Apr 20, 2026 · 37m · startup-ideas
Hermes Agent: The New OpenClaw?
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this technical masterclass, Greg Isenberg and guest Imran provide a comprehensive guide to Hermes Agent, detailing its persistent memory, step-by-step setup, cost-cutting automation techniques, Android edge deployment, and productivity integrations with Obsidian and Garry Tan's G-Stack.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Greg holds 26.1% of the talking time here. How this is scored →
speaking balance: gold is Greg, purple is the guest (3 minute bins)
When Greg asks if setting up the full agent workflow takes 20 days, Imran pushes back gently to clarify that it's about forming the behavioral habit of defaulting to agent delegation rather than a fixed number of days.
Hardest push from Greg ▶ 3:41 Greg challenges the friction of switching ecosystemsGreg stops the flow to challenge whether Hermes is just another short-lived tool that users will abandon for OpenClaw in a week, demanding proof of long-term utility.
Biggest teaching moment ▶ 10:53 Imran explains deterministic code generation to slash token costsImran teaches Greg that recurring agent tasks should generate deterministic Python/Bash code once rather than repeatedly invoking LLM reasoning loops, dropping expenses by 90%.
Greg holds their own ▶ 35:49 Greg contextualizes agent time savings into executive financial returnGreg builds directly on Imran's venture deal flow metrics by quantifying the financial math of valuing an operator's time at 500 dollars per hour.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Greg as informed peer | Guest teaching | Guest disagreement | Greg pushing back | Why |
|---|---|---|---|---|---|---|
| Episode Roadmap and Core Agent Capabilities | 2 | 6 | 1 | 0 | Greg sets up the premise asking Imran to explain Hermes Agent clearly so viewers can set it up immediately. Imran explains the core technical limitations of OpenClaw (lack of built-in memory, unstable gateway, opaque token spend) and how Hermes solves them with SQLite logging and stability. | |
| Selecting a Long-Term Agent Ecosystem | 3 | 5 | 0 | 1 | Greg expresses the practical user hesitation about committing to an agent ecosystem rather than jumping between tools weekly. Imran validates the concern and demonstrates the pre-packaged 40+ native tools and built-in Mac integrations that make Hermes a permanent choice. | |
| Security Audits and Flexible Deployment Environments | 3 | 6 | 0 | 1 | Greg raises security questions about granting tools broad permissions. Imran provides practical architectures including prompt-based self-audits, Docker containers, and serverless Modal deployments. | |
| Model Selection, Anthropic Support, and OpenRouter | 4 | 6 | 1 | 1 | Greg asks about Anthropic API key compatibility and notes runaway token costs on OpenClaw. Imran details OpenRouter model pricing, cost differentials between models, and the software engineering approach of writing deterministic code once to cut recurring token costs by 90%. | |
| Telegram Integration and Mobile Deployment via Termux | 2 | 7 | 0 | 0 | Imran demonstrates his Telegram integration with multiple Muppet-named agents and walks through deploying Hermes on a Solana Seeker Android phone using Termux and the Termux API to access hardware sensors. | |
| Business Opportunities with Mobile Agent Automation | 4 | 5 | 0 | 0 | Greg prompts Imran to brainstorm entrepreneurial and monetization use cases for mobile agent instances. Imran suggests scalable social media posting directly from hardware Mac addresses to bypass scheduling API rate nerfs. | |
| Agent Memory Querying and Personal Life Auditing | 3 | 6 | 0 | 0 | Greg asks about auditing personal life routines and suggests querying the agent directly. Imran queries Hermes live on his personal habits, revealing detailed context about his daily life and dietary routines. | |
| Software Updates and Remote Access via Tailscale | 4 | 5 | 0 | 1 | Imran reviews maintenance practices such as daily git updates and setting up Tailscale mesh networking. Greg probes whether users should build specialized single agents or distinct personas. | |
| Obsidian Integration and Automated Daily Dashboards | 3 | 6 | 0 | 1 | Imran demonstrates his daily Obsidian dashboard generated automatically via Markdown manipulation. Greg questions whether building such a workflow requires 20 days or can be set up in a week. | |
| Essential Meta-Prompts for Maximizing Productivity | 3 | 5 | 0 | 0 | Greg asks Imran to write down the core meta-prompts that yield the highest productivity returns. Imran outlines prompts for procrastination auditing, automated tool construction, and daily task automation. | |
| Turning Knowledge Bases into Custom Agent Skills | 4 | 5 | 0 | 0 | Imran showcases converting a historic chatbot Wikipedia page into a custom Hermes skill. Greg reinforces that developing the mindset to convert daily resources into skills is the true modern AI capability. | |
| Must-Have Skills and Garry Tan's G-Stack | 3 | 6 | 0 | 0 | Imran recommends vital skills including the Obsidian integration and porting Garry Tan's G-Stack into Hermes. Greg asks for a concise primer on what G-Stack is and how founders can apply it. | |
| Avoiding the Customization Trap and Fund ROI | 4 | 5 | 0 | 0 | Greg and Imran warn against the 'tuner car' trap of endlessly tweaking agent configurations without executing real work. Imran shares how Hermes improved deal flow and founder conversations at his investment fund, which Greg calculates into substantial hourly ROI. |
Statements from this episode (0)
Nothing in this episode matches those filters. clear them