kenny miller

A little funny, scary thing we did. featuring the talents of @brianwankum and @rakelle cc: @MachineCinemaAI @CuriousRefuge youtube.com/watch?v=QiHadn…

new model for engineering team structure in 2026: 2 people only one pirate and one architect the pirate's job is to move as fast as possible to develop valuable, shipped product features by vibe coding. the architect's job is to turn the product surface discovered by the pirate into a reliable, structured machine—also by vibe coding, but at a slower, more well-reasoned pace. every product needs a pirate but most product's only need an architect once they some form of PMF, and in that case they usually don't need one full-time. architects can work across many codebases and solve interesting technical challenges. pirates go hard on a product that they own end-to-end.

“ video generation becomes a live creative loop.” Let’s go team @runwayml

Directable AI Set to Disrupt Film and TV Biz: ‘It’s Going to Help Small Film Industries in Asia Compete’ variety.com/2026/film/asia…

Junkyard King - Ep. 0 | MAKING OF ⚙️ HERE’S HOW I MADE THIS, STEP BY STEP 🧵👇 1/14

"Quite literally every frame in this movie is a painting.” FRANKENSTEIN, a film by Guillermo del Toro, wins the Academy Award for Production Design.

Spent about $1000 in credits on Seedance 2.0 over the last few weeks,and here are a few thoughts: First, the main thing that strikes me using a state-of-the-art model from this new generation is how hard it still is to scale beyond short form. Getting great animation is fast. Getting multi-cut sequences that make sense is possible. Consistency with Omnireference is actually very good. But the moment you move into real narrative work, things change. Multi-character exchanges, long sequences, maintaining visual continuity across shots, keeping tone, pacing, and staging consistent… it’s not impossible, but it is still a lot of work. And with generation costing somewhere between $2 and $7 per ~15 seconds, it adds up very quickly. As models improve, producing good looking short content is becoming trivial. Building something that holds together as a story is still not. Continue Video in Seedance is clearly trying to address part of this, but in my case it has been broken for the last couple of weeks, so none of my longer attempts would go through. In theory, you could imagine a small team of 5–10 people generating all day from the same storyboard, using a shared visual reference as a single source of truth. That alone shows how close we are to something that starts looking like a real production pipeline. But we are not fully there yet. Right now it still feels like we can touch the future with the tip of our fingers, while at the same time struggling to precisely steer a model using mostly words, references, and iterations when the narrative becomes complex. Short clips are easy. Worldbuilding is not. And storytelling is still the hardest part.

The part nobody talks about in the AI film debate: The traditional pipeline was already brutal before any of this. You graduate from film school - almost always in debt unless you are born very lucky. You spend years as a PA or an assistant for someone awesome (if you’re very lucky.) But most just grind on sets for nothing. You try to get into the guild. You wait for someone to give you permission to make something. It only comes if you’re very very very lucky… sometimes talented.., but always lucky. And most, even the most talented of the top film schools in the world wash out and do something else by the time they’re 30. Not because they lacked craft. Because the system has a thousand gatekeepers and almost no doors. AI isn’t breaking the system. The system was already broken. We just have a scapegoat now…. And an opportunity.

@graveair When prestige and hundred million dollar budgets are at stake, you get “safe bets,” but AI film balances us back towards new voices trying new things at much lower costs - I have faith we get a New Hollywood like the 70s.

While we're talking about the brilliance of Spielberg. Here's a thread of blocking breakdowns that I've made over the past few years. Munich has some of my favorites.

The token cost to build a production feature is now lower than the meeting cost to discuss building that feature. Let me rephrase. It is literally cheaper to build the thing and see if it works than to have a 30 minute planning meeting about whether you should build it. It’s wild when you think about it. This completely inverts how you should run a software organization. The planning layer becomes the bottleneck because the building layer is essentially free. The cost of code has dropped to essentially 0. The rational response is to eliminate planning for anything that can be tested empirically. Don’t debate whether a feature will work. Just build it in 2 hours, measure it with a group of customers, and then decide to kill or keep it. I saw a startup operating this way and their build velocity is up 20x. Decision quality is up because every decision is informed by a real prototype, not a slide deck and an expensive meeting. We went from “move fast and break things” to “move fast and build everything.” The planning industrial complex is dead. Thank god.

We built our own Brand Agent for Lovable 🔥 Our team can now search, request, or even generate brand assets directly through an AI agent. It’s trained on our brand guidelines, values, and design system so whatever you ask it stays on brand. Why this is powerful: 1/ One source of truth for all brand assets 2/ Consistency across brand, design, and marketing 3/ Brand knowledge embedded directly into an AI agent 4/ Accessible to anyone We love building internal tools!! Go build your own brand agent!

Video.js v10 beta is here! Video.js v10 is a complete rewrite combining four (!!) open source players into one. With 88% smaller default bundles, first-class React and TypeScript support, composable architecture, and beautiful new skins, you’re sure to fall back in love with your video player ♥️ Read @heff’s blog (and check out the site’s new look): videojs.org/blog/videojs-v…

𝗪𝗵𝗮𝘁 𝗶𝗳 𝟰𝗗 𝗩𝗼𝗹𝘂𝗺𝗲𝘁𝗿𝗶𝗰 𝗩𝗶𝗱𝗲𝗼 𝗰𝗼𝘂𝗹𝗱 𝗯𝗲 𝘀𝘁𝗿𝗲𝗮𝗺𝗲𝗱 𝘂𝘀𝗶𝗻𝗴 𝘀𝘁𝗮𝗻𝗱𝗮𝗿𝗱 𝘃𝗶𝗱𝗲𝗼 𝗰𝗼𝗱𝗲𝗰𝘀? 𝗜𝗻𝘁𝗿𝗼𝗱𝘂𝗰𝗶𝗻𝗴 𝗣𝗮𝗰𝗸𝗨𝗩 (accepted at @CVPR)! In 𝗣𝗮𝗰𝗸𝗨𝗩, we make 4D Gaussian scenes video-native. [1/4]

holy moly... effect medley #carsonmars #gamedev #allcode codec have mercy... :-S

me: "can you use whatever resources you like, and python, to generate a short 'youtube poop' video and render it using ffmpeg ? can you put more of a personal spin on it? it should express what it's like to be a LLM" claude opus 4.6:

New iteration on Parallel Timelines (working title). FLUX.2-Klein 4B now runs at a satisfactory real-time speed, balancing performance and visual quality. Hand tracking controls the split, letting me define where the transformed output appears; a more gestural and playful approach than the previous eye-tracked version. #ai #realtime #realtimeai #StableDiffusion

Doug Liman's Killing Satoshi starring Pete Davidson and Casey Affleck eliminates physical locations through AI generated backgrounds. The production uses AI to adjust actor performances in post production, replacing traditional reshoots. UK casting notices explicitly state the film will use AI for locations and performance modifications. Producer Ryan Kavanaugh positions the technology as "tool, not replacement" while implementing workflow that removes location scouts, background artists, and traditional post production roles. The Bitcoin biopic dramatizes the search for Satoshi Nakamoto, the anonymous creator of cryptocurrency. Kavanaugh frames the project as "Social Network of Bitcoin" through Proxima Media and Aperture Media Partners financing. The approach raises questions about labor displacement in production departments typically employed for location work, background elements, and performance continuity. SAG-AFTRA negotiated AI protections in 2023 contracts, but this production tests boundaries of performance modification consent. Note that Nick Schenk wrote the screenplay and Dixie Chassay handles casting. More info. studio.aifilms.ai/blog/killing-s…

Paul Schrader on AI films But what does this Paul Schrader guy know about films anyway so

Gaussian Splat relighting using @DSkaale's incredibly clever technique! Rendered in real-time on the web using @sparkjsdev, @threejs, and @theworldlabs

Love it, great work @GenMagnetic cc: @MachineCinemaAI