Tony Dinh: a $1,000 video in 30 minutes
Tony Dinh says a video he paid over $1,000 for a year ago took him under 30 minutes with Opus 5.5, and Claude picked the brand assets on its own. Meanwhile Matt Shumer called local models a nerd trap and got an argument.
Real money
Greg Isenberg says the average human now makes more mistakes than frontier agents
Greg Isenberg says we have reached the point where the average human makes more mistakes than the average frontier model agent. He stops short of calling it AGI, and posts no error rates, no task list and no comparison behind the claim.
I don't know if it's AGI but I think we've gotten to the point where the average human make more mistakes than the average frontier model AI agent
Tony Dinh says a video that cost him $1,000 last year took 30 minutes
Tony Dinh says he made a motion graphics video with Opus 5.5 in under 30 minutes that he paid roughly $1,000 or more for a year ago. He published the full prompt, a 15 second showreel-style brief asking for real product screenshots, logo and music-matched motion. He says Claude gathered the brand assets itself with no direction from him.
Claude gather the brand assets by itself btw, I gave it no direction
Founder fights
Matt Shumer calls local models "a nerd trap", builders push back
Matt Shumer posted that local models are useless and doubled down by calling them a nerd trap. Sam Wasserman asked whether he had run GLM 5.3 locally; Shumer's reply was to ask whether Wasserman had used Opus 5.5. Nobody in the exchange posted a benchmark, a cost figure or a task either model failed.
If local models are useless, why are developers still building them?