for this learning is X-Aeon Agents Skills.
For now this is my take on AI
The AI buzz that I hear/read often is oversold. Creating everything from scratch with little to no help, inventing new things from mere ideas, this is clearly not happening. Sadly a lot of people believe it. Jobs are being decommissioned not because of the advent of AI. Instead, hasty managers think AI can replace people. They believe this will help them cut a lot of costs in the short term.
- The same happened in the 2000s with the dot-com bubble. Management decisions were taken to remove jobs and former business models just because a company has a my-company.com website which should guarantee visibility and success. Then the bubble exploded, most of those companies went bankrupt. It didn’t mean that the web was a useless fad. It just meant that a lot of managers believed the web would replace humans in marketing their ideas. Reality proved them wrong. The web was a revolution changing the way people work and market ideas. Yet, it has not removed the humans from the equation. It even created a lot of new job positions that did not exist before. I think the same goes for AI. Unfortunately we will have to suffer bad management decisions driven by the AI buzz for some time.
I realized that for now AI is mainly driven by vertical scaling: better performance is achieved by better models.
- Back in the 90s, it was the same pattern with databases: scaling with more hardware and budget. Better models need more tokens, resources and budget. I believe the future is in horizontal scaling for models as well. Agentic systems thrive at it, but they are still under-utilized and still rely a lot on expensive models.
The usage of AI is worrying from an ethical point-of-view. Like any invention, we see people that already use AI to scam other people. They steal others’ money, power or resources. They confuse others with deepfakes or realistic generated content.
- I believe that the answer to those issues will not come from a technical stand point. We didn’t neutralize bad usage of nuclear weapons by inventing a technical anti-nuclear-bomb weapon. Instead we managed to stop its usage using regulations, human relations and diplomacy. I think the same should handle ethical AI issues. For example by always crediting AI where it is due.
What’s next?
After 1 year, I am still only starting this journey. It is exciting and frustrating at times. It opens a lot of new opportunities that I only dreamed of a few years ago.
My goal is 3-fold:
- 🦸Be a better developer by finding and industrializing better ways to create applications.
- 🎮Master the underlying architectures of neural nets, generative AI and evolution algorithms to later build awesome projects.
- 🎁Share everything I learn and create as OSS, so that anyone can build on top of it.
I promise I will continue to share this journey in this blog, without waiting for the next 12 years 😉
Am I alone?
Is somebody still reading my blog after 12 years? Would love to read about you and your projects in the comments😅