I've built a product for fun and personal use, but also to see how much sense fullstack and/or product builder/developer/maker/engineer profiles make these days.
It's Abelard, an assistant/tool that helps you take care of your plants, whether you have a couple and you're new to all this, or you have thirty and you vibe-plant your way through caring for them; it also lets you keep the history of every plant in your pocket, so you can see how they've grown.
It isn't a product meant to be sold, nor to explore the GTM side of things, but it's a good example that lets me put into practice the more than 12 years of work and study I have in my craft, mixing what I've learned about design, technology and product. It lets me take on the roles of front end, back end, AI, systems, devops, design and product, and it even opens the door to combining software + hardware + AI with a mini bot or smart planters that help you with your plant-care tasks (something I'm also already exploring and that I think has more future than ever).
The conclusion: Yes, the coding part is solved, but with nuances.
And other lessons we'll go through:
- Vibecoding is not the same as agent-guided software development (neither is better than the other, they're different things, and both are useful and have their place).
- The false sense of learning with AI, where it feels like you learn much faster, but maybe you don't.
- The SaaS/Product world has changed, not died. I don't think everything will be replaced in the short term by software built by end users themselves (beware of thinking your echo chamber is reality, which happens to me too).
- Again, some opportunities and doors close with this technological shift, but as programmers and/or curious people, we're privileged and more doors are opening for us than ever.
Let's get to it
The idea that you can build a SaaS or small throwaway tools with AI... go for it, I build plenty in my day to day. Now, if you have a bit of common sense and can do the math, unless it's purely for personal enjoyment or learning, it does NOT pay off to spend time and tokens on a tool or system that other people have already thought through and refined. We fall into the trap of thinking our system or way of doing things will be better if it's custom-made, but maybe not… The whole point of paying for software that's already been thought through and worked on is that the hard part, the thinking and reaching solutions, is already done or being done by someone else, other people or entire teams.
Vibecoding vs. agent-guided development
There's a vital distinction to make here: vibecoding is not the same as agent-guided development. The first is throwing a prompt at the AI, going with the "vibes" and trusting that the code works, accepting black boxes. The second is collaborating with agents while you act as the architect. You're the one who defines the structure and supervises the result. And this is where the weight of the craft comes in: AI, even if it seems so, doesn't make us all equal, it's a multiplier. If you've been at this for a while, what you learned before AI gives you calluses that are now worth gold for developing that instinct, which is more valuable than ever, so it acts as a multiplier. If you're a junior, be curious and use it as a multiplier too: try things, research, ship things.
The danger of creating being cheaper than ever, ending up like Crimson Desert, and thinking anyone can build things that make sense
If you follow video game releases at all, this game came out not long ago, and it suffers precisely from one of the problems that come up when you start tinkering with AI: you get carried away building and end up with an app that does everything and nothing. And even though it has brilliant parts, because it has so much, it's quite complicated to use.
You have to know when to stop.
Then there are the people who say "everything will be voice" and fire off posts or tweets saying "UI is dead" to scrape some likes or views and carry on as if nothing happened, when it's a debate or topic we could spend hours and days discussing. Are there problems that have a UI now because it's been the best tool to solve them, and won't need one when the technology allows? Yes. Does this mean NOTHING will need a UI? Well, I don't know what you understand by UI (User Interface), but for me there are certain systems or problems that, without a polished UI and well-thought-out UX, I honestly can't see how they could be solved. Voice is part of the game, it's not exclusive.
Nor do I buy that the solutions will sit right at the LLM frontier. We're losing the plot and suffering from an echo chamber at insane levels in this industry… lots of people only hear about AI on the news, some more have tried Gemini or ChatGPT as a chat, and that's as far as they go. A few days ago I was reading a news piece showing the percentage of people who pay (which doesn't mean they use it much better than the rest) and it was far fewer people than we think.
Are you going to tell me that an average user will be able to explain to an LLM how to generate software tailored to the problem they have? If so, tell me how you see it; I'm all ears, make your case. My position changes as I explore and read every day, and I'd sincerely appreciate it. Models and tools get more powerful every day, but I struggle to see just anyone implementing solutions, not only because of the technology side, but because of the human side: people pay so they don't have to complicate their lives and research, and creating means thinking and getting your hands dirty.
OK, so now what? How do we navigate this? What am I doing?
I wish I knew how to answer that, this is changing more than ever, so I'm trying to learn and improve in areas I think will make me stand out:
Problem 1: Keeping the ability to learn, and the loss of focus, which matters more than ever
And working with AI doesn't make it easy. What do you do while the agents work? Do you review, or work with several sessions at once? Are you training your brain with other things so you don't get dull or lose your ability to think and learn?
What I'm doing: Paying attention to focus and to reading on paper, away from screens. I'm reading La Biblioteca Polymata by Val Muñoz de Bustillo and Viaje by Sergio San Juan Escarramán, and staying optimistic about the future with books like Tiempos de Futuros by David Alayón. I'm also cutting down the time I spend on social media. X, for example, is a good source of news, but it's a constant bombardment of a thousand things: sign up for newsletters that sum up what's new, like Café con Codely by Codely, or use agents to summarise, every day or week, whatever interests you from the sources you point them to.
On top of that, narrowing down what I want to learn and putting time into it. I set up blocks where I allow myself to explore and research, and others where I try to have hours of deep learning.
Problem 2: What's worth learning now?
What I'm doing: Staying a generalist and keep learning things I think will later let me connect the dots and better solve the problems I face; there's no other way than to trust and stay positive.
- Learning more about product with the Product Management with Applied AI programme by La Nave Nodriza.
- On the tech side, using Claude with /teach by Matt Pocock whenever I need to learn new skills (both tech and other topics), and good structured courses like the AWS and Cloud one by Commit Academy, which also make sense and have their place.
- In general, as I mentioned with Polymatas and Viaje, learning things both inside and outside product and technology, to keep improving the ability to truly learn at an ever faster pace and to offer different approaches to the problems I may face, and to keep feeding my curiosity with side projects or by researching topics that before AI I saw as impossible or very hard to get into.