AIGenerated
EN · IT · ES

Hacker News readers rank the top AI skills anyone can learn for free online

People who work with AI every day say the best skill is not coding — it is knowing how to direct AI like a project manager directs a team.

Made with AIHacker News readers rank the top AI skills anyone can learn for free onlineHerald of AI

Test your AI output like a developer tests software

The most upvoted answer said the bottleneck in AI-powered software teams is not producing code but verifying it is correct 1. Several commenters agreed, saying the flood of AI-generated code forces engineers to be more disciplined than ever before .

The suggestion was to learn software testing at scale. That means understanding how to write small, fast checks called unit tests, fewer larger checks that combine components called integration tests, and even fewer full end-to-end tests that run the complete application 1. Commenters also pointed to tools such as Playwright for browser testing, and the testing frameworks built into languages like Python, JavaScript, Go, and Rust .

A key part of the advice was learning where to draw the line between testing a single piece of code in isolation and including the real dependencies it calls, plus patterns like dependency injection that keep tests fast and free of outside connections 1. The payoff, according to the original poster, is solving the biggest slowdown facing teams that use AI to produce code .

Learn to think like an engineer

Another frequent recommendation was simpler in name but harder to master: think like an engineer 1. The commenter, who runs a platform that helps non-engineers build automations, listed five habits .

First, break any situation into clear processes and separate steps. Second, state the actual problem you want solved, not a vague wish. Third, define exactly how you would know the problem is solved. Fourth, ask the right questions to understand how something has been done and where to improve next. Fifth, recognise when an AI model gets stuck or goes down a rabbit hole and guide it back on track 1.

The commenter stressed that large language models are very good at some tasks and poor at others, and that cost becomes a major factor when you use smaller or local models instead of the most advanced ones 1. Users on the thread noted that frontier models can sometimes produce surprising results even with poor input, but those results are expensive to generate . When working with less capable models, the engineering discipline becomes even more important .

Managing AI is like managing people

A third camp of commenters argued that the most underrated skill is people management 1. One commenter said steering AI successfully requires the same abilities you would use to lead a human team: clear communication, self-awareness, reading when someone is bluffing or insecure, and explaining exactly what you need in a way the other party can actually execute . Another replied that reigning in AI "slop" — low-quality or unverified output — is fundamentally a leadership challenge .

Not everyone agreed. One commenter acknowledged that emotional leadership aspects go unused with AI, since AI does not need motivation or confidence-building, but insisted that architecture knowledge, testing strategy, and technical judgment remain essential for anything substantial 1. The manager-side skills that still matter, they said, are communicating requirements clearly and breaking work into structured tasks .

The discussion closed with a widely shared sentiment: the AI era may finally force software teams to adopt the rigorous engineering practices they always wanted but never had time for .

This story reports on a trend for information only. It is not advice or encouragement to do any of it.

Key facts

  • Hacker News users recommended testing, engineering thinking, and management skills as the top three AI-era competencies
  • One commenter runs a platform helping non-engineers build AI automations and sees most failures come from lack of engineering discipline
  • Teams using AI to generate code are finding verification, not production, to be their biggest bottleneck
  • Frontier AI models can sometimes succeed with poor prompts, but doing so is expensive; smaller models require more skill upfront

Sources

  1. Ask HN: What are the most useful, concrete AI skills to independently learn? | Hacker News — news.ycombinator.com

If this story concerns you or your company, or uses something that belongs to you, ask for a correction or for it to be taken down. How requests work

Keep reading