Researcher & SWE works like machine & optimizer stay curious, do art & engineering
San Francisco, CA
Joined May 2026
- Tweets1.3K
- Following95
- Followers2.7K
- Likes2.3K
Pinned Tweet
Meet Oh-My-Hermes
I burned 200B tokens tuning my Hermes Agent setup, and now you can get all of it with a one-line install.
It comes with 150+ skills, a long-term memory system that lasts twice as long, graph engineering, and memory-management skills. It also includes subagent routing across model chains, plus optimizations like parallel tool calling and evals.
You don't need to do much, though. Just install it and give it a task. Everything is optimized to route automatically. Hermes Agent is excellent, and I love this technology.
I put a lot into making it really good, then open-sourced it so you can enjoy it right away. Meet oh-my-hermes.
GitHub: github.com/rlaope/oh-my-herm…
#hermes
HOPE. retweeted
MEET DEEPSEEK ECO !!! ITS CRAZY @@@@
Same DeepSeek V4.1 Flash. 1/10 the price. !!!!!
👉 opengateway.ai
We're testing how cheap we can serve it before it gets worse. It hasn't yet.
That's a real run in DeepSeek Harness. The answer streams at real speed.
At peak hours it might make you wait a beat. Whales come up for air too. 🐋
Point your agents at it and let them run all night.
#dsh #deepseek #eco #ultrafast
MEET DEEPSEEK ECO !!! ITS CRAZY @@@@
Same DeepSeek V4.1 Flash. 1/10 the price. !!!!!
👉 opengateway.ai
We're testing how cheap we can serve it before it gets worse. It hasn't yet.
That's a real run in DeepSeek Harness. The answer streams at real speed.
At peak hours it might make you wait a beat. Whales come up for air too. 🐋
Point your agents at it and let them run all night.
#dsh #deepseek #eco #ultrafast
HOPE. retweeted
Burning through your Grok Bot usage on one workflow?
X is flooded with Grok Bot usage screenshots. 👀
Plan just reset, one workflow, and a quarter of the max plan is already gone.
You don't have to ride that meter. Take the whole bot to Hermes:
hermes migrate grok-bot
▸ Routines become Hermes cron jobs, paused until you're ready
▸ Skills, memories and your bot's rules come with you
▸ Apps with an MCP URL get wired in as MCP servers
Hermes writes the export prompt, Grok Bot writes its own export file, one command finishes the move. Fresh installs offer it during setup.
Same workflow. Any model you want. Your own machine.
ALT Grok Bot rug pull? Move in. Grok Bot's colorful bots under a weekly usage bar at 100%, an arrow pointing to the Hermes Agent desktop app, and the command hermes migrate grok-bot.
i built deltarune-maker #deltarune
An unofficial DELTARUNE fan studio for game experiments and short films.
github.com/rlaope/deltarune-…
Oh My Hermes has a skill for budget vs actual reviews.
Give it the GL export, the budget and the period, and it writes the variance and cash brief: which lines moved, why they moved, and what that does to runway.
Totals have to tie back to the source, and the math is recalculated with a tool. An assumption you haven't approved, like the burn rate, gets flagged as unapproved instead of passing as fact.
If something doesn't tie out, the brief goes on HOLD with an owner.
The picture is a mockup with a sample company.
It won't post journal entries, approve payments or file anything.
Try it with /omh-finance-analysis
github.com/rlaope/oh-my-herm…
Oh My Hermes has a skill for game sprites.
Image models draw each pose at a slightly different size and height, so frames you cut by hand jitter in play. This skill takes a grid sheet on magenta, aligns every frame to one baseline, and writes a QC file. If the scale or the feet drift past the limit, it tells you to regenerate the sheet instead of nudging frames.
The QC numbers in the picture are from a real run. Two sheets of our mascot Kami failed, and the third one passed.
Try it with /omh-sprite-forge
github.com/rlaope/oh-my-herm…
Meet DeepSeek Eco (DEEPSEEK ULTRA)
Same DeepSeek V4.1 Flash. Same brains. A tenth of the usual price.
We're trying out serving it this way, and honestly? It's working.
The catch: when everyone shows up at once, you might wait a sec.
Your overnight agents won't notice. Your bill will.
Give it a spin and tell us how it feels.
opengateway.ai/models
You can grab this entire skill pack for your Hermes Agent completely free.
Get it running with a single install around 150 skills ready to use with zero overhead.
oh-my-hermes is an all-in-one repository integrating various plugins and skills built specifically for the Hermes Agent.
github.com/rlaope/oh-my-herm…
burned nearly 1t tokens fine-tuning hermes agent
coding intelligence long-term memory and 150+ skills no annoying setups or config hell needed
because i already did all the heavy lifting just one install and you're good to go
github.com/rlaope/oh-my-herm…
Hey everyone, I found a way to run DeepSeek at 1/10th of the price.
No expiration date, and we also have an ultrafast model pushing 1,000 TPS.
The model is called deepseek v4.1 flash eco. While it's not quite as fast as the ultrafast version, it’s 90% cheaper and definitely not slow by any means.
We’re intentionally taking a loss on serving this because we just want you all to enjoy using it.
It's available on OpenGateway
go check it out here: opengateway.ai/models
#deepseek
HOPE. retweeted
90% off DeepSeek. No end date ??!!!!!
Stop paying full price for DeepSeek.
DeepSeek V4.1 Flash Eco is now on OpenGateway. Same model, 90% off. 🐋
That's 10x more tokens for the same budget:
→ let your coding agent run all day without watching the bill
→ ship the side project you shelved because the API cost didn't add up
→ give every user in your app the full model, not a cut-down one
→ stop trimming prompts to save money
Nothing important left behind:
• Tool calling and thinking work just like the standard model
• Prompt caching: resend a long context and it comes back fast for almost nothing
• OpenAI-compatible API. No subscription, pay per token
Switching takes one line:
model="sionic-ai/deepseek-v4.1-flash-eco"
check it out now: opengateway.ai
90% off DeepSeek. No end date ??!!!!!
Stop paying full price for DeepSeek.
DeepSeek V4.1 Flash Eco is now on OpenGateway. Same model, 90% off. 🐋
That's 10x more tokens for the same budget:
→ let your coding agent run all day without watching the bill
→ ship the side project you shelved because the API cost didn't add up
→ give every user in your app the full model, not a cut-down one
→ stop trimming prompts to save money
Nothing important left behind:
• Tool calling and thinking work just like the standard model
• Prompt caching: resend a long context and it comes back fast for almost nothing
• OpenAI-compatible API. No subscription, pay per token
Switching takes one line:
model="sionic-ai/deepseek-v4.1-flash-eco"
check it out now: opengateway.ai
oh-my-hermes bots desktop
it's very insane. release soon.
GitHub: github.com/rlaope/oh-my-herm…
i've burned through hundreds of billions of tokens with Hermes Agent, and every friction point I hit along the way became a workflow.
oh-my-hermes packages those workflows into skills and systems you install once.
It never patches or replaces Hermes. You keep talking to Hermes as usual, and OMH picks the right skill and model for each lane and runs the work in parallel
Your goal becomes a graph of small nodes.
The ready ones run in parallel as Hermes subagents, each on the model its category calls for. A separate node verifies the whole goal, and failed nodes get repaired before you see the report.
The OMH pane shows the graph live as each node goes from running to accepted.
What else is in the box:
- 100+ expert skills (frontend, backend, security review, verification gates) that the router loads for you
- A model chain per work category, which moves to the next model instead of silently downgrading
- Long-term memory that keeps, refuses or defers each candidate with a written reason
- Evidence before "done": nothing is reported verified until a test, review or CI gate actually passed
GitHub: github.com/rlaope/oh-my-herm…
Web: omher.app
#hermes #skill #packages #omh
Almost Free DeepSeek V4.1 Flash + GLM 5.3 Flash.
@OpenGatewayai
DeepSeek V4.1 Flash is 70~90% off and running at ~1,000 tokens/s.
web: opengateway.ai
model: opengateway.ai/models
We built Oh My Hermes for the jobs that are too big for one agent turn: a couple of features to build, docs to write, and someone has to check all of it before it gets called done.
Oh My Hermes is an operating layer for Hermes Agent. You keep talking to Hermes the way you already do, and OMH picks the skill and the model for each piece of work. It never patches or replaces Hermes itself.
For a bigger job, start your message with /dag-ulw. The goal becomes a graph of small nodes, and the ready ones run in parallel, each on its own model and reasoning effort. Two nodes running at the same time never write to the same files.
A separate node checks the goal at the end. Failed nodes are repaired before you get the report, and when a provider rejects a model, that lane moves to the next model in its chain instead of quietly downgrading.
Every status says how far the work actually got: Plan · not run, Code · reported done, Test · verified. A run only counts as verified when a test, review or CI gate passed.
It comes with 150+ specialist skills for frontend, backend, Rust, security review and more. You don't call them by name; the router loads the ones the task needs. Long-term memory works the same careful way: nothing is saved silently, and every candidate is kept, refused or deferred with a written reason.
Install once:
github.com/rlaope/oh-my-herm…
