Huge congratulations to @NousResearch and team @dillonrolnick @Teknium @karan4d @theemozilla @gradyacc and many others on closing their $90M Series B at a $1.5B valuation. @Delphi_Ventures co-led their seed round, doubled down in their Series A, and participated again here. This marks us tripling down on Nous. Following Nous' journey over the past few years has recalibrated my view of top tier founders The team has always been months to years ahead at every juncture. They did synthetic model evaluations years before people got sick of benchmark maxing, distributed training runs before it was interesting, YARN to massively increase context length back when you couldn't dump endless text into LLMs and millions of downloads for their fine tuned Hermes Models. There were also a lot of fun projects along the way like WorldSim (massive world simulation) and FORGE. Fast forward to today and there are 24M+ downloads of Hermes Agent since February, #1 among apps on OpenRouter by daily token usage, and 250,000+ GitHub stars along with just mass virality. The Nous team estimates Hermes drives approximately 2.5% of all global token usage. Nous has created two things that I think are both foundational and sticky First, Nous created a religion and an authentic cult like following. They achieved this through grassroots interest and involvement from their community and users, and by actually embracing them. A recurring example is Teknium replying to users at all hours of the day, working with them and integrating their pull requests into Hermes agent. The project’s ethos is open source and against a totalitarian end state where a room full of folks controls AI and reality itself, and that resonates globally. The second reason is that Hermes agent unlocks a fundamental new spot on the AI optimization frontier as the open source non lab owned option. Every major competitor is owned by an AI lab (Muse owned by Meta, OpenClaw by OpenAI, Grok by SpaceX, Instinct which is closed source). With Hermes Agent you own everything you create around an AI model so you have that as a long term evolving asset (Business or personal) instead of getting IP farmed Also with Hermes agent you can drive differentiated outcomes versus using the same model as everyone else. Individuals and businesses have a comprehensive set of owned depth around the model (Evaluations, memory, interactions, data, private IP, variety of open or closed or fine tuned models, and more). This lends itself to an AI agent that grows with you that you own for life. Both @satyanadella and @PalantirTech’s have shared this view that corporations don’t want to give away their IP anymore and be farmed. They want to own it as an asset. I’ve shared my full thesis on this concept of owning vs renting your intelligence here: xitter.cf/Shaughnessy119/status/… Hermes Agent is not just a random spark, it’s the result of a team that has spent years building products, building systems and being fully enveloped in the AI world. The thousands of micro decisions and the culmination of the team led to Hermes Agent. So what comes next? When you have tens of millions of users already, they can become the champions within a business to bring their Hermes build to all employees of a company. They can start sharing skills, interactions, interop, data and more and build something truly multiplayer for enterprise. Having Nvidia, Microsoft and Samsung as investors doubles as enabling them to be extremely smart in potentially onboarding a massive enterprise user base to Hermes Enterprise. This isn't just private businesses using Hermes internally; it's Fortune 500 companies who can offer customized and unique agents to their customers and users built on Hermes Enterprise. Hermes' agent was not built to be a casual or fleeting personal agent and this lends itself to a larger business expansion. As Teknium shared “I fully intend and have plainly stated many times that I always built hermes to be the most powerful AI Agent. "Normie" consumers can leverage that and we do what we can to make it an experience they can approach (the mobile app will be almost exclusively focused on consumer) - but they are not our only demographic. I want the scientists, the cybersecurity engineers, the developers, the knowledge workers, the creatives, and everyone else to be able to have an agent that they can leverage that is maximally useful to them too.…” Nous has captured the hearts and minds of the smartest user base on earth and that's the wedge. Nous is building for a future where you have sovereignty over your AI stack. They are the center of gravity for Open source AI developers, users, investors and supporters. The AI discussion has been dominated by fear mongering driven by centralized labs to obtain a regulatory moat. This is very obvious. The safest path in my opinion is open source and acceleration. An entire world of agents pressure testing, building and dealing with each other aligned to the individual vs aligned to a centralized team of decision makers at a lab. Nous is the answer. Nous has now raised $158M to date to execute this vision. Internally we have people who have built insane Hermes agent builds including @KSimback and @pierskicks. I have built an insanely deep Hermes build that takes in 40+ private and public data sources and finds me new founders to invest in, helps manage my liquid portfolio, and tracks my entire historical journey as a VC (past founders, deals, much more). A few fun content items through the years has been our AMA with Teknium and our Podcast with Jeff AMA: xitter.cf/Shaughnessy119/status/… Podcast: members.delphidigital.io/med… Very excited for @robotventures (@tarunchitra and @rleshner) to lead the round alongside @nvidia @Microsoft @Samsung @usv (@fredwilson), @MenloVentures (@deedydas and team), @ycombinator and @1789Capital. Tarun is extremely deep on AI, inference (see his latest article) and open source and is the ideal lead investor in Nous. Also amped for the Hermes Mobile App Our goal at Delphi Ventures is to be the biggest supporters and investors of our portfolio companies. We are extremely excited for the future of Nous. wsj.com/pro/venture-capital/…

Huge Bifurcation in American AI War 1/ True current frontier labs like Open and Anthropic. Google is not frontier. These labs currently have the best models but are arguing for max regulatory capture and to basically control the pace of intelligence. Safety is clearly used to propagate a reg cap agenda. Their models are truly insanely good and the final jump in intelligence can create massive amounts in economic benefit, its an exponential style return not linear. They should also own a % of the technologies their models create as per token billing makes no sense. There is zero shot these companies slow down on training (maybe slow down what they release) given China will not slow down and if we lose AGI, we lose everything. Model commoditization and chinese AI labs doing actually cool inventive stuff makes their business model a lot harder. 2/ The open source and sovereign first AI companies like Meta (open models and muse), Palantir (own your intelligence stack), Microsoft (Satya bullish open source and serving this to enterprise) and soon Google since they lost the frontier race and will roll out massive open source support imo. Nvidia included. These companies do not offer frontier models they offer 90% as good for 1/10 the cost but you can actually own your AI stack as a core business asset instead of being IP farmed. They’re not using doomerism as reg capture. They benefit from Chinese open source models and a lot of value is returned to enterprises and people. It’s a bet models commoditize and we get millions of not specific models (my old thesis) but AI architectures around models (data, memory, interactions, internal apps etc) you can point at any model i.e. Nous Research Hermes agent for enterprise is the best example of this thesis. Actually it may be more than 90% as good since its a custom offering for your business at a way lower price point and you can sell the IP later on. Closed vs Open Source Safety vs Accelerationism Reg capture vs market economy This dynamic has many names Who wins?

A fantastic article covering all the routes for Enterprise AI adoption by @KSimback Historically I always said we would have millions of fine tuned models that are highly specialized in a domain that don't have to reason about the world like a general model My view has shifted and expanded I now think we will have millions of proprietary AI workflows owned by each enterprise. The data, memory, evals, models, interactions, history, all of it will be owned by each business as they aggressively shift to owning their intelligence These hyper specific AI stacks (think a laser focused AI model for X, Y, Z) will compound and create entirely new services/apps within their swim lane a foundational model can't offer. This will compound. We haven't seen the fruits of the proliferation of fine tuned enterprise models but I'd expect a wave of narrow AI applications that you specifically use for each experience you are looking for The Global model -> Enterprise Owned Stack flow is happening and I think @PalantirTech, @satyanadella, @KSimback and yours truly feel strongly about it. The creativity and diversity comes FROM differentiating your AI stack not using a general model that is an average

If you’re an early stage founder solving for this thesis, get in touch! My thesis is enterprises mass switch to open source AI and creative solutions to cut costs as usage balloons

This is an incredible report by @KSimback on decentralized inference It's possible the U.S. government restricts access to god tier open source models in the future so this its worth understanding how this models works - specifically what parts are decentralized vs not

The most basic way AI could blow up imo. I'm not saying it does but this is the most obvious way I can see it happening - Per seat subscriptions are massively subsidized. The flat fee was priced way below what heavy usage actually costs - For real business use you have to move to the API anyway. Data protections, work integrations and compliance officer approval - On the API you pay metered rates, and businesses are burning credits way faster than the per seat pricing ever led them to expect - This is everywhere right now. Internally for us, Codex users, Uber torching its entire 2026 AI budget in 4 months, the Microsoft comments. Just go try an API I shared more on this here: xitter.cf/Shaughnessy119/status/… - And I don't think most businesses have the money to keep paying increasing API rates without a real change to how they operate (caps needed) - Because they have a cheap alternative. They can reach open source models through any aggregator (OpenRouter, Venice, Baseten, Together) and still get strong privacy. Venice private data centers, or E2EE/TEE serving GLM 5.1. More on open source inference provider raises here: xitter.cf/Shaughnessy119/status/… - And the discount is enormous. DeepSeek V4 codes within a hair of Opus on SWE bench at roughly 1/30th the price, and the cheapest open models run closer to 1/100th - Chinese labs open source frontier grade models. The model is the single biggest cost an inference provider has, and they get it for free - This idea dies if China goes closed source. That is actually bullish web2 AI labs, because if everyone is closed you pay up for the best intelligence. China goes closed source if they are tired of giving away an asset and they want the revenue and data flow to train new models - Is this showing up in web2 AI lab revenue yet? No. Revenue is off the charts. Anthropic went from 9B to 47B run rate in five months - So go forward, what happens? - I think revenue slowly starts leaking to the open source inference providers (see Venice usage, OpenRouter's $113M raise, Baseten is raising at $11B or triple its valuation in three months, on revenue that went from $200M to $600M annualized in a single quarter) - It doesnt move overnight, but it caps the labs ability to raise prices, and margins are already deeply negative. OpenAI is reportedly running near negative 122% - With margins that bad there is no cash flow, so the labs are fully dependent on outside capital to buy GPUs, train models, and keep subsidizing usage (I.e. see Google tapping $80b equity sale, granted 30b for employee RSU taxes. Clearly they think Equity is overvalued or you wouldn't sell it) - The break comes when that capital stops. Pricing is capped so margins cant improve, and the moment investors lose conviction on payback, the whole flow reverses - Why would they lose conviction on payback? Back to the start - the inability to improve margins or get businesses to pay more - This is also limiting, if we start making new drugs with AI or create entirely new businesses, you better believe people will pay up to the max for AI usage

I pulled 3,182 tweets analyzing @NousResearch Hermes Agent versus Claude Code to understand exactly why developers are choosing one or the other I want to figure out why people use Hermes vs Claude/ Claude Code Here is the breakdown using 3,182 tweets to distill the answer vs guessing 🧵