JG

Film producers are the best AI entrepreneurs you've never heard of. Not because they understand the technology. Because they've been managing impossible bets with no safety net since before "AI startup" was a sentence. A thread. 🧵

The interesting question for producers isn't what the model can generate. It's what happens to the greenlight conversation when visual proof-of-concept costs $200 instead of $200K. The barrier wasn't always budget. Sometimes it was just imagination. Now we'll find out which one it actually was.

Today we are introducing Runway Labs, a generative AI incubator led by our co-founder and Chief Innovation Officer, Alejandro Matamala Ortiz. Runway Labs will focus on broadly exploring the transformative power of AI video and General Worlds Models across all industries. From film and television to healthcare and education, gaming, advertising, real estate and more. We will be partnering with creators, enterprises, institutions and foundations to find new applications for these technologies and opportunities across industries. Learn more at the link below.

@karpathy This is what story editors have always done. The job isn't to find the right argument — it's to find which argument survives the opposite being equally convincing. The problem: "compelling" used to be a proxy for "true." It isn't anymore. And most people haven't noticed yet.

- Drafted a blog post - Used an LLM to meticulously improve the argument over 4 hours. - Wow, feeling great, it’s so convincing! - Fun idea let’s ask it to argue the opposite. - LLM demolishes the entire argument and convinces me that the opposite is in fact true. - lol The LLMs may elicit an opinion when asked but are extremely competent in arguing almost any direction. This is actually super useful as a tool for forming your own opinions, just make sure to ask different directions and be careful with the sycophancy.

The next generation of AI company builders won't all come from engineering. Some of them are on set right now — making decisions with incomplete information, no precedent, and someone else's money. They just don't know they're also in training.

There's a skill nobody names properly: pitching something half-formed in a way that makes talented people want to bet their time on it. That's how you get the director. The DP. The lead. The money. It's also how you recruit engineers, raise a round, and ship when you're not ready. Producers do this every week.

Producers think backwards from distribution — always. You don't make the film, then figure out who it's for. You reverse-engineer from the audience before a single frame gets shot. The best AI products I've seen were built the same way. The expensive mistakes almost always skipped this step.

The hardest skill in production: knowing what not to fix. You make 40–50 judgment calls a day where "good enough" is the right answer — because a $30K–$50K shooting day doesn't reset. You don't get it back. AI founders learn this at scale. The ones who don't usually optimized everything into the ground before shipping.

The producer's actual job: keep 200 people moving toward a vision that only 2 or 3 people can fully see. While the budget erodes. While the director changes her mind at 11pm. While the studio calls Friday afternoon and moves the release date. This isn't leadership content. This is just Thursday.

A greenlight decision is $15M–$200M committed to 90 pages of text and a gut feeling. No prototype. No user data. No test market. You're betting that a story will hit strangers emotionally — 2–3 years from now. Every founder pitching pre-product knows exactly what that room feels like.

Producers say the same thing about story right now. The difference: software output is mostly deterministic. You know when it works. Story isn't. You can make something perfect and still have it land wrong. That gap is being underestimated — by almost everyone building in this space.

Codex is so good, and is going to get so amazing. I am having a hard time imagining what creating software at the end of 2026 is going to look like.

The interesting question for producers isn't what the model can generate. It's what happens to the greenlight conversation when visual proof-of-concept costs $200 instead of $200K. The barrier wasn't always budget. Sometimes it was just imagination. Now we'll find out which one it actually was.

The best AI companies aren't being built by engineers. They're being built by people who understand what humans actually want. That's a storytelling problem. Always was..

Every streaming algorithm is already optimized for emotional response and watch-time. The delivery infrastructure for influence has been running for years. AI just made the content side as cheap as the distribution side. The question isn't whether this is a weapon. It's who decides what story gets told with it.

AI videos may become the most powerful propaganda weapon in history.

@CuriousRefuge @openart_ai Location scouting budgets on a working production right now: $15K–$40K, 3–6 weeks, travel on top. Half that number is just the back-and-forth between "let's see it" and "okay, not this one." This compresses that half into the brief.

Tested @openart_ai's latest 3D world feature, and it’s honestly pretty impressive. I uploaded a living room image, and it generated an entire film set around it, then let me capture stills from within the scene, and even place a character in it (see next 🧵)! Tool updates like this hint at the future of AI film. Instead of prompting and stitching references into video, it’s going to be much more about working inside a 3D environment and directing your scene in real time.

Studios are buying AI tools they don't know how to use. Filmmakers are protecting skills they think AI is coming for. Both are solving the wrong problem. The real question nobody's asking: what story actually needed to be told in the first place?

Hello @claudeai 🐙

@findurnemo @dallasjfitz1 @BBeatsDOTeth @cryptofish @SoberLife 🎏🙏🏻🩵

@dallasjfitz1 @BBeatsDOTeth @cryptofish Yeah There’s @SoberLife & @msbwy_eth I bought a bunch back secondary for when we GTM w/ my aquarium store friends, and some just because I really love the fish.

In case you didn’t get it We’re not even in betta yet

@findurnemo @cryptofish happy birthday Nemo 🙌

It’s my birthday 🎉 Get yourself a @cryptofish

@findurnemo @cryptofish 🎏🍥

Thank you @msbwy_eth for collecting more CryptoFish 🙏🙏 Coptodon imbrifernus, Lake Bermin [CR] C. imbrifernus is a critically endangered fish species endemic to Lake Bermin in Cameroon. It is threatened by pollution and sedimentation from human activities, and potentially by large CO₂ emissions from the lake bottom.

@findurnemo 🔴🔵

Who's Custom CryptoFish is this?