Creativity Needs Context
- Jun 4
- 9 min read
Why the next decade of filmmaking belongs to connected tools
A director gives a note in a review session: "Make the forest feel like it's grieving." By the time that single line of intent travels from the screening room to the artist re-lighting the shot, it has passed through a producer's notebook, a spreadsheet, three chat apps, a task tracker, an email thread, and at least one phone call. At each handoff, a little meaning evaporates. The artist receives something closer to "darken the forest." Two weeks and several thousand dollars later, the note is re-explained, and the work is redone.
Nobody in that chain was incompetent. The talent was there. The vision was there. What failed was context — the connective tissue that lets an idea survive the journey from one mind, and one piece of software, to the next.
This is the quiet, expensive problem at the heart of modern production. And it is exactly the same problem the AI industry spent 2025 learning to name. When we built Movie Colab, we weren't just building another tool to add to the pile. We were building a way to keep context from leaking out of the pile. This post is about why that matters more than almost anything else right now.
01 — The hidden tax: the cost of living between applications
Start with a number that should alarm anyone who runs a creative team. In a five-week study across twenty teams at three Fortune 500 companies, researchers found the average worker toggled between different apps and websites nearly 1,200 times a day. Each switch costs only a couple of seconds — but they add up to just under four hours every week spent simply reorienting. Over a year, that is the equivalent of five full working weeks, or about 9% of all time on the job, lost to the friction of moving between tools.
~1,200 app switches per day · 4 hours lost each week · 9% of annual work time · 5 weeks per person, per year.
Researchers gave this drain a name: the toggling tax. At one company, employees switched roughly 350 times a day across 22 separate applications and websites. The tax isn't only about clock time, either. The deeper cost is cognitive. Work by attention researchers at UC Irvine found it takes an average of 23 minutes and 15 seconds to fully return to focus after an interruption. Separate research cited from Cornell put the recovery time after a single application switch at close to ten minutes.
The more things you are juggling, the steeper the curve. The computer scientist Gerald Weinberg modelled this decades ago, and the shape of it is brutal:
Hold this thought, because it is the whole argument in miniature: fragmentation has a cost, and that cost compounds. Now consider an industry built almost entirely out of fragmentation.
02 — The worst-case scenario: filmmaking is a chain of lossy handoffs
Film and animation production is, structurally, the toggling tax taken to its extreme. A single shot is touched by a screenwriter, a storyboard artist, a previs team, a director, a production manager, a dozen specialist artists across modelling, animation, lighting and compositing, a review supervisor, and an editor. Each of them lives in different software. The screenplay sits in one app. The storyboards in another. The schedule in a third. The 3D work is split across Maya, Blender, Unreal, Houdini and a render farm. Notes live in email and chat. Files live on a server somebody has to remember the name of.
A modern pipeline, as one industry guide puts it, is really a controlled handoff of creative intent, technical constraints and production data — moving from editorial decisions to assets, from plates to pixels. When the handoffs are clean, the audience never sees the machinery; they only feel the story. When they aren't, the mismatches surface late, in compositing, where fixing them is most expensive. Clean version control and clean handoffs are precisely what stop a small issue from becoming costly rework.
Here is the part nobody likes to say out loud: every one of those handoffs between disconnected tools is a place where the director's original intent quietly degrades. We can draw the decay.
The talent in this industry is not the bottleneck. The seams between the tools are the bottleneck.
This is why production-tracking systems exist at all — tools like the platform formerly known as ShotGrid. They try to bolt a layer of coordination on top of the chaos. But coordination is not the same as continuity. Knowing that a note exists is different from the note arriving with its full meaning intact, attached to the exact frame, the exact version, the exact asset it refers to. Most legacy systems were designed for the VFX workflows of a previous era — not for real-time, game-engine, AI-assisted production where iterations happen by the hour.
03 — The AI parallel: the whole industry just rediscovered the value of creativity context
Here is what makes this the right moment to talk about it. In 2025, the most important shift in applied AI had nothing to do with bigger models. It had to do with context.
For a couple of years, the craft of getting good output from an AI was called prompt engineering — finding the right words. Then, around the middle of 2025, the field's leading voices reframed the problem entirely. Andrej Karpathy popularised the term context engineering: the art and science of filling a model's working memory with exactly the right information for the task. Anthropic described it as the practice of curating and maintaining the optimal set of information available to a model as it works.
The crucial insight underneath the rebrand was this: most AI failures are not model failures — they are context failures. When an agent gives a wrong or irrelevant answer, the underlying model is usually working fine. What's broken is the information it was handed. A support agent that quotes last quarter's prices doesn't have a reasoning problem; it has a context problem. Research that year on what some called "context rot" even showed that models use their context unevenly — pile in too much disconnected, poorly-curated information and performance gets worse, not better.
The three phases of working with AI:
Copy-paste. Paste text into a chat box, get a reply. Useful, but every session starts from zero. No memory, no continuity.
Frozen context. Pre-loaded assistants with fixed instructions and documents. Better — but the context is static and disconnected from what's actually happening in the work.
Agentic, living context. Agents that take real actions across real systems, drawing on live, connected context. This is where the leverage is — and it only works if the context is continuous.
Read those three phases again with a production hat on. They are exactly the maturity curve of film software. The disconnected stack is Phase 1 — every tool starts from zero, knowing nothing about what came before it. A bolted-on tracker is Phase 2 — some shared documents, but frozen and detached from the live creative work. The thing everyone is racing toward, in AI and in production alike, is Phase 3: a system where context flows, continuously, so that neither a human nor an AI agent ever has to be told twice.
In AI, the model is becoming a commodity. The context is the moat. The same is now true for filmmaking.
04 — Our answer: vertical integration is context engineering for film
This is the idea Movie Colab is built around, and it is why we made the unfashionable choice to build a vertically integrated suite rather than one clever app. Six tools, one continuous spine of context. The script knows about the storyboard. The storyboard knows about the schedule. The schedule knows about the asset. The review knows about all of it. Information is published between the tools instead of being re-entered at every stage.
Script Viz (Story) — AI breaks the script into a shot list, generates style-controlled storyboards, and runs a script analysis with sentiment graphs plus estimates of shots, characters and props. Produces the first structured context.
Table Read (Sound & pacing) — turns the script into a voiced performance with per-line control over tone and pace. Timeline layers keep notes and metadata attached to every line. Adds performance and timing context.
Projects (Production) — the mission control. Tasks link directly to their creative versions, so planning and execution stay in sync, with a full audit trail of who did what. The live context backbone.
Sync Space (Assets) — a single source of truth for files across hybrid studios: version history, file locks and change lists that keep everyone on the correct version. Keeps the asset context coherent.
Movie Colab VR (Review) — step inside the scene at true scale in a private theatre, reviewing the work with the immersion it was made for. Context experienced in space.
Artist Hub (Talent) — match the right artists to the work, with a match score and a portfolio that travels with them, onboarded with full project context from day one. Connects people to the context.
The clearest proof of the idea is a feature inside Projects we call the Review Agent. After a frame-accurate review session — annotations, comments, version comparisons and all — you activate it. It reads the entire conversation, summarises the creative direction, and turns it into a clean list of tasks for the artists. No note is lost. No intent is paraphrased into something weaker by the third re-telling. The director's "make the forest feel like it's grieving" arrives attached to the exact frame it was said about.
This is the toggling tax run in reverse. Instead of a human spending four hours a week re-orienting and re-explaining, the context carries itself.
And because context is structured and continuous, it also reaches outward. Our plugins let artists publish straight from Unreal, Maya, Blender, Unity and the Adobe Suite into Movie Colab — no manual export, no re-upload, no lost link between the work and the task it belongs to. The interoperability isn't a convenience feature; it is the point. It's how the context spine stays unbroken even when the creative work happens in third-party software.
05 — The compounding return: why connected context is worth more than the sum of its tools
Here is the economic heart of it. When tools are disconnected, value is additive at best — each app does its job, and the gaps between them quietly subtract. When context flows, value compounds. The script analysis makes the storyboard smarter. The storyboard makes the schedule more accurate. The accurate schedule makes the review more focused. The structured review makes the next iteration faster. Each stage inherits the intelligence of the one before it.
This is also what makes a connected pipeline the perfect substrate for AI. An AI agent is only as good as the context it can reach. Point an agent at a fragmented stack and it inherits all the gaps — it becomes another worker paying the toggling tax. Point it at a continuous context spine, and it can finally do the thing everyone wants from AI in production: read the whole situation, understand the creative intent, and act on it reliably. The pipeline becomes the context layer that the agents run on.
Everyone else is selling you a model. We're building the thing that makes any model useful: a place where your creative intent never has to be repeated.
06 — What it means for you: the practical payoff
If you run productions, the implications are concrete. You stop paying to re-explain decisions. Notes survive intact from the screening room to the artist's desk. Onboarding a new artist mid-project stops being a week of catching them up, because the context is already there to inherit. Rework caused by drifting intent — the most demoralising and expensive kind — drops, because intent stops drifting. And when you bring AI into the workflow, it works with your full production context rather than guessing in the dark.
The film industry has spent decades treating its fragmented toolchain as the unavoidable cost of doing business. The AI era has made the alternative impossible to ignore: in any complex creative endeavour, the team that preserves context end-to-end out-creates and out-paces the team that doesn't. Not because they have better people or better models — but because their ideas survive the journey.
That is what we are building. Not another app for the pile. A way to keep the most valuable thing in any production — the original spark of an idea — alive all the way to the final frame.
Context makes communication clearer. Context makes production faster. Context is the craft now.
Sources & further reading
Harvard Business Review (2022), How Much Time and Energy Do We Waste Toggling Between Applications? — the ~1,200 switches/day and 9% figures.
Gloria Mark, UC Irvine — research on attention and the ~23-minute return-to-focus after interruption.
Gerald Weinberg, Quality Software Management: Systems Thinking — the context-switching loss table (Chart 01), via Zoho Workplace.
RingCentral — workers switch communication apps ~10×/hour; two-thirds lose up to 60 min/day navigating between apps.
Karpathy & Anthropic (2025) on context engineering (CIO; Stacker).
Chroma (2025), Context Rot study — models use context non-uniformly as input grows.
VFX pipeline as a "controlled handoff of creative intent" — MimicVFX; CADA.
Product details from the Movie Colab site: Projects, Script Viz, Table Read, Sync Space, Integrations.