Rec · CutToAi · London
We do not sell AI. We test it, teach it, and tell the truth about it.
Tools, research and training for filmmaking in the age of AI, built from real productions and real cohorts rather than demos.
The work
Evidence, not claims
A running record of what has been built and tested, with the failures published alongside the results.
The book
Filmmaking in the Age of AI
A practical guide from story development to distribution, with the filmmaker's judgement at the centre.
View on Apple Books →
The platform
CutToAi Experiments
Nine working instruments built from real production problems, each with what it does not do stated on the card.
See the instruments →
Thirty years inside the work, before any of it was pointed at AI
The studio is built on three decades of international broadcast and documentary production and eight years teaching it at Levels 4 to 7. That is why the tools here are judged against a working pipeline and a real cohort rather than a demo reel.
The practice behind it30 years
Broadcast and documentary production
Best Film
UK Competition, Sheffield Doc/Fest
Experience across film and education
What we do
One practice, worked from three sides
Filmmaking, film education and critical AI practice are not three separate businesses here. They are one practice, and the speaking is where most of it reaches people.
How it works
From enquiry to the room
01
The enquiry
Tell us the audience, the date, the format you have in mind, and what you want people to leave with.
02
The call
A short call so the session is pitched at your people rather than a generic audience. If we are not the right fit, we say so.
03
The session
Tested examples, the failures included, and enough hands-on time for people to try things themselves.
04
The follow-up
Slides and examples shared afterwards, with any AI used in preparing them disclosed.
The method
What if the question were not what the tool can do, but what you can hand over?
Four tests, one task
- Delegate What can be handed over without loss?
- Describe Can you specify what you want?
- Discern Can you tell whether it is any good?
- Diligence What must be checked regardless?
Four tests run against a specific task, and any one of them failing is a reason to keep the task. Read the framework. The 4Ds were developed by Rick Dakan and Joseph Feller in partnership with Anthropic, and are used here under CC BY-NC-SA 4.0.
Talks
Three rooms we work in
Sessions are built to a brief. Fees depend on format, preparation and travel, so they are given on enquiry.
Six subjects are listed in full on the speaking page, and sessions are also built to a brief.
Recent writing
No predictions. No tool of the week.
Findings from the lab and the classroom, written up as they happen.
In the founder's words
Why this exists
“I started CutToAi because the industry's response to artificial intelligence is fracturing into blind enthusiasm and defensive panic. Neither serves the craft. Drawing on many years in UK higher education and active post-production workflows, I build solutions and frameworks that test these tools against reality. This is where I do that work in public, holding AI accountable to the discipline of filmmaking.”
Mostafa Nagy. Thirty years in broadcast, eight teaching it
Read the full statementNew tools and findings, when there is something worth sending
No schedule, no filler. You will hear from us when a tool ships, a claim gets tested, or something turns out to be wrong.
Or read the archive on Substack.
Book a talk
Tell us who is in the room
Programme review, AI-adoption strategy, workshops, or a production that needs a second pair of eyes on where AI actually belongs. We take on a small number of engagements at a time, and say no to the ones we cannot do properly.





