01What Is the Hermes AI Agent and How Does It Work?
The Hermes AI agent is a self-improving, self-evolving AI agent that doesn't just run tasks—it learns and improves as it works. Unlike tools like OpenAI Codex or Claude Code, Hermes gets better the more you use it, making it structurally different in how it operates.
In this episode, Julian Goldie explains that Hermes recently beat Codex in key speed benchmarks, and not by a tiny margin. This matters because speed directly affects cost: faster agents run more tasks on the same budget, while slower agents eat up tokens and give less output. So, Hermes is not just a minor upgrade; it represents a shift in how AI agents can perform.
02Why Did Users Start Switching from OpenAI Codex?
OpenAI Codex, a popular AI coding agent, launched with a lot of hype. Developers loved it, and business owners used it to build automations without needing a developer on their team. But over the last 24 hours, users reported performance drops: responses got slower, tasks took longer, and reliability wobbled.
The conversation on X shifted from "Codex is great" to "I'm switching." There were over 1,700 posts in 24 hours from people reporting the same issues. That's not a few complaints; that's a pattern. When tools degrade, users don't stick around—they move, and many are moving to Claude Code.
This shows a key point: the agent you use today might not be the best one next month. The gap between tools is shifting quickly, so you need to stay alert.
03How to Get Started with Hermes AI Agent
If you're considering trying Hermes, here’s how to think about it. While the episode doesn't give a step-by-step setup, Julian emphasizes evaluating it carefully. Here’s a practical approach based on the episode:
- Monitor benchmarks: Look for consistent speed wins on tasks that matter to your workflows, like client outreach or lead generation.
- Test it on a small project first: Don't switch everything at once. Try Hermes on one task to see if it's faster.
- Compare costs: Faster agents mean more tasks per budget, so calculate potential savings.
- Watch for updates: Hermes benchmarks are early, but the signal is real. Stay updated on new developments.
- Build your workflow in layers so you can plug in new tools without breaking what you have.
04How to Use Hermes AI Agent in Your Business
In the episode, Julian explains that Hermes is useful for tasks like client outreach, lead generation, lead follow-up, and data pulling. Because it's self-improving, it gets better at these tasks the more you use it.
To use it effectively, don't treat it as a permanent tool. Instead, build your workflow in layers. Julian suggests creating an agent operating system where you can plug everything in.
For example, if Hermes gets better, you can plug it into your existing dashboard alongside other tools like Claude Code or anti-gravity CLI.
This layered approach lets you benefit from new tools without rebuilding everything from scratch. You can also use tools like Codex's new feature for autonomous long-horizon tasks, where agents work for hours, all within your dashboard.
05Results and Comparisons from the Episode
The episode shows that Hermes beat Codex in speed benchmarks, but the benchmarks are early. The signal is real, though: Hermes is consistently faster on tasks that matter to real workflows, not just developer tests.
Julian also mentions that thousands of people are already publicly switching from Codex due to performance issues. He notes that this cycle of tools improving and users switching used to take years but now takes days or weeks. For example, Hermes has also overtaken Open Core in speed.
These results highlight that staying ahead requires watching benchmarks and user feedback. The teams that catch these shifts early get a head start on better results before everyone else figures it out.
06Tips, Limits, and Mistakes to Avoid
The biggest mistake people make is treating every tool like it's permanent. They get locked into one agent, workflow, or platform, and when something new comes out, they either ignore it (because switching is painful) or jump ship immediately and rebuild everything. Both are bad ideas.
Instead, build your workflow in layers. This way, you can switch individual tools when better ones appear without breaking your entire system. Also, be careful about early benchmarks—they're promising, but not final.
Don't drop everything for Hermes today, but watch it closely.
Another limit is that tools like anti-gravity CLI have token limits, so you can test them for free but with constraints. Finally, remember that infrastructure changes, like Google's improvements, will end up inside products you already use, so stay aware of updates.
07Who Should Use Hermes AI Agent and Why It Matters
Hermes is for business owners building with AI in 2026. If you're using AI agents for client outreach, lead generation, or automation, speed directly affects your costs and results. Faster agents mean more tasks completed on the same budget.
This matters because the AI industry moves fast. Tools like Codex were big just last month, but now Hermes is beating them. The cycle of building, improving, and switching now takes days, not years.
By paying attention to tools like Hermes, you can stay ahead of the curve.
Julian also mentions that if you want coaching on evaluating new AI agents, you can join the AI Profit Boardroom, which offers weekly coaching calls, templates, and a community of 3,100 members. But even without that, the key takeaway is to stay flexible and keep learning.
08Related AI News: Anti-Gravity CLI and AI Satellites
Besides Hermes, the episode covers Google DeepMind's anti-gravity CLI, a free command-line tool for terminal-based coding agents. It lets AI agents run more complex coding tasks directly from a terminal, replacing Gemini CLI. It's free to test, with token limits, and could end up inside products like Gemini and Google Workspace.
Also, Elon Musk announced he will present detailed AI satellite designs in a few weeks. This project, TerraFab, involves Tesla, SpaceX, and xAI, aiming for large-scale chip production in orbit. Space-based data centers could use constant solar power and vacuum cooling, potentially lowering AI compute costs dramatically.
This could make every API cheaper and push back limits on what's viable.
These stories show the rhythm of the AI industry: new tools emerge, improve, and sometimes replace old ones. Staying informed helps you make better decisions for your business.
