rip

I've been using Hermes as my personal agent for about a week now. Some observations: 1- My setup is a little unusual. Hermes runs on @Railway and connects to all of my tools through my own MCP gateway (email, calendar, Slack, X, LinkedIn, GitHub, Obsidian, DNS, etc). It runs code in a separate @vercel sandbox. 2- I started out communicating w/ my Hermes agent on Telegram but moved to @discord after just a few days... Telegram doesn't really do threads or channels, and honestly the mobile app was so buggy that it was hard to use. In Discord I have a channel for each area (email, X, GitHub, calendar, reminders...) and Hermes posts updates there throughout the day. 3- I also added Hermes to the @sazabi Slack workspace so that I can communicate w/ it in full-view of my team. The catch: it ONLY responds to me. It ignores DMs from everyone else and only replies in channels when I tag it. I didn't want my teammates (or customers in Slack Connect channels) to be able to get at my inbox through it. 😅 4- I gave Hermes access to my @obsdmd vault; it uses Obsidian like a notebook. My vault is synced to the cloud and I run a small MCP server (vault-cortex) that gives Hermes read/write access. Hermes keeps a note on every person I interact with (built from my Slack, email, meeting notes and WhatsApp), writes a recap of my day every night, and saves things like PRDs. 5- I also built my own memory system, which is also hosted on Railway + exposed to all of my agents over MCP. Claude Code, Codex, Cursor and Hermes all read from and write to this one memory store. The first time I started Hermes, it onboarded itself by reading all of these memories, which was pretty sick. 6- Things Hermes has been most useful for: - inbox triage (it drafts replies and I approve them one by one; I am not just triaging my email inbox, but also Slack, LinkedIn, X, iMessage, etc.) - morning brief + evening recap - prospecting/sourcing - working on side projects - tracking and landing PRs 7- Because Hermes triages + summarizes so many different inboxes for me, I find that more and more of my "work" happens in Discord; I'm just talking back and forth w/ Hermes all day. 8- I mostly interact w/ Hermes using voice notes or @WisprFlow , but it also has the ability to "talk" w/ me over a regular phone call w/ @Vapi_AI . It supports both inbound + outbound calls. The first few calls were kinda rough so I configured Hermes to reviews every call afterwards, reflect on what went wrong, and implement fixes. It's working way better now. 9- One obvious but huge part of Hermes that gives it the edge over local coding agents: it keeps getting better. Every time I correct it or give it some info about how to perform a task, it saves this to memory and references it in the future. Other coding agents do this too, but Hermes seems to do a better job of "learning" and getting smarter over time. Also, because there are no "sessions" , Hermes gives you the feeling of infinite context, which is really nice. This post got way longer than expected, but there ya go. My impressions from one week w/ Hermes. TLDR - I'm a fan. I am tempted to spend more time w/ other personal assistants like @Muse and @instinct_xyz but , to be honest, after seeing a post from the @NousResearch team about how "Hermes will always be for power users", I am doubtful that I will prefer either of those...

EU sorry. We’re making it so that (a) if you had created a dot elsewhere and find yourself in the EU it will keep working (b) dots are coming to the EU in the future

So this is cool. If you ever need an ephemeral dev environment and need to scale up resources try out running k3s in our @CoreWeave sandbox platform. Using private networking via @Tailscale we can spin up a k3s cluster with many worker nodes. Perfect for local e2e testing or dev environments.

I'm really trying to stick with Android, but the Slack app sucks soo bad. Android was my daily driver for a decade, only been Apple for a few years. Funny how difficult this has been.