I didn't have some dramatic first moment with ChatGPT. I just started using it. At first, probably like most people, I was basically using it like a smarter Google. Ask it a question, get an answer, move on. Then I kept using it. And the more I used it, the more I figured out what it could actually do.

That doesn't mean I think it's magic. It's not. But the gap between someone who uses it well and someone who doesn't is real.

What it actually does well

Rubber ducking. If I've got a half-formed idea and want an opinion on it, ChatGPT is good for that. It doesn't have my same blind spots, and sometimes just framing the problem clearly enough to explain it is half the work.

It's also useful for writing. Not code. Emails, messages, anything where I know what I want to say but the words aren't coming out right. I'll describe what I need to communicate and have it take a pass. Saves me from staring at a draft for twenty minutes.

And honestly? I use it to make me a playlist every morning so I don't spend half an hour having an existential crisis about what music I'm in the mood for. That might not be the use case OpenAI had in mind, but it works.

For context, I've also been using GitHub Copilot heavily. The two tools occupy different spaces. Copilot lives in my editor, close to the code. ChatGPT is where I go when I want a second opinion or need to think something through out loud.

What it does not do well

It hallucinates. That's still true. It'll confidently cite something that doesn't exist, or generate code that looks right but quietly doesn't work. It's gotten better at this. It's just that "better" still includes occasionally inventing things with complete confidence.

The thing that gets me more is when it confidently tells you it can do something, you say "great, go ahead," and it responds with "I can't actually do that, but here's a guy who can." Very helpful. Thanks. I'll just go ask the other AI then.

Privacy is worth thinking about seriously. By default on Free and Plus plans, your conversations can be used to train OpenAI's models unless you opt out in settings. OpenAI updated its privacy policy in February 2026 alongside the launch of ads on the Free tier. If you're working with anything sensitive (client code, proprietary logic, internal documents), know what's being retained. Enterprise and Team plans don't train on your data by default, but most individual users aren't on those tiers.

The pricing question

ChatGPT currently offers several tiers: Free ($0), Go ($8/month), Plus ($20/month), Pro ($200/month), and business plans beyond that (full breakdown).

For most developers, Plus is the sweet spot. You get GPT-5, generous message limits, image generation, and access to advanced voice and Deep Research. The Free tier works for occasional use, but the message caps and ads make it frustrating for daily work. Pro at $200/month unlocks unlimited everything and o1 Pro mode for heavy reasoning tasks. Unless you're pushing it to its limits daily, that's a hard one to justify.

Tips for getting started

Be specific. More context upfront gets you better output. "Write me a function" gets you something generic. "Write a TypeScript function that takes an array of blog posts sorted by date and returns the three most recent non-draft entries" gets you something useful. OpenAI's own prompting guide emphasizes this.

Ask it to ask you questions first. Before a complex task, try: "Before you start, ask me the five questions that would most improve what you produce." The questions it asks will surface things you forgot to specify. The output after this is almost always better.

Give it your voice. If you want it to write something that sounds like you, paste two paragraphs of your own writing and tell it to match the tone and style. This works better than trying to describe your voice in the abstract.

Use Memory deliberately. ChatGPT's Memory feature lets it retain context across conversations. Tell it things like "I work in TypeScript and Next.js" or "I prefer concise responses" and it'll carry that forward. You can review, edit, or delete stored memories in settings.

Verify everything. Treat its output like a first draft from someone who's confident but occasionally wrong. Don't ship code without reading it.

What I actually think

The people who get the most out of it are the ones who know what they want before they start typing. If you go in with a clear problem, you'll get a useful answer. If you go in hoping it'll figure out what you need, you'll get something vague back and wonder what the fuss is about. It's not the tool's fault. That's just how it works.

Best way to learn it: use it for a week on actual things you need to get done. Not demos. Real work. You'll figure out where it fits faster than any guide will tell you. Including this one.

Working on a project where AI tooling could speed things up? Let's talk.