Hey everyone — Jacky here, just joined from Switzerland. I build AI-powered SaaS tools (image generation, video pipelines, brand intelligence). One thing that surprised me shipping AI to real users: the model matters way less than people think. We tested Gemini, Kling, Flux, Stable Diffusion — the biggest quality jumps always came from improving the workflow around the model, not from switching models. A great process with an average model beats a messy process with the best model. Every single time. That's exactly why this community's focus on "real & repeatable results" resonated with me. The repeatable part is where the magic actually lives. Excited to share what I've learned building these tools (especially the failures — those teach faster) and to learn from how this group approaches AI. Genuine question for you all: do you use AI mainly to speed up what you already do — or to create entirely new things you never could have done before?

In this video, I go over the various updates and releases from Google and Anthropic, discusses the upcoming AI hardware releases from Apple and OpenAI, tests out a frankly creepy demo of a live interactive AI avatar, and more. Enjoy!