Cris Valenzuela on Creative AI, Runway and Building Through Change

Cris Valenzuela met Runway co-founders Alejandro Matamala and Anastasis Germanidis at NYU’s ITP program, which combines art and technology. Drawing on film, programming and business backgrounds, they experimented with early machine-learning tools for creative work. He says the company emerged from wanting to continue that exploration together.

Their experiments included video search and browser-based writing tools built before today’s generative AI wave. They later explored image and video generation. Valenzuela describes Runway as an effort to bring research into artistic practice, and as co-founder and CEO has a direct interest in the company and its products.

Valenzuela has often felt like an outsider: he studied business while making films, explored art while learning programming, and started a company from art school. He says a beginner’s perspective encourages first-principles questions but can make progress slower. For him, combining disciplines is also a source of creativity.

Growing up in Chile, he experienced the culture as traditional and restrictive and wanted room to follow his curiosity. His concern is becoming too comfortable to push intellectually or creatively. These are personal reflections about his own experience, not a general assessment of Chile or its people.

At Runway, he values the ability to figure things out and finish work, especially in a fast-changing field. The founders had to learn how to start a company and remain in the US. He describes learning by building a neural network from the ground up, curating data and working through obstacles rather than relying on reading alone.

When learning becomes difficult, Valenzuela recommends asking capable and kind people for help. He says fundraising, hiring and releasing Runway’s Gen-1 and Gen-2 models have all required a willingness to be wrong. Actual use can overturn assumptions about what creative people want from a model.

One lesson, he says, was that a polished interface matters less than helping people get value from the models. Because the technology was changing quickly, a carefully designed interface based on an early assumption might soon need to be discarded. This is his product judgment, not a claim that usability or accessibility never matters.

Stebbings asks how users can have a good experience without guidance. Valenzuela says Runway must test with creators: video generation is a new medium, and people may discover forms its builders did not anticipate. He compares it to putting an early camera in artists’ hands and learning how they use it.

He says educating users is hard when capabilities change rapidly. At the recording, he says Gen-2 had moved from four-second to 18-second generation, changing what users could attempt; this is a description of Runway at that time. He favours early public releases because feedback from filmmakers and artists can challenge the company’s assumptions.

Fast time to value still matters. Valenzuela says a preview-frame feature made the generation process easier to understand and increased usage, but gives no figures. He also explains that free access and an unlimited plan encourage creative experimentation; pricing and plan details are his account of the product at the time.

Stebbings asks whether AI is enabling creators or ultimately replacing them, citing the writers’ strike. Valenzuela argues public discussion is too focused on language models and that creative work involves feedback, iteration and people. He wants a wider range of stories about AI and invites sceptics to try the tools.

Asked directly why screenwriters were striking, he questions the idea that typing a prompt can replace the whole process of making a script or film. This is the view of the CEO of a company selling generative tools. The interview does not hear from writers or resolve their concerns about employment, compensation, consent or the use of their work.

On model size, Valenzuela says larger models can support broader capabilities, while smaller models may be suited to specific tasks. He does not expect one model to serve every use case. These are his 2023 expectations, and actual performance depends on task, training data and implementation.

He says Runway was updating models on roughly a weekly cadence. In his view, a particular model is not a lasting moat: the people building it, the speed of learning and feedback from users matter more. That is a competitive thesis rather than proof that model technology has no defensible value.

A false factual answer is a problem when a user asks for a fact; in image or video creation, an unexpected output may sometimes be useful. Valenzuela treats this as a context-dependent distinction, not a claim that unreliable outputs are harmless. Creative users still need to decide what to keep or reject.

He sees open models as a way to reach more people and invite unexpected contributions, but argues that openness is only one product decision. He says Runway has released some models openly while keeping others proprietary and may not open all of them. His position reflects choices made by the company he leads.

The constraint he most wants to remove is speed: compute, model training, deployment and efficiency can all slow progress. He says Runway had about 55 people at the time and deliberately kept the team lean. As an organisation grows, he expects coordination to become harder.

Valenzuela says he has made hiring mistakes by placing too much weight on credentials and previous employers rather than a candidate’s drive and ability to execute. Runway uses interviews and exercises to assess both baseline domain skills and getting things done. He considers the process strong while acknowledging his own bias.

Humility matters because teams may spend months on work that has to be abandoned. He looks for people who ask questions, adapt and are willing to learn rather than insist there is only one way. Stebbings adds that he values capable people without arrogance.

Valenzuela is sceptical of universal company-building recipes. Runway organises teams in groups it calls ensembles; he says an attempt to adopt objectives-and-key-results practices because an investor recommended them did not fit its needs at that stage. His lesson is to test practices against the work rather than copy them automatically.

Fundraising initially felt like a cold start because he did not know Silicon Valley investors. He advises founders to question prospective investors as closely as investors question them. In one case he says an investor offered attractive terms but described Runway’s purpose in a way that did not match the founders’ vision, so he declined.

He believes investors can help with introductions, but cannot substitute for a team building the company through difficult daily work. His hardest round was the Series A, when he says many investors rejected the idea that generative AI would transform media. He recalls doubting himself but continuing to pursue the thesis.

Looking back, he wishes he had communicated Runway’s long-term vision more explicitly rather than assuming others understood it. Products and models change, he says, while the team’s shared direction gives continuity. His claim that generative tools could transform media is a forecast and a thesis underlying his company.

Asked about a high valuation, Valenzuela says optimising for valuation can distort decisions away from building a durable company. He imagines a much larger outcome if Runway succeeds, but that is his aspiration rather than a forecast supported by financial results. The interview does not establish current revenue or cash returns.

Venture funding was, in his account, the practical route for three immigrant founders without significant savings or family capital. He says outside investment helped them hire and pursue the work. He does not believe adding a single famous person to the cap table would replace the efforts of the team.

Valenzuela believes the scale of AI’s eventual impact is not yet fully understood, while admitting current assumptions may be wrong. Asked about work and family, he says he does not see a sharp separation between them and emphasises caring for people around him. This is a personal view, not a tested prescription for balancing parenthood and work.

In a quick-fire round he says AI founders need not be based in Silicon Valley and reiterates that the value of a creative model matters more to him than a fancy interface. He describes building a company as persistently difficult, requiring resilience and attention to what can be controlled.

Looking ten years ahead, he hopes AI will become ordinary infrastructure people use without naming it, much as they use the internet. His ambition for Runway is to make movies. Both are forecasts or aspirations voiced in 2023, not guarantees about future technology or the company.