There are approximately nine thousand AI tools lists published every week. I know because I’ve read too many of them, and they’re almost all the same twelve names in a slightly different order, written by someone who spent forty minutes with each tool and then ranked them.
Read more: The AI tools I’d actually tell you to use in 2026I’ve been at this long enough to have strong opinions and a graveyard of subscriptions I cancelled. This is what I’d tell you if you asked me over a drink, which means some tools get a paragraph and some get a sentence, based on how much I actually care.
One rule before the list: test any tool on three real tasks from your actual life before you pay for it. An email you struggled with, a doc you had to summarize, something you coded or built. Twenty minutes of that is worth more than any review. Okay. List.
ChatGPT is still where I’d tell most people to start, because it’s reliably good at everything, which matters when you’re still figuring out what you even want from these tools.
Claude is what I reach for when the writing actually matters. The thing I didn’t expect when I started using it: it’s noticeably more willing to say “I’m not sure” instead of just making something up confidently. After a while, that starts to feel really important. It’s also better at holding onto context over long documents without quietly losing the thread halfway through.
Perplexity is a search engine that actually reads the pages for you. Whenever I catch myself about to open five tabs, I try Perplexity first. It shows its sources. You can click through. That’s the whole pitch, and it’s enough.
DeepSeek, if you haven’t tried it, try it before paying for anything else. Open source, launched a new version in April, genuinely good. Free via the chat app, very cheap on the API
For the coders
Cursor is what most developers I know have switched to. It’s VS Code under the hood so the transition is low-friction, and having an AI that actually understands your codebase rather than just autocompleting variable names is a different experience. Windsurf is cheaper and has a better free tier, I’d try it first and only pay up for Cursor if you find yourself wanting more.
GitHub Copilot gets a mention because if your company is already paying for it, you probably aren’t using it as well as you could be. Learn it properly before shopping around.
Claude Code is its own thing and I want to be clear about what it is: you give it a task, it goes and does it, for example, opens files, writes code, runs the tests, reads what broke, tries again. You come back and review what it produced. It’s not magic and you still need to understand the output, but the kind of thing that used to eat an afternoon doesn’t anymore.
For building things without coding
Lovable is the one that still surprises me. You describe what you want, you get a working web app with a link you can share. Real code, not a mockup. I’ve watched people with no technical background ship things with it and I still find it a bit hard to process. The code is clean enough to hand to a developer if you want to keep building.
Bolt if you want to find out fast whether an idea is even worth pursuing. Rougher output, much faster. Good for killing bad ideas before they cost you anything.
Replit Agent if you’re building internal tools, like, a dashboard, something for your team and you want the database and hosting handled in the same place. Boring compared to the others. Probably the most useful.
Images
Midjourney is still the best if you care about the image actually looking good. V8 landed earlier this year and the gap widened. The learning curve is annoying but real.
Adobe Firefly if someone in your company has ever said the words “IP concerns” it’s trained on licensed content, so you’re covered. The output isn’t always as beautiful but “legally safe” is a feature with a very specific value.
Ideogram for anything with text in the image. Posters, thumbnails, social graphics. It’s the one thing Midjourney still fumbles and Ideogram handles reliably.
Magnific for making an image you like into something print-quality. It fills in realistic detail when upscaling instead of just blowing up pixels. Last step, not first, but worth knowing about.
Video
I’ll be honest: this whole category is moving too fast for any list to stay accurate for long. What I can say is Veo (Google) is the strong all-rounder, Kling gives you similar quality at roughly half the cost, and Runway is what you use when you need to actually direct something with specific camera moves, consistent characters, that kind of control.
HeyGen is different from all of them. You write a script, pick an avatar, get a finished talking-head video with synced audio in whatever language you need. No camera. No editing. OpenAI and PepsiCo use it for training videos and marketing. It’s well past the novelty phase
Audio
ElevenLabs for voice. Huge library, clones from a few minutes of audio, not expensive. If you make any kind of voiced content, this is just the infrastructure choice.
Suno if you want actual music. Full songs, lyrics, vocals. I still find it slightly surreal that this is a thing you can do.
Descript is criminally underused outside podcasting. You edit the audio by editing the transcript. It also cleans up bad recordings automatically. If you’ve ever had to sit through audio editing, you know why this matters.
The ones most people haven’t found yet
NotebookLM, Google made this and it’s free and almost nobody uses it. You upload your own documents, ask questions, get answers that come only from what you uploaded. Not from the whole internet. Just your stuff, cited. For anyone whose job is reading and synthesizing a lot of information, this is the tool.
Gamma for presentations. Give it a topic and a slide count. Get a designed deck. It takes about a minute. For how often most people have to make decks, this one pays for itself in the first week.
Granola records your meetings through your system audio, no bot, nothing joining the call and gives you a summary and action items after. Small thing. Appreciated every single time.
What I actually pay for
A coding tool, Granola when I’m in a meeting-heavy stretch, a voice plan, and a general purpose ai tool on the free tier. Total is less than a dinner. The whole professional setup doesn’t have to cost much if you’re not just subscribing to whatever’s getting the most press that month.
The people I’ve seen get the most out of this stuff aren’t the ones who’ve tried everything. They picked two or three things, got actually good at them, and stopped looking.
Leave a Reply