Guest
Karan Malhotra
Co-founder, Nous Research
Real Alpha has cut 8 moments of Karan Malhotra (Co-founder, Nous Research) talking, from an interview on Behind the Craft published Aug 2, 2026.
What Karan Malhotra said
- 26:55
Hermes can turn a childhood game idea into a top-tier mod
“it was able to get into this niche and perform at the top 1% of modders.”
at 33:19 - 11:11
Adversarial context can reduce an AI’s sycophancy over time
“The most powerful thing for a model is in context learning. ICL is more powerful than everything else, fine tuning, whatever.”
at 12:51 - 23:30
One Kanban board can coordinate people and AI agents
“So the fact that I can have a human project manager on the Kanban board that's manually using it while Hermes agents are filling up the pieces”
at 24:06 - 0:56
Hermes uses context to align models with users
“By putting all this context together, we've taken Claude's main allegiance away from Anthropic to you.”
at 7:24 - 42:35
Hermes Agent began as an open harness for reinforcement learning
“And we built Hermes Agent initially to let anybody do RL inside of a harness.”
at 43:53 - 16:30
Open systems give good actors a fairer path, Malhotra argues
“An important note in the argument for open source is that an open system is much fairer to a good actor than a closed system with models”
at 18:01 - 36:03
Nous grew from two Discord friends into an open-source lab
“hundreds of thousands of downloads in just a few days.”
at 37:52 - 20:00
Hermes sells convenience and custom support around its open agent
“If you want to do image generation or audio or VPS spin up or web search, we have all of those subscriptions included inside of ours.”
at 21:02
- 26:55