AI on the Lot reports filmmakers and studios are using AI in active projects

• According to venturetwins' event report, the strongest signal from AI on the Lot was not abstract enthusiasm but filmmakers, executives, and studios already using AI on real productions.
• venturetwins' Jon Erwin quote framed hybrid production as a jobs story in Los Angeles, with Erwin saying one project employs about 600 people and arguing that AI-backed workflows can pull work back from overseas outsourcing.
• In zaesarius' Gareth Edwards recap, Edwards said he has spent nine months testing diffusion models and now uses them for idea organization, concept testing, and development imagery, even as he hesitates to lock a feature onto tools that keep changing every quarter.
• zaesarius' Paul Schrader recap pushed the conversation further, saying the box office test is a fully AI-generated protagonist, "not a hybrid," then opening ChatGPT on stage to draft a treatment live.
• The cultural fight is already hitting creators directly: venturetwins' post about Jorge, wilfredlee's reaction, and Uncanny_Harry's reaction all describe backlash severe enough to push at least one filmmaker away from an opportunity.

You can see the mood shift in venturetwins' floor report, the jobs argument in the Jon Erwin clip, and the creator-side tooling in Kling AI's RAPHAEL showcase. There were also more practical tells around the event, including Dave Clark's seminar on hybrid workflows, a CapCut and NAKID Pictures workshop post, and a session on world models.

The clearest change at AI on the Lot was how little of the conversation sounded hypothetical. venturetwins described a day spent with filmmakers, executives, and studios already using AI in active projects, then called it a tipping point.

That lines up with venturetwins' Jon Erwin post, which put labor and production economics at the center of the pitch. Erwin said one project employs roughly 600 people and argued that hybrid production can bring work back to Los Angeles because so much conventional production has already been pushed to Eastern Europe.

A second attendee report from BLVCKLIGHTai's day-one recap said the most concrete conversations were around distribution and production decision-makers, not just model demos. That is a more useful signal than stage rhetoric, because it suggests buyers and platform teams were in the room.

According to zaesarius' recap of Gareth Edwards' comments, Edwards has been personally testing diffusion models for nine months and already uses them to organize ideas, test concepts, and generate development images. He drew a hard line at authorship, saying the filmmaker still owns the vision and the reason the story needs to be told.

The interesting part was the hesitation. In the same recap, Edwards said he has not committed to a project yet because the tools are changing every three months, which makes a fixed production timeline feel unstable.

That combination, active use plus reluctance to lock a production around the current stack, is probably the most honest creative read from the event. GenMagnetic's post confirms Edwards appeared onstage at AI on the Lot, but the sharper details came from the longer attendee summary.

Paul Schrader's threshold was more specific. In zaesarius' recap, Schrader said the real milestone is a film with a fully AI-generated protagonist that makes money at the box office, "not a hybrid."

The same post says Schrader opened ChatGPT on stage and wrote a treatment for a project called "The Collection Agent," prompting for a protagonist in the mold of Clint Eastwood. That is a small scene, but it turns a culture-war abstraction into a very literal workflow, veteran director, live text box, instant story draft.

The same recap also folded in a March 2026 interest study claiming 33 percent of U.S. consumers believe AI can make emotionally resonant content, versus 78 percent of industry professionals. Even with that gap, the fact that Schrader's benchmark is now box office proof, not technical feasibility, says a lot about where this conversation has moved.

The counter-signal was immediate and personal. venturetwins' post about Jorge said the creator described AI as a "miracle" for getting new series made and for bringing jobs back to Los Angeles, then still ended up getting canceled for it.

wilfredlee's reaction said Jorge deleted an original tweet after mass backlash and attacks directed at his family, while Uncanny_Harry's reaction described the episode as bullying a filmmaker out of an opportunity. starks_arq's post argued the decision handed power to the loudest anti-AI voices.

Whatever side someone takes on the tools, that is already a production fact. The pressure is no longer limited to studio policy, festival panels, or union statements. It is landing on individual creators in public, in real time.

The event was also thick with workflow vendors and hands-on sessions. Kling AI's RAPHAEL showcase positioned its feature-film case study as an end-to-end workflow, from ideation to final cinematic frames, which is exactly how these tools are increasingly being sold to filmmakers.

A few other posts fill in the floor plan:

• promise_ai's Dave Clark seminar post described a master's seminar on hybrid production workflows.
• dustinhollywood's workshop post promoted a CapCut and NAKID Pictures session.
• DavidmComfort's retweet of the Stage 15 session advertised an "Inside the AI Film Studios" block built around prompt, render, iterate, refine.
• promise_ai's world models post showed Mariana Acuna Acosta speaking about world models.
• Autodesk Flow Studio's Cannes post tied Wonder Dynamics' tool pitch to filmmaker control and production practicality.

That matters because it adds one more concrete read on AI on the Lot: the story was not only directors talking philosophy. It was also a market of specific production tools, training sessions, and pipeline language that already feels normal on the conference floor.