The freedom to create is the freedom to make the invisible visible. Every object around you was once just a thought. The sacred act isn't the thinking, it's the making real. We are sculptors of reality itself🪄 #creativity

Multimodal is the future, again! 𝗧𝗵𝗲 𝗸𝗲𝘆 𝗶𝘀𝗻'𝘁 𝗷𝘂𝘀𝘁 𝗺𝘂𝗹𝘁𝗶𝗺𝗼𝗱𝗮𝗹𝗶𝘁𝘆; 𝗶𝘁'𝘀 𝗵𝗼𝘄 𝘄𝗲 𝗯𝗹𝗲𝗻𝗱 𝗺𝗼𝗱𝗮𝗹𝗶𝘁𝗶𝗲𝘀 𝘁𝗵𝗮𝘁 𝗺𝗮𝘁𝘁𝗲𝗿𝘀. Groundbreaking early work like Meta's ImageBind paved the way by unifying six different modalities into a

LLMs can write poetry, paint pictures, and spin up code. But ask them to follow rules, and they break. A legal clause that complies with regulation. A financial report that obeys accounting standards. A design that satisfies physical, functional, and aesthetic constraints. Even

𝗠𝗲𝗺𝗼𝗿𝘆 𝗶𝘀 𝘁𝗵𝗲 𝗻𝗲𝘅𝘁 𝘂𝗻𝗹𝗼𝗰𝗸. 𝗪𝗶𝘁𝗵𝗼𝘂𝘁 𝗶𝘁, 𝘆𝗼𝘂𝗿 𝗔𝗜 𝗵𝗮𝘀 𝗮𝗺𝗻𝗲𝘀𝗶𝗮! Every design system, whether for art, branding, or even mechanical depends on continuity, not just creativity. Style is memory. Brand is memory. Consistency is memory.

🎉 Our spotlight paper X-Edit: Detecting and Localizing Edits in AI-Altered Images (in collaboration with Valentina Bazyleva and Nicolò Bonettini @ Reality Defender) is being presented today at the XAI4CV Workshop @ CVPR 2025! We share a method to spot and pinpoint subtle

Clarity comes from action ‼️

tfw you know a paper's going to be good

Physicists and mathematicians really have the wriest humor.

Ikigai:

Nothing more satisfying than a sales call where you barely have to sell. Just had a discovery call turn into, “When can we get started?” 🚀 Early validation that we’re solving a real pain. #build #win #physical #design

Machines that build machines - von Neumann would smile. First prints 🚀 #hardware #design #physical #AI

“Everyone knows” what an autoencoder is… but there's an important complementary picture missing from most introductory material. In short: we emphasize how autoencoders are implemented—but not always what they represent (and some of the implications of that representation).🧵

Huge thanks to my friend and co-author Mashrur Morshed , my mentor Jacob Seidman, and my advisors Vishnu Boddeti and Gaurav Bharaj for their help and collaboration. Paper: arxiv.org/pdf/2509.15551 Webpage: sepehrdehdashtian.github.io/Papers/PolyJui…

Language models choke on CAD code. We’re fixing that. Autoregressive and diffusion language models don’t actually read design files. They break them into tokens, their own "vocabulary blocks." A single design can already blow past the context window, blocking larger designs and

Join our official a16z Tech Week after hours event! are you a CAD+AI person? developers, researchers and founders in AI+CAD, come grab a beer at the Romantic meetup in North Beach. Participation is subject to approval so we keep a tight, subject specific crowd and no