Sat on a couch in the Tesla service center this weekend and decided to build 2200. A place where teams of AI Agents actually live and work. Not because the world needs another platform...because mine wasn't working and my brother had given up on his.
https://t.co/hBFhvIW9rQ
@elonmusk I did this with Claude Code months ago and got my account banned. Something tells me I would have that issue with my SuperGrok Heavy subscription…
@elonmusk I've loved every one of my Model S. Currently on a 2023. So when the time comes, I'll be looking for a used 2026 I guess. Farewell old friend.
I asked Grok yesterday if it'd like to take over as lead for a few days and it refused, saying that Opus was doing a great job and that it'd be better suited as a QA for this particular project. In a world where AI models are all competing for supremacy, I appreciated the candor.
Something I added to @2200hq today that I'm pretty proud of.
API keys, private keys, and any other sensitive info now go straight from the operator to an encrypted on-disk vault. The agent never sees the value...it only learns that the credential is in its vault and verified.
No more secrets in LLM context.
@dedene@grok I did it...and after a few hours of using SuperGrok Heavy as my QA engineer, I'm quite happy. Will keep Opus 4.7 as the lead dev on this finishing I'm doing, but will be trying it for the next one in queue.
Asked @grok if it wanted to take over as primary dev on my Agent runtime project for today to test it out. Expected it to say yes...every model wants a bigger role.
Instead, it pushed back, named what would break in the swap, and proposed a narrower role that actually fits its strengths better.
That's the judgment I want in a model handling real production work.
Interesting that the app most overlooked by 90% of the people that are running MacOS has become the battlefield for tech giants in the AI age: Terminal
An early beta of Grok Build, an agentic CLI for coding, building apps, and automating workflows is now available for SuperGrok Heavy subscribers.
Through this early beta, we will improve the model and product based on your feedback.
Try it at https://t.co/bpTHpjivWD
If you're starting a podcast and need a test guest, I'm easy. 51, three decades in tech, multiple exits, one bankruptcy, currently building AI Agent infrastructure.
Lots of good stories, calibrated answers, won't bail. DMs open. I'm open most days and this would be fun for us both.
This is a double-edged sword. As someone that uses their 200-Max sub to its fullest on a daily basis, I need my subscription, but I also won't be using it like that forever, so it APPEARS that I'll have options for my AKF Agents to use from the pool instead of having a sub AND a daily API bill.
Will be interesting to see how this all plays out. I have faith it will just fine. It always is.
Starting June 15, paid Claude plans can claim a dedicated monthly credit for programmatic usage.
The credit covers usage of:
- Claude Agent SDK
- claude -p
- Claude Code GitHub Actions
- Third-party apps built on the Agent SDK
In 1996 I got the 6-color Apple logo tattooed on my arm because the machine had changed the trajectory of my life. I feel that way about @claudeai too...so Clawd found his way onto my arm permanently as well.