code is cheap, execution is everything.

some thoughts on building products as a cs major.

code is cheap. it used to be expensive a couple decades ago, but now with ai, almost anyone can build something. what matters more now is execution, choosing the right idea, and actually getting people to use what you've built.

i think one of the most important skills for a cs major is knowing how to market yourself and your product. getting technically cracked could definitely help, but if you don't know how to present your work, you're missing out on a lot of opportunities.

to begin with, find a problem you've faced in your own life, and vibe code a solution out of it. chances are, if you've faced a problem, someone else has faced it too. now whether they're willing to pay for your solution is a completely different question.

another thing i've noticed is how much of the startup market is just copying. one company comes up with something new, and suddenly 100 others are building pretty much the same thing with minor improvements. but that's also how innovation works to some extent.

with the recent advancements in ai, i think the opportunities are pretty insane. someone can technically build a working prototype by literally yapping their thoughts out to an ai. of course, building a serious, scalable product is a different story, and good engineers still matter.

more specifically, a market i see myself potentially working in is applied ai. i remember lucy guo talking about this in one of her podcasts, where the idea was to use ai to automate repetitive tasks for businesses and sell them the solution.

businesses don't necessarily care whether you're using a massive llm, a small language model, or some complicated agentic system. they care whether you're solving their problem and how much it's going to cost them.

the hard part is finding those problems and convincing businesses to try your solution. cold emailing could work, but even finding the right person to email is a problem of its own. so random.

this is also why i find slms interesting. if a smaller model can handle a specific task reliably at a fraction of the cost, why use a much larger one?

i'm still learning about all of this, and i haven't actually built a business around any of these ideas yet. but i do think the barrier to building software has dropped significantly. the difficult part is figuring out what to build, who needs it, and whether anyone actually wants to use it.

writing code is getting easier. building something people want? not so much.