I started web development 13 years ago, at 16, on a Nokia C1 mobile phone, not a laptop or MacBook.
I built websites on Wapka for fun and, if I’m being honest, to impress a crush. Then I abandoned it.
At the time, it felt like something I was simply doing, not something that would shape my future.
I went on to study Accounting at university and did well. I understood principles, systems and numbers, but I kept wondering why I was more interested in improving systems than simply understanding them.
Six years after graduating, I pursued a Master’s in Data Science and Artificial Intelligence.
Looking back, none of it was wasted.
Web development taught me how to build.
Accounting taught me how businesses actually work.
Data science taught me how to find patterns.
AI taught me how to turn ideas into working systems much faster.
For years, those skills felt unrelated.
Then I started building products.
Building CTRON Loop forced me to bring everything together. Understanding business workflows mattered just as much as writing code. Knowing how people make decisions mattered just as much as choosing a model. Solving the right problem mattered more than using the newest technology.
It made me realise that careers don’t always move in straight lines.
Sometimes the things you think you’ve left behind become the foundation for what you’re meant to build next.
Looking back, the Nokia phone, the Accounting degree, the AI master’s and everything in between weren’t detours.
They were preparation.
#ctron #ai #agent
GPT 5.6 Sol and GPT 5.5 are so good. They’re straightforward and fast, and most of the time I’m not worrying about token max or rate limiting, unlike Fable 5 and Opus 4.8
Typing every line of code was never what made someone an engineer.
Before the internet, developers relied on books and libraries. Then Google, Stack Overflow, GitHub and modern frameworks changed how software was built.
Now it is Codex, Claude Code, Cursor and AI agents.
The tools changed. The thinking did not.
The real measure is still whether you can understand the problem, make sound decisions, recognise bad output and build something that actually works.
Every generation mocks the tools of the next one.
Then eventually, everyone uses them.
With AI getting better at everything from content creation to military strategy, what’s one skill you’re doubling down on learning this year that *won’t* be replaced anytime soon?
Drop your answer below 👇
#Tech#AI#FutureProof#Learning
@sahildarz Exactly. AI still needs clear context, specifications and judgment. My point is that directing the work well is becoming as important as writing every line manually
Typing every line of code was never what made someone an engineer.
Before the internet, developers relied on books and libraries. Then Google, Stack Overflow, GitHub and modern frameworks changed how software was built.
Now it is Codex, Claude Code, Cursor and AI agents.
The tools changed. The thinking did not.
The real measure is still whether you can understand the problem, make sound decisions, recognise bad output and build something that actually works.
Every generation mocks the tools of the next one.
Then eventually, everyone uses them.
Lot of companies I talk to are quietly pulling back from just throwing everything at OpenAI or Anthropic.
Not because the models are bad, but because they’re realizing they’re handing over way too much of their actual workflow and data.
The open source + self-hosted route is looking more practical for anything that actually matters to the business.
Seeing this shift in your circles too or is it still mostly hype?
Been messing with agents the last couple weeks.
Even with all the new frameworks, getting them to do anything non-trivial reliably is still painful. Memory leaks, tool failures, context getting lost... it adds up.
Starting to see why some teams are just waiting for proper managed agent infrastructure instead of building it themselves.
Anyone else finding this way harder than the demos make it look?