Загрузка видео...
Не удалось загрузить видео
"Why should I use T3 Code instead of Claude Code or Codex?" Short answer: T3 Code is way better for developers. Long answer:
326,525 просмотров • 1 день назад •via X (Twitter)
Комментарии: 31

every tool is the best one

Long answer is 41 mins, my attention span is 41 seconds

Can we get a longer answer that’s still text? I gotta feed the baby and she can’t see videos

For me, the biggest issue with T3 is integration. Building a frontend with it is genuinely harder than using the harness’s native UI. A lot of features feel nice on paper but not fully thought through. I’d love a lite version. No snoozed threads, no extras. Just T3 with the freedom to move a thread to Claude whenever I want. Until then, it feels like a pretty UI to interact with opencode, and occasionally with Claude and codex

revert the layout pls. was way better before.

My test for any coding agent: drop it into a messy 2 year old side project, not a fresh scaffold. Where does T3 Code pull ahead there, context handling or the review flow?

Can’t go back to using anything else. Haven’t heard the fan of my laptop in a long time. Thanks

The part that sold me is one orchestrator handing work off to subagents on different models side by side. Hard to go back to one model per window

The quiet shift here is where loyalty now sits. When the same subscriptions plug into any client, the model stops being the product and the shell becomes it. A three person open source team won ground simply by fixing the small frictions the labs considered beneath them.

Genuine question: does T3 Code run on top of an existing Claude Code or Codex login, or does it need its own API keys? That decides whether I could try it alongside what I already use.

T3 Code has become my daily driver. It's fast, the UX is great, and I even control it from my own agents. Thank you for building it! If I could ask for two things, I'd love prompt suggestions like Codex has, and sending voice messages, ideally on both mobile and web. Would those be on the roadmap?

T3 code is so bad you can’t even charge for it.

not going back to a terminal tbh

A few months ago, before I’d heard of T3 Code, I’d also thought about building something similar UI/UX-wise. Early on, the ChatGPT and Claude desktop apps had pretty bad UI/UX compared to Cursor, so I thought about building a universal UI/UX, where users can simply connect to their subscriptions . But I didn’t, because I asked my friend for approval ;)

Harness matters more than model now. Same model in two harnesses gives me very different PRs, mostly down to how much context it reads before writing. Let me try this.

Kudos to T3 code Earlier today I have 20+ agents from 3 different providers (Codex, Claude, GirHub Copilot) being orchestrated by Opus

bold claim with a 40 min video to back it up. ok I'll actually watch this one

I started using T3 Code 2 days ago, perfect timing for the video. Love the demo.

the simplicity and speed of T3 Code can really boost dev productivity

tried it, liked it at first, then the flaws piled up and i ended up back in the cli where I feel more in control of my context. Always have been terminal first but gui was an interesting experiment

saving the long answer for my next flight like it's a netflix show

Long answer is a 41 minute video. Short answer it is.

already paying for two claude max accounts, so the long answer only matters if that login works in T3 Code

What about superlogical, herdr? Are they obsolete?

the real test is what happens when two agents want the same file. whichever UI makes that boring wins

There’ve been a lot of progress since June. T3Code sucked before then.

I don't even need to watch the video. I am already sold. Unfortunately the agent-sdk bug 😢

I've been seeing a lot of people praising t3 code, even here in brazil... you're doing a great job

i especially love using my different subs, and have t3code spawn threads across different ones

What we don't want is to be part of the AI lab wars. We don't want to be trapped in any silos. We need an open source company OS, that let's us use the best models for the job, and let the agent get IDs, and interoperate.

Should I finally try it out?

