everyday i wake up and think about how much i appreciate black boxes and the magic they let me wield

this seems...relevant. excited to try it out, and congrats to the team on launching this!!

Launching tomorrow, a new way to edit videos automatically with AI…

man, function calling with llms is so hot. and pydantic goes so incredibly hard.

gpu rich, cpu poor 🧐

the video editing landscape is such an interesting one. literally every single player now has AI top of roadmap each with there own mini-take on what's important and how to do things.

this wasn't the case at the beginning of last year.

dubbing is such a fun use case to do with ai. its such a clear example of ai models seeming more like building blocks versus the end experience / product.

youtube served 100,000 videos/day for just $4k/mo

ByteDance does $135B/yr in revenue (more than Meta). More small startups will eat their lunch because they can hit scale with fewer and few engineers (even in video which is complicated), and have plush ground to innovate on with AI.

diglett ai

which LLM API provider has got function calling enabled on this bad boy

you know facebook really has their content targeting on lock when ur cofounder sends u an instagram short like this

poor Sieve getting abused

the best part about work is that i get to watch tron legacy and speed racer instead of doing work

the quality of VLMs is getting shockingly great. super excited about Reka and a big fan of Reka's younger sibling, CogVLM which we work with quite a bit Sieve.

Really excited to unveil our first case study! It’s been amazing working with the Zight (fka CloudApp) team, especially Sachin and Séraphin Hochart, and really awesome to see them tell the story of using Sieve for Zight’s AI capabilities.