Nous Research launches Hermes agent and hackathon, positioning open-source as cheaper alternative to Claude
Mar 4, 2026 with Dillon Rolnick
Key Points
- Nous Research launches Hermes agent, an open-source AI tool priced to undercut Claude by targeting users balking at costs above a few dollars per day.
- The company is using a hackathon as a community discovery mechanism to surface real use cases and feed learnings back into open-source model improvements.
- Nous frames Hermes as a feedback loop: deploy the agent to real users, gather what actually matters, then fine-tune open-source models to match those demands.
Summary
Nous Research launched Hermes agent, an open-source AI agent positioned as a cheaper alternative to Claude, and announced an online hackathon. The company's pitch centers on cost and accessibility rather than raw capability.
Dillon Rolnick, Nous Research's COO, describes Hermes agent as a hybrid combining coding agents and orchestration tools, built on fine-tuned open-source base models. The core positioning is price. While Claude is "extraordinarily capable," users often balk at costs. "I'm willing to spend a couple bucks but maybe not like a $100 a day," Rolnick says. Hermes agent targets that gap with agentic capability at a lower price point.
Nous has traditionally built its brand through post-training and fine-tuning of open-source models like Llama. The agent launch creates a feedback loop. By releasing an agent that touches real users and real work, the company gathers signals on what post-training improvements actually matter. Rolnick describes the strategy as putting the agent into the field, learning what people actually need it to do, then shaping open-source models to fit that use better.
Rolnick faces a go-to-market challenge. The gap between what users think AI can do and what it actually can do is wide. Even savvy non-technical users are surprised by Hermes agent's capabilities. But specificity matters. When pressed on what users should try first, Rolnick defaults to broad advice: identify the task you spend the most time on that you don't want to do, and ask Hermes to solve it. That is a direction, not a killer use case.
Rolnick positions the hackathon as a community discovery mechanism. Nous released Hermes as "a gift to the community" and is using the hackathon to surface what actually resonates. The company appears to be in play mode, testing what sticks rather than following a hardened product roadmap.
The bet is that open-source agents can undercut proprietary alternatives on cost while building a tighter feedback loop between deployed tools and model improvement. Whether that is enough to compete with Claude's capabilities, or whether cheaper is a sustainable wedge, remains unproven.