I wanted to create a tool to help me learn Grok Build while I used it. I tried a few ideas then decided to give Grok the lead.
I asked Grok to create what would best serve xAI’s mission. Landed on turning new users into true high-agency builders.
Grok Forge is its vision.
@morganlinton Completely agree.
I think for me it stems from my desire of wanting to be apart of the push outward and to create amazing things. I don’t have a background in engineering, so a lot of work is needed for me.
@morganlinton Not gonna lie, I have thought about doing this too. My problem is I get so tied up in to wanting to know more laterally and not deep dive at scale.
@morganlinton@nikitabier Same. Been spending the last week getting back into x to learn ai, agentic coding and to see all the amazing things going on. Rockets included!
@Vivek4real_ If the engineers are 2x, 10x or 100x their output then obviously ai cost will be more than just human output.
If everyone is using it wrong then the cost is not worth it. The desired outcome is to significantly increase productivity.
@Tesla@grok What’s the chances of getting a Tesla map app for navigation and Grok voice for us non Tesla owners. Apple map seems to want to give me ads now.
@arimedai@tetsuoai@morganlinton Grok Build is still in early beta, so it’s not as efficient yet. But fixes and improvements are coming pretty consistently. Running with the heavy plan is still more cost efficient than other models.
@tetsuoai@morganlinton Glad to see this. I implemented better defined gates and review agents yesterday and had far better results in my phases. I’ll look into god-functions and pattern logic next.
I’ve gotten on average 10mins max on larger tasks. Ive been running separate tabs for agent reviews for major components and has been better at finding false claims for “completed” tasks.
Subagents in the main build have the same issue for saying work is completed. This is why I isolated reviewer agents in tabs.
I’m still learning to build properly as it is, but to me it’s still very powerful and promising. The experience feels more intuitive, the overall flow and responsiveness is great.
I’ve have used Claude and codex. I find codex builds better than Claude and has minimal errors compared to grok.
Adding absolutes and not to stray about what needs to be done has helped a lot with my interactions.
I need to see how the new updates feel.