On Monday, Google DeepMind and A24 announced a research partnership - Google investing a reported $75m, according to the Wall Street Journal - to develop AI tools for filmmakers. Plenty has already been written about what it means for the industry. I'm more interested in what it changes a few steps downstream, in the rooms where the graphics and props actually get made.

The terms are worth reading carefully. It's a non-exclusive research partnership: A24 gets access to DeepMind's research and infrastructure, and DeepMind gets a hand from A24's filmmakers in shaping the tools it builds. It explicitly does not include A24's content library or its data - Google isn't training models on the films. The studio is opening its process, not its catalogue. For a studio whose reputation is built on how it works, that's the more interesting thing to hand over. You can read the announcement on the DeepMind blog.

Why this lands in the art department

The early use everyone keeps pointing to is workflow - AI-assisted storyboards have been mentioned specifically, the kind of thing Scorsese recently put his name to. Storyboards sit upstream of almost everything I do. If previs and the look get locked faster, the art department gets the same questions earlier, and more often. Faster decisions upstream don't reduce the work down here; they just move it forward in the schedule.

The part I watch closest

Anything that touches on-screen content and clearance is where I pay attention. Tools that can draft interface mock-ups, signage or background graphics are useful, but they don't remove the decision that actually matters: what's real, what's fictional, what's cleared. That call still belongs in prep, with someone who knows the legal and period reasons it matters. A faster way to draft a phone screen doesn't change the fact that a person has to decide it shouldn't be a real logo - and handle it live on camera rather than in post.

The open question

The caution is real, and some of it is coming from inside the building. A24's audience skews young and largely wary of generative AI, and at least one of its own directors has called it a source of "creative rot". Both A24 and DeepMind are framing this as tools that stay invisible and in service of the work - not the prompted generation that makes people uneasy. That's the right framing. The test is whether it holds once the tools are in daily use and there's a schedule to hit.

For now it's a small, early deal - $75m is a modest bet next to an outright acquisition. But the direction is clear enough, and it's worth understanding from where I sit: not as a threat to the craft, not as a shortcut, but as one more thing that shifts when decisions get made, and who's in the room when they do.

Greece has quietly become one of Europe's busier shoot destinations, and the 40% cash rebate is almost always the headline. But after years of watching productions land here - features, series, commercials - the rebate is rarely the thing that trips them up. The friction comes from a handful of practical details that nobody mentions until they bite. These are the five I'd sort out first.

1. Lock the incentive paperwork before you roll

Greece runs a 40% cash rebate, capped at €8m per project, alongside a separate 30% tax-relief scheme - the two can be combined. Both are administered through EKOME and the Hellenic Film Commission, and the application goes in before production starts, after a short cultural test. The rebate is then paid out after the work wraps and spend is verified.

Two things catch people out: commercials, reality, sports and current-affairs formats don't qualify, and the paperwork rewards being early. Get a Greek line producer or production-services partner onto it in prep - not somewhere around wrap.

2. Don't default to an August shoot

July and August are peak heat and peak tourism at the same time. Locations, crews, hotels and suppliers are all stretched and more expensive, and a good number of vendors slow right down or close around mid-August. The light is beautiful, but you're paying a premium for the hardest version of the schedule.

Spring and autumn give you the same Mediterranean look with far less friction - and far more goodwill from the people you need.

3. Treat island work as its own logistics problem

An island shoot is not a mainland shoot with a ferry attached. Wind, sailings and thin local infrastructure mean almost everything gets shipped in, and a cancelled crossing can quietly eat a day. Confirm early what is genuinely sourceable on the island versus what has to travel from Athens, and build buffer days for weather into the plan rather than hoping.

4. Decide your on-screen and print graphics in prep

Real logos, newspapers, packaging and phone interfaces all carry clearance risk, and physical print has its own lead times - made worse by those August shutdowns. The cheap fix is deciding in prep what becomes fictional or cleared artwork, and locking print quantities early, rather than discovering the problem in the edit.

This is the part I live in, and it's why I build tools around it: Skenic catches these items in the script breakdown, and Mombie handles live phone screens on camera without post replacement.

5. Get the Greek details right

Signage, paperwork, licence plates, currency and period references read instantly wrong to a Greek audience when a foreign team guesses at them. It's the kind of mistake that looks fine on the monitor and then surfaces the moment someone Greek watches the cut. A local eye on the background keeps the world believable - and saves the reshoot.

None of this is a reason not to shoot in Greece - it's one of the best places in Europe to do it. It's just that the rebate gets you in the door, and these are the things that decide how the shoot actually feels.

Look at almost any contemporary scene and count the screens: a phone in someone's hand, a laptop on the desk, a TV in the background, a tablet at the till. Every one of them is a graphics decision, and every one is a potential clearance problem. The trouble is that screens are easy to wave away in prep - "we'll sort it on the day" - and expensive to fix once the footage exists.

Three ways to handle a brand on screen

Broadly, you have three options for anything visible and recognisable. You can clear it - get permission to show the real thing, which is slow, sometimes costly, and never guaranteed. You can avoid it - frame around it, blur it, turn it off - which limits the director on the day. Or you can design a fictional or cleared version - a believable not-quite-real newspaper, a packaging label, an app that reads true but belongs to no one. The third route is usually the one that keeps everyone safe and keeps control on your side.

The hard case: live screens

Static graphics are manageable. The genuinely awkward one is the interactive phone - an actor reading and sending messages, scrolling, taking a call, while the screen is on camera. Traditionally that means one of two things: shoot it live with a real interface and replace the screen in post (slow, costly, and a clearance exposure if the UI is a real product), or fake it with a static playback that the actor can't really interact with.

This is exactly the problem Mombie was built to remove. It's a real-time messaging-prop simulator: fully controllable, believable phone screens that the actor interacts with for real, on camera, with no screen replacement in post and no clearance risk from a trademarked interface.

Catch it in the breakdown

The common thread is timing. A screen flagged during the breakdown is a quiet design task; the same screen discovered in the edit is a reshoot or a VFX line item. That's why I push screen and graphic items into the script breakdown early - and why Skenic surfaces them scene by scene, so nothing recognisable reaches the floor without a plan.

Screens aren't going to get rarer in the films we make. Treating them as a prep decision rather than a post surprise is the whole game.

"Print-ready" and "printed" are not the same sentence. A design can be approved and perfect on screen and still be days away from existing as a physical object an actor can hold. On a shoot, that gap is where things go wrong - so it's worth understanding what actually happens between the file and the prop.

Lead times are real

Specialty stock has to be sourced. Certain finishes - foil, emboss, specific weights, uncoated period papers - take time and sometimes a minimum order. Aging and distressing a document by hand adds another pass. None of this is slow exactly, but it isn't instant, and it doesn't compress well the closer you get to the shoot day.

The August problem

In Greece specifically, a lot of print and finishing suppliers wind down or close around mid-August. If your shoot lands in that window, the network you're relying on is thinner than usual. It's an easy thing to plan around once you know it's there - and an expensive surprise if you don't.

Always print extras

Hero documents get crumpled, stained, signed, torn and reprinted across takes. A single perfect copy is a liability. Quantities should account for handling, multiple setups, and the reshoot nobody's planning for - it's far cheaper to print ten than to re-run a job at midnight.

Same-day changes and remote delivery

When a name changes on the day or a date needs to move, a local print network is the difference between a reprint by lunch and a problem that follows you home. For productions shooting in Greece, that's handled end to end through an Athens supplier network. For everyone else, the files travel instead - print-ready PDFs and layered source delivered to whatever the pipeline needs.

Print is unglamorous and easy to leave late. Treating it as a lead-time item, the same way you'd treat a build or a costume, is what keeps the hero prop on the table when the camera turns over.