Descriptions:
Nearly every hand in the room went up when she asked who had recorded video at the conference. Almost none stayed up for who had actually posted any of it. Ekaterina Deyneka counts herself in that gap, and Reelful is her answer to it: drop in raw footage with a line of direction, and an agent finds the usable moments, cuts them together, and generates captions, music, voiceover, and b roll around them.
Her framing for an AI engineering audience is that an agentic video editor is structurally the same thing as an agentic app builder. A prompt goes in, a sandbox spins up, an agent works inside it with tools and skills, and something renders out the other end. The difference that matters is editing rather than generating. A blank canvas lets an agent do anything it likes, while real footage forces it to judge which take is best and what to drop, and to produce something polished from material that is often messy or incomplete. The composition layer is Remotion, which expresses video as React code, chosen precisely because agents write code well. Skills carry the taste: cut rules, font pairings, when a cutaway actually helps. A verification pass catches compositions that will not render and sends the agent back around. All of it hides behind mobile templates, since the point is that a consumer never sees the pipeline at all.
Speaker info:
– https://x.com/katedeyneka
– https://www.linkedin.com/in/katedeyneka
– https://www.katedeyneka.com
Timestamps:
0:00 – Who recorded video here, and who actually posted it
1:29 – What agentic video editing means
3:33 – The same shape as an agentic app builder
4:10 – Editing real footage is harder than generating
5:30 – The pipeline, from media understanding to a creative plan
6:50 – Remotion, video as React code, and the verification layer
8:49 – Hiding all of it behind mobile templates
