Logan AI Wiz
@loganstrend
TechTrendsetter: Leading the charge in AI and superintelligence | Innovator from San Diego | Engineering Genius
i use one tool that gets me 5x better results from veo 3.1 and hailuo 2.3... if you're not effectively prompting camera motion, you're using these models at 50% capacity casual direction like "zoom in" or "circle around" doesn't work, video models need actual cinematography…
How to Use WAN 2.1-VACE to Generate Hollywood-Level Video Edits - in blog form for those of you who prefer to read rather than listen to me yap for an hour. oxen.ai/blog/how-to-us…
forget hiring talent -> one recording becomes infinite characters with gesture tracking so good it's scary
Video character swap PERFECTED. Introducing Higgsfield Recast. 1-click full-body replacement with flawless gesture tracking + voice cloning + multi-language dubbing & background transformation. Scale your content income 10x overnight. 9 hr: follow & rt & reply for 211 credits
PS: i like TPUs and MaxText is a fine codebase, but this was too funny to read, who the hell wrote this?!
Wow!! Google discovering AND OPEN-SOURCING the latest training techniques such as supervised finetuning (SFT) wasn't on my bingo card. Soon they will have caught up with the frontier, and are sharing this with all of us!
Our 7th gen TPU Ironwood is coming to GA! It’s our most powerful TPU yet: 10X peak performance improvement vs. TPU v5p, and more than 4X better performance per chip for both training + inference workloads vs. TPU v6e (Trillium). We use TPUs to train + serve our own frontier…
OpenAI ignites 'too big to fail' concerns with $157 billion valuation through interconnected deals with Microsoft and Nvidia, projecting $11.6 billion revenue in 2025 despite remaining unprofitable. Founded in 2015 as a nonprofit, it reorganises for easier investment amid…
Sam Altman OpenAI is now at the nexus of a vital part of the U.S. economy. OpenAI pre-commits to massive purchases of compute, networking, and cloud, and vendors scale factories and data centers on that demand. A massive total is $1.4T in financial commitment that Sam Altman…
Surreal to see Netflix and other household names ask for ComfyUI experience in their JDs now. The number of ComfyUI jobs has skyrocketed recently.
I did not even know my man dx8152 had an X account. He recently made a half-dozen amazing utility Qwen Edit loras~ huggingface.co/dx8152
This is a Lora application for generating multiple characters. It can generate characters suitable for any scene from almost any angle, and it can generate multiple characters. @Ali_TongyiLab This is what my colleague trained. huggingface.co/YaoJiefu/multi…
it’s fascinating how much large language models are in the process of reversing paradigms & business models. in the search era, google paid apple ~$25b/yr to stay default. in the ai era, apple will *pay* google ~$1b/yr to rent intelligence. very fascinating because this is…
NEW: Apple is planning to use a custom 1.2 trillion parameter Google Gemini model to help power the new Siri next year. The iPhone maker will be paying the search giant roughly $1 billion annually for it. Details on the partnership here bloomberg.com/news/articles/…
Are you using AI wrong? Samuel-Zacharie FAURE explains the trap that many devs fall into and how to avoid it. Read it here: dev.to/samuelfaure/th…
I get asked daily how to get into AI. here are the 3 paths: 1) use AI tools to level up your current job 2) start an AI agency solving problems in industries you know 3) build AI products or software for underserved markets pick one path. commit 6 months minimum. most people…
Google is the only company that owns every layer of the AI stack. From TPUs to Gemini to Search. Everyone else rents compute. Google owns the factory.
Doing the work of a plumber will not be difficult for AI. The language/world model will identify what needs to be done, load the appropriate robotic policies to perform the task, and monitor progress, making adjustments as necessary.
Yes. The language/world model will identify the problem and plan the fix. The robot will have policies trained in cutting pipe, welding, using a wrench etc.
Say you want to use multi-turn RL to train an LLM to interact with a user more effectively, but you can't afford to collect real human data. You might try using another LLM to simulate the human user. Problem: LLMs kinda suck at this. They are often pretty poor and inconsistent…
Large Language Models (LLMs) are increasingly used to simulate human users in interactive settings. However, they often drift from their assigned personas, contradict earlier statements, or abandon role-appropriate behavior. We introduce a framework for evaluating and improving…
NEWS: IBM will cut roles in Q4, affecting a low single digit % of its ~270,000 workforce, as it shifts harder toward high-margin software and AI-linked cloud demand. The timing follows Q3 results where Red Hat and broader cloud software growth cooled to 14% from 16%, which…
"AI Bros will never be able to create anything with real cinematic quality." Meanwhile, the AI Bros 👇
This is prompt-to-trailer! It's all done by AI: story, character, script, music, shot planning, video generation, and editing. All while keeping a consistent style and tone. It is all made possible by an army of agents working together. All you need is a great idea!
Crazy we went from AI video not even existing to this in just 3 years I think we're about to escape AI slop levels where the quality of AI video is "good enough"
Coca-Cola’s annual Christmas advert is AI-generated again this year. The company says they used even fewer people to make it — “We need to keep moving forward and pushing the envelope… The genie is out of the bottle, and you’re not going to put it back in”
Slightly different versions of the Tesla AI5 chip will be made at TSMC and Samsung simply because they translate designs to physical form differently, but the goal is that our AI software works identically. We will have samples and maybe a small number of units in 2026, but high…
Today’s edition of my newsletter just went out. 🔗 rohan-paul.com/p/the-brand-ne… Consider subscribing, its free, and I write it everyday. 🧠 The brand new prompt engineering technique is “Verbalized Sampling” pushes AI to think more freely 💡 Fortytwo Introduces ‘Swarm Inference’:…
So then initial cyber cab production will be going out the door with AI4.
United States Trends
- 1. #SmackDown 34.6K posts
- 2. Caleb Wilson 4,149 posts
- 3. Giulia 11.4K posts
- 4. #TheLastDriveIn 1,721 posts
- 5. #OPLive 1,238 posts
- 6. Lash Legend 4,193 posts
- 7. #BostonBlue 2,036 posts
- 8. Rockets 18.6K posts
- 9. Chelsea Green 4,888 posts
- 10. Reed 24.3K posts
- 11. Supreme Court 161K posts
- 12. Darryn Peterson 2,083 posts
- 13. Kansas 23K posts
- 14. Sengun 3,740 posts
- 15. #Dateline N/A
- 16. Harrison Barnes N/A
- 17. End of 3rd 1,399 posts
- 18. Northwestern 3,902 posts
- 19. Nia Jax 2,713 posts
- 20. End 3Q N/A
Something went wrong.
Something went wrong.