Theo - t3․gg
AI coding assistants, LLM APIs, and developer platforms with a focus on usability and developer experience.
Nutrition Label
Theo provides high-energy, opinionated coverage of the AI engineering landscape, prioritizing "vibe checks" and live usage over scientific benchmarks. While his titles often lean into sensationalism or satire, his actual analysis offers grounded, practical insights for developers. Transparency regarding sponsorships is consistently excellent, and he frequently demonstrates tools live to highlight real-world friction.
Strengths
- +
- +
- +
Notes
- !Titles are often satirical or sensational; the actual verdict is usually more nuanced and critical.
- !Analysis relies on live usage and "vibe checks" rather than controlled scientific benchmarks.
Rating Breakdown
Breakdown across the key dimensions we rate. Methodology →
Recent Videos

Are you f**king kidding?

It's finally here.

gpt-5.4 is really, really good

The drama never ends...

OpenAI uses Elixir + Coding CLI showdown

I can't believe nobody's done this before...

I'm so f***ing tired of Obsidian.

Cursor, Claude Code and Codex all have a BIG problem

Software engineering is dead now

Trump actually threatened Anthropic (this is bad…)

They cut Node.js Memory in half 👀

My new app is really stupid (I wrote none of the code)

Anthropic is lying to us.

Delete your CLAUDE.md (and your AGENT.md too)

I Made a New Thing
Why this rating
Evidence receipts showing why each dimension is rated the way it is.
“So I actually got to ask Sam Altman a question... here is the clip.”[0:05] →
The video delivers exactly what the title promises: footage of the creator asking Sam Altman a specific question about coding tools.
“Thank you Dnsimple for sponsoring! Check them out at: https://soydev.link/dnsimple”[Description] →
Explicit disclosure of sponsorship in the description, meeting high transparency standards.
“It just hallucinated... It thinks it ran the command. It didn't run the command.”[04:12] →
Demonstrates a direct problem encounter where the AI tool fails to perceive the terminal state correctly, showing real-time friction.