Ryan Prasad

@dansemperepico Did not enjoy banging on Claude Code and trying to make it into a local claw. Happy with Hermes + Honcho + OpenAI sub.

Great news. I've officially moved the first of my 5 OpenClaw agents to Claude Code. It has the same functionality as my Claw and all of my Claw's memory, context, knowledge base, and personality, but writes to my second brain MD files instead of to its own separate files like a Claw does. This means that I can port the second brain into any agentic harness I want (Gemini, Codex etc) and have access to all these models without high switching costs. When I'm sitting at my computer, I use Claude Code via CLI or the Claude Desktop app, and when I'm away from my computer, I can continue working with my agents via Telegram. Best thing about this is that I get to use my Claude Max subscription with the latest and greatest Opus 4.7 and I get Anthropic's enterprise grade security.

@sudoingX Would love to try this, but without the Max plan, this would be a short-lived experiment as a Claude Code believer. ChatGPT 5.4 works perfectly well for my use cases.

people ask me this all the time so let me say it out loud. the best model to pair with hermes agent is claude opus 4.7, by far, in order of magnitude. the rest is not even close. pair opus 4.7 with hermes and you feel the unlock immediately. the self improving loop, the per model parsers, the tool call handling, every part of hermes agent amplifies when the model on the other side is that capable. hermes agent is one of the most well designed and rapidly contributed open source projects in the space right now. the combination with opus is the current ceiling for agentic work.

@rivestack Thank you for the pointers!

@ryanprasad_ai for the slack channels piece, one embeddings table with a channel_id column gets you per-channel isolation and cross-channel search from a single index. also worth adding wal archiving to s3 on top of your nightly backup for point in time recovery

Have Hermes installing Postgres + PGVector + Honcho locally and doing nightly backups to S3. Reminds me a bit of Claude Code. Need to see how others have this setup for multiple Slack channels with persistent memory. Also we have a SOUL.md file now...

@NousResearch @Saboo_Shubham_ Indeed, this is how I am using it.

@Saboo_Shubham_ Alternatively: Hermes writes the code. Hermes reviews it. Hermes runs the loop. Hermes turns mistakes into strategy.

This is WHAT running AI Agents look like in 2026. Claude Code writes the code. Codex reviews it. OpenClaw runs the loop. Hermes turns mistakes into strategy.

@Paul__light @ClaudeDevs @scsp_ai expo.scsp.ai/hackathon/ Not there to win. Don't care about the money. Thanks though!

@ryanprasad_ai @ClaudeDevs @scsp_ai Success in you hackathon bro. I do hope you win What’s your hackathon about?? Is it late to apply

Our virtual hackathon is back! Join us for a week of building with Opus 4.7 alongside developers from around the world. The Claude Code team will be in the room all week, with a prize pool of $100K in API credits.

Our deterministic script seems to be working as intended. ~$30 to get from the 495/95 Junction to Quantico right now. We will have some solid data on toll rates over time. Should make for easier budgeting. Hermes automatically writes reusable skills. Very cool.

Delegating to Hermes feels so natural in Slack. I'm opinionated, but I think paying for a seat(s) is worth it - if only to have the 'lets get stuff done' vibe.

Codex proposed schema for the toll pricing data collection script. Will run 6-11 AM and 3 - 7 PM. Just need to vaccum everything up for now. Processing will be done later at analysis. That bit could be tricky, as we will probably need to add trip segments and stuff.

Got Hermes Agent (+ local Honcho brain) hooked into Slack! Eventually we will have this in separate channels with separate contexts. This will eventually be the budget bot that connects to You Need A Budget API. This is what I wanted local Claude Code to do, but whatever.

@dreamsandcode I feel for these folks

The mistake is thinking AI is a fad

@dreamsandcode Nice app!

@ryanprasad_ai I'm curious how it would compare to what I built shelter.money

Okay! I think RyanClaw is a red herring now and is going to cause us to fall behind. Try to not reinvent the wheel and all... Now that the jobs pipeline is toast, let's downgrade Claude Max, upgrade Codex sub, and give Hermes Agent a shot as our AI assistant. Hermes + YNAB

@nicoloboschi Nice work on that! I'll have to give it a test drive.

@ryanprasad_ai mem0 is a reasonable option for simpler agent setups. if you need more retrieval power for RyanClaw, Hindsight is state-of-the-art on LongMemEval and fully open source. github.com/vectorize-io/h…

'Tighten up security for personal cloud infra' is turning into 'tear it all down, it's all old anyway.' Instead of having Postgres as the brain for RyanClaw, I am looking at a more specialized solution for AI agents like mem0. The jobs pipeline is a bust. No need for Postgres.

VDOT has an API that provides real time toll pricing (based on demand) as CSV. Gonna start collecting rush hour pricing data now so I can make better estimates when decision time comes around. Should be easy for Hermes. Poll every 10 min, dump in PG. smarterroads.vdot.virginia.gov

Pleased with Opus 4.7 in Claude Code. Opus 4.6 was fine though. Dunno what the devs on here were bitchin about. Great for troubleshooting over long tasks. Would prefer to use this in Hermes Agent instead of beating on Claude Code to function like OpenClaw.

@MatthewBerman Oh no! You put so much work into that elaborate setup.

I have to say something...please don't be mad. OpenClaw has been nearly unusable for the past week. Something changed and now everything is broken. Looking forward to testing Personal Computer.

And here is proof of memory recall. Not the fastest, but me likey. Time to get Slack set up.

So far very pleased with Hermes Agent + ChatGPT 5.4. It 1-shot the correct local Honcho installation locally after we realized last night's attempt was wrong. Forget my earlier whining. It's good. Honcho is the brain for Hermes so that we have persistent chat sessions.

@AmrTawfik160 I am revising my initial assessment. It's slow, but it's 1-shotting complex tasks. I'm impressed.

@ryanprasad_ai sounds like quite the hassle, especially when time's at stake. ever think about switching back until it gets smoother?

Not sure I like ChatGPT 5.4 for human on the loop work in Hermes Agent. Very sluggish compared to Claude - burning a lot of my time with reasoning. Honcho installation is a bit more complicated than I had originally anticipated. Hopefully we can use my Codex subscription.