Dust AI Startup: Model‑Agnostic Strategy, Fundraising & Future of Work

Stan, co-founder of Dust, shared insights on building a company in the AI space, particularly in the shadow of large frontier labs like OpenAI and Anthropic. He discussed Dust's mission, the evolving nature of work, the challenges and benefits of being a horizontal platform, and the strategic decision to remain model-agnostic. He also touched upon fundraising in a competitive market and the unique aspects of building a tech company in France.

From OpenAI to Dust: A Journey to Product Building

Stan, an early employee at Stripe for five years, transitioned into AI, joining OpenAI when it was still possible for engineers to engage in research. He spent three years there working on large language models and mathematics. Despite the significant financial opportunity he left behind at OpenAI (stock options he gave up were worth more than Dust's valuation at one point), Stan has no regrets. His motivation for leaving was a desire to return to building products. He found research to be an intense but fleeting experience, providing brief highs that quickly faded, prompting a continuous search for the next discovery.

Dust's Vision: Applying LLMs to the Workplace

Dust's initial motivation was to apply large language models (LLMs) to the workplace, a focus that was considered niche in 2022 and early 2023. Stan believes that work will be profoundly transformed by this technology. He draws a parallel to how our great-grandparents, whose work was primarily physical, would perceive our current desk-bound jobs as "not work." Similarly, he anticipates that in two to three years, our current work will seem antiquated.

Dust was founded on the conviction that AI would disrupt work, a belief that is now proving true as the way people work, especially in development, has radically changed in recent months. Stan admits he was wrong about an "earlier plateau" in AI development; the technology continues to advance rapidly, constantly redefining work.

Navigating the Giants: OpenAI and Anthropic

Operating alongside giants like OpenAI and Anthropic presents both challenges and benefits for Dust. Dust chose an early horizontal platform approach, which, while powerful, makes the go-to-market strategy more complex compared to verticalized AI products.

The benefits of these large labs include their role in educating the market. Stan observes a convergence in the product space, with many products moving towards a "productivity suite" model for human-agent interactions, similar to Microsoft or Google Drive. This shift implies less product depth but broader integration.

Dust differentiates itself through a focus on collaboration and multiplayer AI, and by being "model-agnostic." Stan emphasizes that being tied to a single model provider is akin to buying machines from an energy provider whose plug only works with their energy, creating a dependency that can be problematic. He believes product players should be agnostic about where they source their intelligence. While a lot of usage might gravitate towards one provider at a given time, this can change dynamically, making model agnosticism a crucial strategy.

Fundraising and Company Building Philosophy

Dust's fundraising journey included an easy seed and Series A, but a more challenging Series B. Stan attributes some of this difficulty to building from France and the market's energy being absorbed by the large labs. He notes that growth investors might prefer to invest heavily in a single lab for quick returns.

Regarding company building, Stan and his co-founder Gabriel are known for their thoughtful approach, particularly in fundraising. They deliberately chose to raise at reasonable valuations, avoiding the "coffin corner" of raising too much money without realizing its potential, which can lead to a downturn. Their mantra was "no GPU before PMF" (Product-Market Fit), reflecting their focus on the product layer rather than heavy infrastructure investment early on.

Stan acknowledges the aggressive approach of companies like Mistral, which raised significant capital early despite its founder having no prior company-building experience. He sees a "gray zone" in balancing aggressive growth with a more measured approach.

Building in France: A Deliberate Choice

Despite the potential ease of building in the US, Stan and Gabriel chose to build Dust in France. This decision was driven by a desire to contribute to the French tech ecosystem and a sense of national preference or sovereignty. While it adds an "additional constraint" or "friction," Stan believes that when a company is successful, its location becomes less relevant.

Verticalized Products and Market Dynamics

For founders building verticalized AI products, Stan advises focusing on creating a defensible "moat," specifically through network effects. As models improve and intelligence becomes commoditized, the value of scaffolding around models diminishes. Products that can leverage network effects, where value increases with more users interacting on the same platform, will be more resilient.

Margins and Pricing in the AI Era

Margins are a significant challenge in the current AI landscape. Labs capture substantial margins at the token level, which can pressure product companies. Dust initially used seat-based pricing to encourage usage but has transitioned to credit-based pricing due to exploding usage and longer agent loops. Stan believes credit-based pricing is essential for product-layer companies to maintain margins, as users will maximize AI usage if given a flat rate.

He also points to historical precedents where companies successfully built products on top of infrastructure that captured margins. He hopes that the rise of open-source models will put pressure on the labs' "humongous" margins (estimated at 70-80%), recalibrating the market and creating more opportunities for product companies.

Final Words of Wisdom

Stan's advice for aspiring entrepreneurs is to find a clear vision or a problem they are passionate about solving. Building a company is inherently difficult, with many lows and few highs. Without a strong, driving purpose, the journey is unsustainable. He encourages founders to identify this core motivation, even if it evolves during the building process, to sustain them through the inevitable challenges.