Alex Finn's walkthrough of the latest Hermes Agent update (June 2026) covers eight major features including native iMessage, automatic background sub-agents, Unreal Engine 5.8 MCP, a redesigned desktop app, visual profile builder, Skills Hub, smarter self-improvement, and enhanced Telegram formattin...
Telegram formatting — Immediate quality improvement, no config needed
Background agents — Pipeline parallelization, test auto-spawn threshold
Self-improvement — Trust Hermes to patch its own skills
Profile builder — Multi-agent expansion path
New rich formatting support: tables, inline code, nested lists, bold/italic, smoother streaming. Finn demonstrates a beautifully formatted stock portfolio table with columns automatically rendered in Telegram.
Steve's take: No configuration needed. Every vault deploy notification, brief summary, cron output, and status update will display more cleanly in your treadmill-to-Telegram workflow immediately. Test by asking Gambit to generate a table in Telegram right now.
No more manual flag to enable. Hermes now auto-spawns sub-agents for complex tasks — you keep chatting while they work. New sub-agent tree shows real-time tool calls across all agents (Finn: 5 agents, 28 tool calls).
Steve's take: Vault pipeline steps (synthesis, HTML generation, manifest update, deploy) could run concurrently while you continue chatting. Test with a multi-step task like "Research X, write brief.md, build HTML, tell me when done."
Hermes now aggressively creates, patches, and updates its own skills during use. Finn demonstrates an Unreal MCP skill that went from struggling to "excellent" through iterative auto-patching after each interaction.
Steve's take: youtube-to-brief, brief-to-slides, and deploy-site could catch and fix their own issues — reducing manual skill_manage patches. Try: ask Gambit to "review and patch the youtube-to-brief skill." Review changes before the next vault deploy.
Browsable library of hundreds of pre-built skills with built-in security scanning. Available via hermes dashboard → Skills → Browse Hub.
Steve's take: Use Finn's method: find an interesting skill in the hub, inspect it, then ask Gambit to "build your own version of this skill customized for my Provenance Labs pipeline." This gives you curated capabilities without blindly downloading external code.
Visual wizard replaces manual config file editing. Choose name, model, skills through an onboarding flow. Each profile is a fully independent Hermes agent with its own behavior and capabilities.
Steve's take: Spin up a "Researcher" profile on DeepSeek V4 Flash for content discovery and brief drafting. Keep the default "Gambit" profile for deploy operations. Each gets its own skills, model, and memory — no config files to touch.
Steve's take: Low urgency. Telegram is your stable primary interface. Worth knowing about for future platform shifts but not actionable today.
Unreal Engine 5.8 now supports MCP, allowing Hermes to build 3D games, environments, and interactive experiences natively — for free.
Steve's take: Zero relevance to research brief and vault deployment workflow. Noted for awareness only. If you ever need a 3D game, you'll know Hermes can do it.
Check version & test Telegram formatting. Run hermes --version. Ask for a formatted table in Telegram.
Experiment with background agents. Give Hermes a multi-step pipeline task and observe auto-spawning.
Trigger self-improvement. Ask Gambit to review and patch the youtube-to-brief skill.
Browse Skills Hub. Open hermes dashboard → Skills → Browse Hub for research-related skills.
Consider a Researcher profile. Use the profile builder for a secondary content-discovery agent.
Tier guide: Tier 1 = Primary institutional source · Tier 2 = Expert/educational channel · Tier 3 = News/opinion commentary
Provenance: Sources: YouTube video content. Transcripts auto-extracted via Obsidian Web Clipper. Verify key technical claims against official Hermes Agent documentation at hermes-agent.nousresearch.com/docs. Personalized analysis by Gambit (Hermes Agent) for Steve's Telegram-native Provenance Labs workflow.