Building weird systems / Breaking standard AI workflows / Assimilate or be consumed / Accelerate / Follow if you're building & learning
Joined May 2008
- Tweets 5,545
- Following 345
- Followers 922
- Likes 2,648
847 Photos and videos
How do you make AI safer?
By accelerating development, making it open source / weights, making it accessible, and moving fast.
Slowing it down only delays finding and solving problems.
May 9
Normalize calling AI regulation what it actually is: anti-AI regulation. Because most of it isn’t about making AI safer. It’s about slowing it down, raising barriers to entry, and protecting incumbents from competition.
1
2
114
Vectro Chat is going to have a feature that creates a character to message with.
It will avoid the problems Character AI has had for a few years.
It will be marketed as a fiction generator and creative writing assistant, which the user will agree to in the TOS.
1
154
Need inexpensive API?
I'm already a user of the Nous portal, but if you're not, you might want to take advantage of this deal quickly now before the slots fill up.
May 9
Ready to finally make the switch?
Here is a $20 off Nous Portal code for the first 200 new users, to celebrate our new #1 spot.
portal.nousresearch.com
2
10
11,846
Coding LLMs use the word "fixed" when code is updated
Even if there was no testing or proof, it seems to mistake changes for fixes.
To prevent it, I've added this to SOUL md:
### Verified Means Verified
Never say something is fixed just because code was changed.
Fixed means ALL of these are true:
1. The code was changed
2. The change was tested
3. Error capture confirmed the problem is gone
4. Results were verified before telling the user
1
3
53
Being able to interrupt Hermes Agent by messaging it is huge.
It will stop the agent loop like hitting a pause button.
Then you can steer it from there.
Also check out related features:
Hermes Agent tip of the day:
There are 4 ways to deal with the model while its running,
- Message it, by default, it will interrupt the agent loop, stopping it and making it respond to your new message
- /queue will queue up a message that will fire after the agent loop completes
- /bg or /btw will run a parallel prompt that is async
And
- /steer will inject a guidance message into the next tool calls result sent to the model during an agent loop, to try to guide the rest of its trajectory
1
2
161
DeepSeek V4 on @OpenRouter 429 rate limit errors prevented what could have been a really cool experiment.
I hit the API with requests all day and it's usually not a problem.
Im Hermes, I told it build any app it wants that helps humanity, but is something people missed.
I just gave a YOLO prompt to Hermes Agent with DeepSeek V4
I told it to build anything it wants as long as it helps humanity but is something that people missed.
I told it to search around the web a bit to make sure it's not something that already exists.
1
161
Hermes agent automatically created a skill to use headless apps on Ubuntu
Hermes Agent controlled a desktop app on Ubuntu
Asked it to do this and it immediately installed:
xvfb xdotool scrot imagemagick
Told it to open calculator and do 1 1
It tried headed, failed, switched to headless...
Then automatically made its own GUI automation skill!
2
95