Sashv Dave

When a computing environment is generally capable enough as AI is, it's not new software capabilities that make it tailored to your needs, but the data. Every company and every person will one day have a magical library that personalizes their computing experience. We're building Pacific to pull that future forward.

introducing chipmunk—a training-free algorithm making ai video generation 3.7x & image gen 1.6x faster! ⚡️ our kernels for column-sparse attention are 9.3x faster than FlashAttention-3 and column-sparse GEMM is 2.5x faster vs. cuBLAS a thread on the GPU kernel optimizations 🧵

I taught an LLM to optimize proteins. It proposed a better carbon capture enzyme. Introducing Pro-1, an 8b param reasoning model trained using GRPO towards a physics based reward function for protein stability. It takes in a protein sequence + text description + previous experimental results, reasons over the information given in natural language, and proposes modifications to improve the stability of the given sequence. 🧵(1/n)

Introducing Wildcard Bridge & agents.json An open source spec connecting AI agents with APIs — built on the OpenAPI standard.

Today, we're thrilled to announce our $3.3M seed round at Delve. AI makes millions of decisions every second. But it’s still regulated with spreadsheets and screenshots. When AI moves at the speed of thought, protecting it with yesterday's tools isn't just inefficient, it's dangerous. We built Delve to help secure the future of AI. Our AI agents turn hours of manual clicks and screenshots into seamless automation and are already securing 100+ of the fastest-growing AI companies. Welcome to the modern age of compliance.

We're thrilled to be launching Tilde. We're applying interpretability to unlock deep reasoning and control of models, enabling the next generation of human-AI interaction. By understanding a model's inner mechanisms, we can enhance both its reliability and performance—going beyond what's possible with traditional techniques like fine-tuning alone. Check out what Tilde can do.

Meet Sohu, the fastest AI chip of all time. With over 500,000 tokens per second running Llama 70B, Sohu lets you build products that are impossible on GPUs. One 8xSohu server replaces 160 H100s. Sohu is the first specialized chip (ASIC) for transformer models. By specializing, we get way more performance: Sohu can’t run CNNs, LSTMs, SSMs, or any other AI models. Today, every major AI product (ChatGPT, Claude, Gemini, Sora) is powered by transformers. Within a few years, every large AI model will run on custom chips. Here’s why specialized chips are inevitable: