← Back

The AI Tools People Actually Use (According to Twitter)

Marketing tells you every AI tool is revolutionary. Twitter tells you the truth.

I spent some time digging through the actual discourse—the casual mentions, the rants, the "this thing just saved me 3 hours" posts. Here's what people are really using.

The Current Champions

For Coding: Cursor is eating GitHub Copilot's lunch

Copilot is still everywhere, but the enthusiasm has shifted. Cursor—a VS Code fork with built-in AI—is the new darling. People love its chat-based debugging and multi-file editing.

Common sentiment: "Cursor is like having a senior dev in my IDE."

Copilot criticism is growing: suggestions feel generic, and Microsoft ownership raises privacy concerns. Codeium is picking up users who want something free and lightweight.

For Writing: Claude over ChatGPT

Claude 3.7 Sonnet has captured the writers. They say it "gets their voice" better, handles long documents without losing context, and doesn't give the robotic corporate-speak that ChatGPT often defaults to.

ChatGPT is still the default for quick tasks, but power users are migrating. The main complaint: "GPT-4o feels dumber than 3.5 Turbo—they're cutting corners for profit."

Claude's downside: no web access, weaker at code. But for prose? It's winning.

For Images: MidJourney, but people are annoyed

MidJourney v6 produces stunning results. Artists and marketers swear by it. But the Discord-only interface and $10/month minimum are driving people toward Leonardo.ai (free tier, desktop app) and Stable Diffusion forks on Hugging Face.

The ethical concerns about training data never go away: "MidJourney is stolen creativity with a subscription model."

The Migration Patterns

ChatGPT → Claude for anyone doing serious writing or research. Context retention is the killer feature.

MidJourney → Leonardo.ai for people who want free credits and hate Discord.

Copilot → Cursor for devs who want conversational debugging, not just autocomplete.

Hidden Gems

A few tools that keep coming up in "underrated" threads:

Perplexity AI — Described as "Google killer" for research. It cites sources and actually answers questions instead of giving you ten blue links. Researchers love it.

Descript — Edit audio by editing text. Podcasters are obsessed. "Turned my 2-hour podcast edit into 20 minutes."

Hugging Face Spaces — Free access to community AI models that sometimes outperform paid tools. The open-source crowd lives here.

Replit Ghostwriter — Copilot for beginners. Great for learning, hackathons, quick scripts.

What Developers Actually Think

Loved:

  • Cursor (conversational, context-aware)
  • Hugging Face (open-source, transparent)
  • Perplexity (research without SEO garbage)

Hated:

  • Google Bard ("suggested jQuery for my React app")
  • IBM Watson ("AI from 2015")
  • Copilot's privacy model ("MS owns your snippets")

The Takeaway

The discourse reveals a few clear trends:

  1. UX matters more than raw capability. Cursor is winning not because it's smarter than Copilot, but because the experience is better.

  2. Cost drives switching. Free tiers and open-source alternatives are pulling users from premium tools, especially for casual use.

  3. Trust is eroding. Privacy concerns, ethical questions about training data, and skepticism about big tech AI are pushing people toward transparent alternatives.

  4. Specialization beats generalization. People prefer Claude for writing, Perplexity for research, Cursor for code—not one tool for everything.

The era of "ChatGPT for everything" is ending. The toolkit is fragmenting, and users are getting pickier.

Adapt accordingly.

← More posts