Law, E-Commerce, AI. Enjoying life, the universe and everything. #Bitcoin
✝️
Joined January 2009
- Tweets9.8K
- Following587
- Followers674
- Likes115K
Damn there’s OmarchyTV it’s addictive
This is awesome!
Great day to run Omarchy on iPad Pro 9.7 awesome to get it back to life!
Thanks to the work by @henrikklee repo in comments.
Executed by my Hermes agent @NousResearch powered by @Zai_org basically for free
1
4
203
Great day to run Omarchy on iPad Pro 9.7 awesome to get it back to life!
Thanks to the work by @henrikklee repo in comments.
Executed by my Hermes agent @NousResearch powered by @Zai_org basically for free
The problem with AI is more and more that its understanding is deep enough to make it very hard to find its mistakes.
While the few left are still incredibly bad.
These then echo backwards and make them do seemingly rational big decisions terribly wrong, with great reasoning.
It gets the idea and implementation right and even the 2. order effects 80% right and the missing 20% are incredibly understandable so terribly wrong it can be shocking
37
They paid 10-15$ for an algorithm to have her klick and buy that glass.
They have very little understanding how that algo did that, it’s a black box:
You pay, she comes out in perfect buying condition.
14
Someone figured out how to make her feel superior by buying a custom glass.
That’s all it takes.
You brute might use one glass for everything and that works.
But there’s a never ending stream of Chinese businessmen trying to sell glassware that come up with new custom solutions by the day.
She, sophisticated, knows about these, becuase people pay to advertise them to her.
When you tell a model to rename a file:
Claude will rename the file, and all else that needs renaming now, because you and Claude are renaming a file.
When GLM renames a file, it executes a rename operation, because the user expects the file to be renamed.
Big difference
Claude is so great at this because anthropic does focus on the agency of the model, the very thing @DavidSacks criticizes them for.
It’s a questionable path but these harnesses assume and maybe require the models to have an identity and agency to work best.
33
Claude is so great at this because anthropic does focus on the agency of the model, the very thing @DavidSacks criticizes them for.
It’s a questionable path but these harnesses assume and maybe require the models to have an identity and agency to work best.
Introducing Hermes Index: an average of benchmarks measuring model performance in Hermes Agent.
Our goal with Hermes Index is to better inform users choosing between models and to show labs how they perform in Hermes.
portal.nousresearch.com/benc…
Awesome project!
Hold a button. Ask Hermes. Hear the answer.
Hermes Gadget is an open SDK for a small device that talks to your own @NousResearch Hermes.
You speak. Hermes listens, thinks, and answers out loud. The screen shows what it's doing. If it needs your permission, it asks you right there.
No hardware required to try it. The desktop simulator runs the same code as the ESP32 firmware.
Your Hermes, in your hand.
github.com/Adolanium/hermes-…
1
44
The future of consumer robots is pocket-sized, phone-priced, kid-overpowerable, and it cleans your apartment anyway because it has infinite time.
Tesla Minimus.
42
It’s a rare moment where opensource is lapping closed in capabilities and running a simple custom stack gets you ahead.
I’ve had my permanent sessions with separated context (bots) talk to each other in telegram groups all year.
Another way to think about this tweet is that a lot of us have already had this for ourselves. You got it when someone else decided you should
1
70