You won't believe who won.
I challenged Claude to optimize Go code. AI: 61x faster. Me: 21x faster.
But I used 50% less memory and wrote cleaner code.
AI brute-forces. Humans profile systematically. For production systems, that matters more than raw speed.
www.youtube.com/watch?v=4ABo...
There is a HUGE difference. It's not about API keys, it's about context switching and constantly coming back to all your sessions, checking PR status, merging PRs, etc. OpenClaw just does it all, every single step. Start to finish. That's AGI, not AI which is claude code cli.
OpenClaw just autonomously worked 10 GitHub issues from one prompt—PRs, code review responses, CI monitoring, Slack notifications.
People think it's overhyped. They're missing the point. This is about autonomous multi-issue workflows, not just faster coding.
www.youtube.com/watch?v=VS9g...
How are people still letting an LLM write their foundational layer... learn about skills folks. Codex, Gemini, and Claude all support them. I have a skill that is ONLY for creating new projects... it knows EXACTLY how to create my foundational layers. It's the only way to do it
If you are wondering what my primary skills are, there are OSS here: I update them often (sometimes multiple times per day). Use at your own risk, they make me beyond productive. #golang #AI #LLM AnthropicAI OpenAI GoogleAI
If you are trying to "master" AI like you would other technologies, don't. Just keep up, try to find what works for you, and know tomorrow, that what you knew today won't be true tomorrow. I know how this is very uncomfortable for most of us in tech. We are used to finding a
I used to spend hours on the boring parts of GitHub issues: checking for duplicates, reading through old threads, writing boilerplate tests. Now I let Claude Code skills handle all of it. Here's my workflow: 1. Feed Claude the issue URL 2. It searches for duplicates
Today I told my openclaw horror stories about the internet and security, and asked it to protect itself. Just like I would talk to a junior dev. I think I've come full circle with AI...
🧵 Any API keys I use are limited scope and easily revocable. It's a sandbox. Treat it like one.
It's fun watching people stress about openclaw security. I have it running on a standalone box that has none of my credentials. It has its own email account, its own GitHub account (which I invite with limited access to repos I don't care if it trashes).
It's not that kind of Go...
I automated my GitHub issue workflow with Claude Code skills.
Duplicate detection → context gathering → failing tests → fix → verify
What took 2 hours takes 15 minutes.
The boring parts are now automatic. I just focus on solving the actual problem.
Walkthrough 👇
www.youtube.com/watch?v=-GKB...
I used to merge PRs only to find out hours later that CI failed.
Then I discovered one gh CLI command that changed everything.
Now my AI agent watches every PR in real-time and catches failures instantly.
The result? Zero failing PRs make it to main.
www.youtube.com/watch?v=N7nJ...
