AI Labs Will Trap You or Replace You - Dillon Rolnick, Nous Research
Key Takeaway
As AI models commoditize and labs attempt to capture trillion-dollar infrastructure investments through proprietary lock-in or direct vertical replacement, organizations must preserve their autonomy by building on open agent foundations like Hermes Agent.
Highlights
AI labs have spent hundreds of billions of dollars building intelligence systems, but models are rapidly commoditizing as baseline intelligence becomes widely available and average token costs decline.
Businesses will organize 70% or more of their workforce output through agentic systems within two to three years, making the agent layer the primary battleground for AI margin capture.
Open source developers reproduced XAI's GrokBots user interface and released it as a native feature within Hermes Agent in a single weekend.
Proprietary software moats are disappearing because the marginal cost to reproduce software has plummeted, shifting future moats to network effects, reliability, and reputation.
Hermes Agent is currently the most downloaded open source agent in the world, with download numbers an order of magnitude higher than its closed competitors.
Timeline
Model Commoditization and the Shift to the Agent Layer
Practitioners readily switch between different frontier models without brand loyalty. Because model margins are shrinking, competition shifts toward the agent layer where customer stickiness secures profit margins. Enterprises represent the most lucrative customer segment for token consumption and economic productivity. Consequently, labs eye high-value industries like financial services as tractable problems to solve directly rather than selling low-margin tokens to external firms.
- AI labs and the broader industry have spent hundreds of billions to trillions of dollars building advanced intelligence infrastructure.
- Average token costs are declining while total token spending surges, compressing margins on performant models due to rising competition and commoditization.
- AI labs pursue two primary strategies to recoup infrastructure costs: competing at the infrastructure or user layer, or building advanced artificial general intelligence to capture lucrative industries.
The Risks of Closed Foundations and Proprietary Lock-In
Decision-makers across executive and engineering roles must choose where to anchor their technological roots. Selecting proprietary systems integrates organizational intelligence into closed ecosystems that eventually trap users and raise prices. Because agentic workflows will soon orchestrate the majority of enterprise output, depending on a centralized third-party platform introduces extreme strategic risk.
- Closed AI labs target meaningful double-digit percentages of the global economy through deeply embedded enterprise tools.
- Organizations that adopt closed tools will harden their processes around them, absorbing proprietary dependencies and losing long-term operational autonomy.
- Foundational reliance on closed providers creates severe vulnerability to future price hikes and adverse incentives.
The Obsolescence of Software Moats in Open Source Development
Historical moats built on proprietary intellectual property no longer protect software businesses from nimble open source ecosystems. When XAI released GrokBots, open source developers deployed Hermes Agent to reproduce the feature within a single weekend. AI model creators frequently underestimate user ingenuity and the speed at which open source alternatives match proprietary software advancements.
- Nous Research was established for moral and philosophical reasons to combat centralized control over foundational technology.
- Proprietary software moats are disappearing as the marginal cost of reproducing software plummets toward zero.
- Open source builders can reverse-engineer and replicate proprietary features within days using autonomous agent workflows.
Agents as Operating Systems and the Open Source Standard
Modern enterprise applications integrate routine code with intelligent reasoning engines. Organizations should build these systems on open foundations rather than transforming into dedicated intelligence providers. Hermes Agent offers the scale and flexibility required to adopt state-of-the-art capabilities without sacrificing independence. Building on open infrastructure ensures that organizations retain control over their destiny as agentic automation reshapes the global economy.
- Agents function more like operating systems and application frameworks than simple conversational chatbots.
- Agentic applications combine deterministic code, non-deterministic data sources, human judgment, and reasoning primitives.
- Hermes Agent provides an open, customizable standard that lets organizations maintain control over their infrastructure and swap underlying models freely.
Description
Dillon Rolnick, CEO of Nous Research, argues that as AI models commoditize, the labs that spent trillions building them have two ways to earn it back: lock businesses into their platforms and raise prices, or point their own systems at lucrative industries like finance and compete directly. He makes the case that the moats of proprietary software are gone, that agents are closer to operating systems than chatbots, and that companies should build on open foundations they own and control, like Hermes Agent, so they can swap models and inherit new capabilities instead of rebuilding them. Chapters: 0:00 Intro 0:32 The models are commoditizing 1:52 How labs plan to earn back trillions 2:48 Why businesses are the real target 4:28 The labs are coming for finance 5:44 Don't build on quicksand 6:48 Why Nous Research exists 7:56 Software no longer has a moat 9:25 Distillation and the fight over open models 11:06 Agents are operating systems 12:17 Build on foundations you own 13:49 Hermes Agent and the new moats Dillon Rolnick https://nousresearch.com Join us in SF for the second annual AIE CODE summit: https://aie.codes our last flagship summit of 2026. Applications live now!