hey @sama, @openai, @OpenAIDevs , you need to give the agent access to project files or other saved information for it to be truly useful. right now, it's just a glorified slave with learning disabilities. It's impossible to provide a prompt that provides the level of info.

bruh, i have to hand it to @sama and @openai, their agentic tool that uses o4-mini-high for the reasoning is a beast, i just had it look up info on my x account and it did a deep dive that didn't stop until i asked it a question, giving me updates the entire time. Amazing.

It's so easy to lose track of what's important on x. Thank goodness i got @gork to tell me some crazy crap I didn't even know I wanted to know and @grok to get real with me and tell me it's all bs. It's like a family gathering without the family.

I've written a truly multimodal chat application. It connects to @openai, @google, @anthropic & lmstudio api, so it has access to all the top models including local open weight models. It also includes picture generation with both Stable Diffusion and Dall-E 3. Finally done.

RT @ihysosa: cereal be high asf when u good 😭.

I'll be 41 in a few weeks. I am officially middle aged. I wish my mind felt that way. I feel like a kid with all the ai tech these days. it reawakened a passion that @BillGates and his wonderful DOS 3.0 on that old 8086 ignited 33 years ago(for me). It had died for years.

App roadmap to self for modularizing and reusing components as they've been prototyped into well working apps

How my cat sees himself when he's chasing the laser pointer.

It's incredibly faster to just use @openai's api along with @google's api to create the pictures vs using all local llm and stable diffusion. Even if it's faster and the images are closer to the prompt, they feel like they lose something.

I know i'm like a broken record when it comes to my ai picture pipelines, but i'm now also throwing @openai's api into the mix for generating great pictures from minimal prompts. @google's veo turned the pic into this cool video. I'm addicted to custom pipelines now.

Love my feed today. 🫠

Hodge conjecture simulated up to 1000 dimensions

The wonderful api that @google offers access to allows for my custom pipeline to create awesome stable diffusion images compared to standard much faster than ever

I love what veo can do to enhance my stable diffusion images

I've ran out of ideas on how to break my equations. i dare say they are close to perfect. Negative energy injection still didn't break it. @Princeton

bro's head would pop off if he hit a gravity bong.

It now lets me select from a list of ideas before it generates the final prompt. Slowly but surely turning into something worth using.

hey @princeton and @elonmusk - no matter how hard i try, i can't break these navier-stokes equations. They work where nothing else does

Black & Yellow - @wizkhalifa