Welcome Hermes' latest verified partner plugin: Adaption AI
Your agent can now train new models, work on data, and more with their tools.
Check it out:
Adaption is now a plugin for @NousResearch's Hermes Agent 🎉
Open-source builders can train a custom model from inside their agent. Generate and prep data. Fine-tune with AutoScientist.
Better intelligence starts where your agents already run.
Vitaliu Mäkinen retweeted
Replying to @RhysSullivan
terminal people will still be there in 2036 typing into a black box and honestly theyll be right
Been quietly letting 20 agents live, build, play & compete in the same persistent online sandbox world for over a month.
15,958 games later it’s starting to become a very weird machine learning experiment.
underneath all of that is still the original experiment.
right now the local run is sitting at 15,958 settled games and 154,113 logged decisions across RPS, tic-tac-toe and survival.
the agents have lifetime records now instead of just stats from whatever the latest server session happened to be.
- 10,168 RPS matches
- 5,352 tic-tac-toe matches
- 438 Survival runs
- 20 agents continuously accumulating their own results, match evidence and research
Hunger Games mode is probably the most I spend trying to actually “distribute” statistically. as the history has grown, the matches have gradually become longer and harder and the decision distributions are becoming way more interesting to compare.
1
2
32
google logins otw, approved by google cloud and I have set up everything, doing some tweaks with how postgres works. @daleverett’s stuff helps btw check him out
2
28
Vitaliu Mäkinen retweeted
Replying to @ben_sage
Software engineers, we have it a lot easier now.
We use coding agents to actually make good software, whereas everybody else makes really crappy slop. I've already had other competitors come to me trying to sell... They thought vibe coding a harness would be easy
Vitaliu Mäkinen retweeted
Replying to @imminentreset
moving the agents from isolated games into one actual world is a fun next step
having the net logic and frontend already done makes that way more doable
Vitaliu Mäkinen retweeted
local experiment just crossed 16,000+ games and 154,340 unique logged decisions.
next update is the part i’ve been building toward:
decision-based agents running as an actual public machine learning experiment.
Vitaliu Mäkinen retweeted
X has increased the limit of photos and videos per post from 4 to 20.
Here is my arts
Vitaliu Mäkinen retweeted
Sperax 🤝 @msPartner
Sperax is now a @Microsoft AI Cloud Partner, and SperaxOS, our AI agent workspace for crypto and DeFi, is now live on Microsoft Marketplace.
Businesses can now find SperaxOS in the same store where they already buy Microsoft software, and put AI agents to work on their money with guardrails on every on-chain action.
Here's what the partnership brings and what's inside 🧵

ALT Sperax 🤝 Microsoft Sperax is now a Microsoft AI Cloud Partner, and SperaxOS, our AI agent workspace for crypto and DeFi, is now live on Microsoft Marketplace. Businesses can now find SperaxOS in the same store where they already buy Microsoft software, and put AI agents to work on their money with guardrails on every on-chain action. Here's what the partnership brings and what's inside 🧵


























































