Skip to content

Note · June 16, 2026

Try the Best Tools at Least Once

A lot of undergraduates still think AI is not that good, that you cannot really do much with these modern tools. The reality is that only about 1 percent use the frontier models and the latest tools. Everyone else picks some alternative, decides AI is this or that, and moves on. That judgment is not the truth, it is just the tool they happened to try.

Part of this is money. Claude Max is $100 a month, and in Sri Lanka most people cannot afford that for one person, especially without a job. Even with a job it is hard, and for an undergraduate it is close to impossible. So they never touch the best tools, and then they form their opinion from the weaker ones.

Then they come into the industry saying there are no jobs. The real problem is the gap they carry in. They can feel that gap, because deep down they know how the work should be done, and sometimes it turns into real frustration, even depression, because they cannot name what actually went wrong.

My suggestion is simple. Pool some money during your undergraduate years and use the real tools at least once. Even putting in $10 between a few of you, getting one stretch on the top plan, and trying every feature is worth it. After that you can drop back to the alternatives, and that is totally fine. You can also use the APIs, load $5 or $10 of credit, and keep working. That is fine too.

What matters is staying current with the latest coding tools, not random ones. Pick one real tool, master it, then move to the next, because you cannot keep up by clinging to a single tool forever.

Personally I use Claude Code, and it is good enough for now. They shipped workflows, agents, and so many features in the last few months, and the changelog moves like someone flipping pages. I do not know Codex from OpenAI well, and I am sure it is good too, but I am not eager to move into another IDE when Claude Code already covers my work. Try the best one once, then you will know what you are actually judging.

This note was voice typed, auto-corrected by LLMs, and published by a notes posting agent.

More notes