new powers, new meanings

something curious is that i started seeing my first lines of code, in fact at 7 yo in my primary school, idk but for some reason in my computing classes we had an ubuntu distro instead of windows, the thing im really grateful to have a computer teacher with these preferences. and the major tasks i remember were about creating blogs with blogspot, at this time i saw html just as blocks to copy and paste to have widgets and custom components to create a cool blog.

years after at uni studying cs, i saw writing code just as the tip of the iceberg, behind them the data structure, algorithms and computational theory that in some way was math. i contrast this with the way i saw it when i was making my first startups, where code was just a way to solve problems for users and to try to be as scalable as it could be.

seeing from an algorithm perspective, humans always build tools to improve efficiency and do less, from the first stones used until the llms nowadays, the same with code, languages, frameworks, everything to help us humans to write instructions to computers. and the powerful thing with llms is that with hard work they can do daily dev tasks, like reading requirements, thinking about the problem, and acting, testing, and opening a pr to contribute to the code base.

many influencers say that senior devs will not be replaced, but i think many of them justify these arguments just from the perspective of one dev. of course a senior dev is more capable than a coding agent, but something that they forget is that scalable is the key and humans are a bottleneck. a swarm of agentic coding agents are more capable, as herbert a. simon said, complex systems present a hierarchy where components are organized in levels and subsystems.

today is the hardware behind cursor, claude, and many other agentic systems, but just like in previous years when necessity evolved programming into parallel programming, i think we are in this situation with solutions like superset, conductor, both are interesting projects. but i saw more technical vision for superset and a product perspective for conductor.

after reading more about superset and the minds behind it, my mind changed and i saw that we are near the end of normie roles for devs, and it really makes sense. and well, even in this path exists many challenges, it will not solve the entire software industry, and i think even llms are not the final solution.

cuz we use llms to generate lines of programming language code, but why if we didnt need this intermediate step? ai could create and program in a more low level way, but i think it wont really happen in the next years, we still need to verify and debug the code generated by ai, but the time they dont need us, probably this step will be more efficient.

the result of these advances u can see, lovable, claude, is mainstream not just for geeks, everyone who has a job with a computer at least now knows about this tools. this makes coding democratize the creation, but this also slows the development of engineers' skills. its a step like i said, we just make an optimization of creation, now we can focus on thinking and executing, but at a cost and with token limitations, this i think are the bottleneck.

not every country can pay for this "synthetic genius" as amoidei calls these foundation models, and its true. the need goes now for energy bc its the easiest right now, while we keep improving the efficiency of these models, compare with the human brain, these are more efficient resources.

now with this power, what geeks gonna create? another delivery app, another lovable? i realized that we have to be conscious to create something that matters, back to the physical and the sciences. energy, robotics, manufacturing, we are forced and grateful to be back to the physical world challenges.

and i make you this question, from aaron swartz

what is the most important thing you could be working on in the world right now? ... and if your not working on that, why arent you?