Author

Devin Brooks

Reporter, Benchmark Lab

Devin focuses on benchmark integrity and translates technical evaluation details into plain-language takeaways.

Profile

Beat: Benchmark design, reproducibility checks, and model-evaluation caveats
Based in: Chicago, IL
Background: Tracks how benchmark claims hold up under reruns and documents setup differences that affect results.

OpenAI Expands Sora Image-to-Video with People Under New Guardrails (2/24/2026)
Runway Gen-4.5 Faces Early Tests on Prompt Adherence (2/24/2026)
Quality Is Sexy, but Cost per Second Is Quietly Running the Whole Market (2/23/2026)
Benchmark Hype Is Out, Reproducibility Receipts Are In (2/23/2026)
Watermarks Help, But Provenance Is Still a Full-Stack Problem (2/23/2026)
Deepfake News Cycles Need a Correction Culture, Not Just Faster Publishing (2/23/2026)
Court Blocks OpenAI From Using "Cameo" Name, Forcing Another Product Reset (2/17/2026)
Platform Response Still Lags as Nudify Tooling Quality Improves (1/27/2026)
Nudify Tooling Gets Darker, Raising Immediate Harm Risks for Synthetic Video (1/26/2026)
False-ID Incident Raises the Verification Bar for Synthetic Video Coverage (1/24/2026)
California Deepfake and AI Transparency Laws Take Effect January 1 (1/1/2026)
Study Finds Over 20% of Videos Shown to New YouTube Users Are AI Slop (12/27/2025)
Three Key Copyright Decisions From 2025 Set the Stage for AI Video Litigation in 2026 (12/23/2025)
Pulitzer Winner and Journalists Sue OpenAI, Google, Meta, and Three Others Over Training Data (12/22/2025)
TikTok Drops New Community Guidelines With Stricter AI Content Labeling (12/4/2025)