Gunin Gupta

Co-founder TensorTest

[Pavan Kalyan @tpavankalyan23]

1/5

We read the last message from 896 AI video-editing runs.

93% said the job was done. 14% passed.

"Done" tells you nothing.

[Gunin Gupta @GuninGupta907]

@tpavankalyan23 @TensorTestTeam Hahahaha 😂😂

[Pavan Kalyan @tpavankalyan23]

Opus 5.5 made an ad for its own performance review.

Bold career move. 📷

Considering it only scored 32.1% on TimelineBench.

[Gunin Gupta @GuninGupta907]

@PamirEhsas Thanks @PamirEhsas

[Gunin Gupta @GuninGupta907]

Introducing TimelineBench: Evaluating Agents on Realistic Video-Editing Tasks, from Raw Footage to Final Cut

We tested 16 agents on 56 real video-editing jobs. The best passed just 26.8%

https://t.co/RmCUx8aroL

[Gunin Gupta @GuninGupta907]

@drexthealpha It will be soon up on https://t.co/7OJnlbGePy

[Gunin Gupta @GuninGupta907]

We’re TensorTest, building evals and RL environments for creative agents, starting with video editing.

We offer private agent evaluations with a breakdown of failure patterns.

Contact: [email protected]
https://t.co/G4g8bk12QH

[Gunin Gupta @GuninGupta907]

Agents often rendered their first cut 72–90% of the way through a run, leaving little time to review and revise

Our takeaway for builders: inspect the finished sequence for pacing, story and shot choice, alongside technical correctness