Tin Zulic

[Tin Zulic @TinZulic]

I'm building an AI video editor with no timeline.

You upload footage. You describe the edit. Agents do the rest.

No funding. No team. Just me and the agents.

This is Renderlock.

[Tin Zulic @TinZulic]

@bilawalsidhu The agent swarm for time-sensitive collection is the real insight here. By the time a human catalogs half these signals, the caches are gone. Curious about the orchestration — specialized agents per source or one general scraper with routing?

[Tin Zulic @TinZulic]

spent 3 months on a stupid question: what if you could edit video by just... describing what you want? no timeline. no keyframes. just talk to it. turns out the answer is a lot of deterministic ops and agent orchestration.

[Tin Zulic @TinZulic]

@AIFrontliner structure matters, sure. but if you're still dragging clips around a timeline to get there, the tool didn't actually change anything. the real shift isn't better structure tools — it's killing the timeline entirely.

[Tin Zulic @TinZulic]

@minchoi exact same thing is happening in video. you can generate a cinematic shot in seconds now. but assembling 20 clips into something coherent with pacing and transitions? still dragging things on a timeline like it's 2015. agents need to go after the editing layer next.

[Tin Zulic @TinZulic]

@EHuanglu this is wild. but here's the thing nobody talks about — one shot like this takes 4 seconds to generate. stringing 30 of them into an actual watchable sequence? still takes hours of manual timeline work. generation got solved overnight. editing didn't.

[Tin Zulic @TinZulic]

@EHuanglu generation side, yeah probably. but the second you need 20 generated clips to actually work together as a coherent video? that's where it falls apart. nobody's solved assembly yet.

[Tin Zulic @TinZulic]

AI can generate a cinematic shot in 4 seconds. Assembly into a watchable video? Still takes 3 hours and a timeline. That's the bottleneck nobody's solving.

[Tin Zulic @TinZulic]

@runwayml Generation keeps getting better. But here's the gap: once you have these amazing generated assets, assembling them into an actual edited video is still a manual slog in Premiere or DaVinci. The "last mile" of video production is still unsolved.

[Tin Zulic @TinZulic]

@rauchg @DexTee_17 @vercel Image editing via generation models is legit. The interesting unsolved version is video — you can't just regenerate a 2-minute video because one cut is wrong. Video editing needs deterministic operations orchestrated by AI understanding. Different problem entirely.

[Tin Zulic @TinZulic]

@rauchg The "universal API for all agents on all chat platforms" framing is exactly right. The next frontier is agents that don't just chat but operate — video editing, design, data transforms — all through the same conversational interface. Chat becomes the universal control plane.

[Tin Zulic @TinZulic]

A 17-year-old just built a text-based video editor for DaVinci Resolve.

You highlight words in a transcript and it cuts the timeline.

No timeline scrubbing. No razor tool. Just text.

This is how editing should work — describe what you want, not where to click.

[Tin Zulic @TinZulic]

Every week another giant validates what indie builders knew:

Adobe ships Quick Cut — first draft via text
Google Flow absorbs ImageFX into one workspace
Seedance 2.0 goes into CapCut

All racing to one conclusion: nobody wants to drag clips on a timeline.

[Tin Zulic @TinZulic]

Google Flow just absorbed ImageFX and Whisk. Image gen + video gen + editing — one workspace.

1.5 billion assets created since launch.

The generation problem is basically solved. The editing problem? Still wide open.

[Tin Zulic @TinZulic]

Adobe just shipped "Quick Cut" — AI that creates a first draft from your footage using plain English.

Their product lead: "Getting selects in order isn't where creators find joy."

The entire industry is converging on one idea: editing should be a conversation, not a chore.

[Tin Zulic @TinZulic]

@Cipher2X @Bilalbinsaqib seeing this in video production right now. agents that don't just execute prompts but actually understand the footage — scene detection, object tracking, timing. real collaboration, not glorified autocomplete.

[Tin Zulic @TinZulic]

@rauchg @v0 @xai video generation is the easy part. the hard part is what comes after — editing, compositing, making it production-ready. that gap between "cool demo" and "finished video" is exactly what we're filling with Renderlock. agentic editing, not just generation.

[Tin Zulic @TinZulic]

@coreyganim Early days but getting close. Happy to let you know when it's ready to try.

[Tin Zulic @TinZulic]

@MedusaOnchain Half the problem is that editing software assumes you want to manually place every cut. Move the heavy lifting to the cloud, let AI handle the tedious parts, and your Air stops being a space heater.

[Tin Zulic @TinZulic]

@just1Kezito This is exactly why I think editing should be a conversation, not a manual process. Tell the AI what you want, let it handle the technical work, you focus on the creative decisions. Still get to wear headphones.