Physical AI Builders with YC, Oak, π

RSVP here: https://aiconference.com/#tickets 30% Discount Code: fregly30 The AI Conference 2026 is the premier destination for companies shaping the future of artificial intelligence. This is an in-person event in San Francisco, September 29–October 1, focused on applied AI. From research and infrastructure to production deployment and business strategy, the AI Conference brings together 5,500+ builders, researchers, founders, investors, and executives across five tracks, with 120+ speakers, hands-on workshops, Hack Day, Startup Showdown, and an Innovation Hub. Learn more: https://aiconference.com/about/ RSVP here: https://aiconference.com/#tickets Discount Code for 30% off: fregly30

Join us for a relaxed social run, followed by time to hang out over a beverage after. All paces are welcome! Note: we do not have a place to store bags, so please plan accordingly!

Join Foresight Institute and Convergent Research on a salon on FRO's. We'll be talking in more detail about: Oath Oath is building tool for overseeing and controlling advanced AIs using formal methods. Future AIs will operate at scales and speeds that mean we can't oversee them directly. Oath's aim is to scale human oversight, with trustworthy mathematical evidence that rules are obeyed. We're starting with powerful reasoning tools such as Lean, and developing ways for humans to wield them that is understandable and robust to AI deception. Our aim is to help humans work safely with AIs in consequential situations. Atlas Ignota Atlas Ignota (previously Atlas Computing) identifies the gaps AI is creating in society’s defenses and turns the most promising opportunities into action. Through field strategy, Atlas maps attacker–defender dynamics, identifies interventions that could shift the balance toward defense, and builds the teams, organizations, and funding pathways needed to execute. Working across epistemic security, cybersecurity, model-weight security, and critical infrastructure, Atlas operates upstream of traditional incubation to create solutions where they are most needed. 5.30 PM: Doors open, food & drinks available 6.30 PM: Fireside chat 7.00 PM: Audience Q&A 7.15 PM: Mingle & Hang 8.30 PM: Event ends Dress code, casual. +1s are very welcome - just make sure they register here so we can let them in. This salon is hosted by Foresight Institute at the Node.

What if software was something we could mold and reshape around the lives we want to live, instead of settling for the addictive defaults sold by big tech? We want to take back that agency — tech should serve us, not the other way around. During this ‘Art of Being Human’, we will explore: What would it mean to rekindle the dream of truly personal software, and what would we create if we could? Event Details: 5:30–6:00pm — Doors open and close firmly at 6pm 6:00–6:30pm — Starting conversation of what's changing in technology, and what the future of personal software may look like 6:30–8:30pm — Breakout into small groups and get hands-on designing a new personal computer for you 8:30–9:00pm — Group closing A note on event format and what to expect: This is a participatory gathering. You’ll spend time in a small group reflecting on the quirks, habits, systems, and preferences that make the way you live and work uniquely yours—and imagining what your digital life could look like if you could shape it around you. Expect prompts, conversation, hands-on making, and some playful reimagining of what a personal computer could look like when you’re the one shaping it. Come ready to play with ideas, make something a little unexpected, and leave with a few possibilities you may not have considered before. — Imbue is an AI company building tools to empower humans in the AI age. We're building toward an open agent ecosystem where you can understand, modify, and verify agents — so they only work in your interest. Learn about our work and what we stand for: www.imbue.com

Anthropic × Cerebras: Cafe Compute Cafe Compute is a casual evening for people who are building on Claude and want faster inference to transform their projects. Grab a cocktail, connect with fellow developers, founders, researchers, and creators, and trade ideas about what you’re building - or what you can’t stop thinking about building. *** Cerebras is the world's fastest AI inference, up to 15x faster than leading GPUs. Cerebras Inference is powered by our Wafer-Scale Engine (WSE-3) - the world's largest AI chip. OpenAI's GPT 5.6 Sol Ultrafast is powered on Cerebras, you can try out one of the open source models like Qwen 3.8 27B. Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems. Join our rapidly growing communities: Discord ↗ Reddit ↗ X ↗

The VAST Builders Challenge is a three-city technical event series bringing together founders and engineers across New York, London, and San Francisco to build production-ready AI systems. Spend the day building agents to unlock the value of video. Because video is hard to search and analyze, most of it stays in storage, unwatched and underused. Experiment with the latest AI infrastructure, collaborate with other ambitious builders, and turn your ideas into shippable prototypes - with access to expert guidance, compute, and the tools you need to ship. Hosted with VAST Data in collaboration with NVIDIA, SpaceXAI, CoreWeave and Weights & Biases. Note: Registration closes on September 29th. Prizes 1st place: NVIDIA DGX Spark™ HuggingFace Microduck Cursor Credits Cash Gift Cards The Builders Stack Here's what you get out of the box, a lightning-quick setup: Video understanding: reason over a video clip with NVIDIA Cosmos. Semantic search: find moments across hours of footage using natural language. Object detection: real-time detection and tracking across frames with YOLO. General-purpose LLMs: models from Weights & Biases to power your application logic (e.g. video agents). The VAST AI OS orchestrates the video ingestion and data layer, models are served on CoreWeave GPUs, and builders ship with Cursor. Build your own video agent or something no one’s thought of. We'll share the videos to get you started on building video agents. Judges Hassan Moustafa - TME, Multimodal AI @ NVIDIA Adam Ryason, Ph.D. - Product @ NVIDIA Ram B. - Sr. Developer Advocate @ VAST Data Brian Verkley - Director, AI Data Platform @ VAST Data Anushrav V. - Solutions Architect, Physical AI Lead @ CoreWeave Arnav Verma - Field Engineer @ SpaceXAI Who should attend Developers, ML/AI engineers, agent builders, and data engineers, from startups to enterprise teams working with video or multimodal AI. While individuals are welcome, we encourage teams of up to four members and will prioritize team