Who's been using Kim K3? I tried it for a couple of days. I hate the output - it's just too verbose OOTB. I think the output token cost is what puts it far behind the closed models (although the AI hype bros have most of us believing it's near Fable 5 and GPT 5.6 Sol). Interested to know other's experience.

Has anyone else tried T3 Code + T3 Connect? I'm absolutely loving it. My main machine is an i9 with an RTX 4070 and 32 GB of RAM, but AI coding agents can still eat through memory pretty quickly. So I started "borrowing" the computers around the office that weren't being used: a Mac M1 with 16 GB of RAM and an i5 Linux box with 32 GB. Now I'm working across all three machines without bringing my Windows PC to its knees. I honestly feel like I've become the master of my own little compute cluster. 😂 Productivity has gone through the roof, and it's been a huge improvement over the Tailscale orchestration setup I was using before. Curious to hear how everyone else is using T3 Connect.

I'm curious what you all think of using Grok. I have to admit I quite like the UX, and also the Grok Build coding agent. I'm quite impressed by it for research with something like Grok Heavy, and also for planning out and implementing features as a worker model, not to mention how fast it is :) I wouldn't be surprised to see 4.6 coming in early to mid-August. I was playing with Grok 4.5 a little yesterday and used the Codex CLI with a Sol model for verification and review. I feel it's not as powerful as something like Opus 5 or Sol. For Sol, I still have the feeling that it wants to write too much code. And I really don't like talking to Opus 5 yet, in terms of its verbosity and the language it uses, but I like the code it writes. I do really feel OpenAI has some work to do on the harness to really make these models shine.. So right now I've gone for the Codex mid-tier with Grok Heavy (they had a nice offer for 3 months), and I can use some Claude for work as well. Maybe I should give Cursor Ultra a try once again in the future. I haven't tried that in a long time. I feel like OpenAI is really generous with their limits, though. I keep hearing that it's not about changing models and trying new ones all the time, but more about what you ship. I do agree with that: to commit to the tooling you're comfortable with and go with that. It's just that there is a lot of choice available and hard to pick and choose sometimes. But it makes me curious to keep learning as well :)

Can someone please validate that using Grok Build gives you free access to X API? That would be awesome!!

Greetings everyone! My name is Paul. I really enjoy tinkering with AI and incorporating it into my workflows when possible. I am excited to be around others who share the same enthusiasm. I might be new to this community but I have been learning from Ray and community members for some months now. I must say, Ray brings integrity and positivity into the AI space. I look forward to collaborating with everyone!