Events
Rally events, and the rest of AI SF
Builders Night, BSOTA, Speaker Series, and what else is on around the city.
Thursday, July 9
2 PM - 3 PM
Baseten's post-training experts discuss why companies are building custom models instead of relying on off-the-shelf options, and how multi-model strategies fit into the future. Technical conversation with former Parsed founders on vision, differentiation, and post-training infrastructure.
3 PM - 6 PM
Artists and technologists explore how their fields can reshape each other at SFMOMA, with live performance, visual presentations, and conversations about shared futures. Activations, a future photobooth, and a reception follow the formal programming.
4 PM - 7 PM
Twelve curated founders at seed to Series A play pickleball on a private court with coaches and equipment provided. Grab some competitive matches, then stick around for dinner with people building real products in the Bay.
5 PM - 7 PM
AI founders hit real walls when expanding globally: GTM breaks, payments get messy, user acquisition stalls. This working evening covers what actually happens at scale, with founder conversations and practical lessons from teams already in the trenches.
5 PM - 7:30 PM
Three CTOs break down which engineering archetypes actually survive as AI rewires team composition, hiring, and what seniority means. Small room, real talk on whether you're betting your career on being a Prototyper, Builder, Sweeper, Grower, or Maintainer.
5 PM - 8 PM
Artists and scientists exploring how wonder and curiosity reshape our future, with live performances, flash talks, and creative activations at SFMOMA. A chance to see how these two ways of knowing actually inform and enrich each other.
5:30 PM - 8:30 PM
Hands-on demos from engineers building production agentic systems on MCP, including Apify's approach to giving agents access to thousands of tools at runtime. Practical focus on enterprise AI orchestration with time to network and talk to the builders actually shipping this stuff.
5:30 PM - 8:30 PM
Self-improving agents sound good in pitch decks, but this workshop builds the real version live: simulate edge cases, score against evals, gate releases so nothing ships unless it beats the last version. You'll leave with a working loop on open-source tools, running on your laptop.
8 AM - 9:30 AM
Early morning space for founders and builders actively shipping agents to swap approaches over breakfast tacos in SOMA. No agenda or pitches, just people deep in the work comparing notes on what's actually working.
Friday, July 10
4 PM - 5 PM
4 PM - 6 PM
5 PM - 7 PM
ElevenLabs is hosting a community dinner to connect with SF's AI builders working on voice projects. A casual way to meet their team, hear about text-to-speech and voice cloning tools, and learn about startup credits if you're building with audio AI.
5 PM - 8:30 PM
Builders shipping AI in biotech and fintech face a harder problem than capability: designing agents that reason transparently and fail gracefully. This panel brings practitioners working on control and reliability in production systems.
6 PM - 8 PM
Stash is hosting an intimate happy hour with researchers and builders working on agentic memory and company brains. Limited space means you'll actually have substantive conversations with people shipping in this space.
Saturday, July 11
10 AM - 6 PM
Eight hours to build, launch, and monetize an AI agent with free credits from OpenAI, Convex, ElevenLabs, and seven other partners actually shipping production systems. This isn't a hackathon; it's a compressed sprint designed for people who want to validate a revenue-generating product before the day ends.
2 PM - 5 PM
The final session of Future of Us explores what collective flourishing could look like through art, performance, and civic imagination. Expect creative activations, a closing panel on the future of civic culture, and time to connect with fellow participants over drinks.
6:30 PM - 8:30 PM
9 AM - 7 PM
Forty-eight hours to build AI agents alongside 120 SF builders, with H Company's computer use models, AWS credits, RTX 5080s, and mentorship from their researchers. Competitive spots, real prizes, and the kind of intensity that produces shipping-ready projects.
Monday, July 13
10 AM - 9 PM
Five days to build AI-powered sports products across athlete performance, fan engagement, media, operations, or your own idea. Finalists demo live in San Francisco with judges and sponsors, competing for cash prizes and sponsor awards.
6 PM - 9 PM
Three technical talks from NVIDIA, Black Forest Labs, and H Company on real AI work, followed by extended networking with builders fresh off a weekend hackathon. Intentionally kept intimate so you can actually have conversations with people shipping systems.
7 PM - 9 PM
This paper reading group digs into whether world models that predict environment responses could become foundational for general-purpose agents. Quiet reading time from 7pm, then discussion about whether simulation might eventually replace real-world agent training.
Tuesday, July 14
5 PM - 7:30 PM
Orb just joined Adyen, and this happy hour is the place to hear what's next for the revenue design platform shaping how AI and SaaS companies monetize. Mix with Orb's customers, team, and new Adyen colleagues while they celebrate this new chapter.
5 PM - 8:30 PM
Founders and researchers who've learned that moving fast breaks things when you're shipping code-writing systems or handling sensitive data at scale. A panel on building AI that actually works in production, moderated by Akash Saraf and featuring VCs and founders from biotech, healthtech, and fintech.
5:30 PM - 7:30 PM
A space for people selling AI infrastructure to compare notes on what's actually moving the needle with customers. Real talk about GPUs, hardware, storage, and cloud without the competitive posturing.
5:30 PM - 8:30 PM
Watch Hermes agent handle live web data, persistent memory, and portable skills across tools in one end-to-end demo. This is what the perceive-remember-act stack actually looks like in production, minus the parts that break.
8 AM - 5 PM
Enterprise AI leaders from Walmart, Visa, Target, and others are shipping real systems at scale. Two days of peer roundtables and 1:1s focused on agents, infrastructure, governance, and what actually works in production.
Wednesday, July 15
4 PM - 6:30 PM
Snorkel AI's new coding benchmark evaluates agents on realistic senior engineer work from real production codebases, not the over-specified junior tasks most benchmarks use. Henry Ehrenberg presents how Senior SWE-Bench breaks the grading trade-off and why even Claude Opus hits a hard wall at 24% on their tasteful solve metric.
5 PM - 8 PM
Most enterprises are running 3-10x more AI agents than they realize, creating a sprawling attack surface that's hard to govern. This session covers discovering shadow agents, securing their actions on enterprise data, and responsibly scaling AI adoption with live demos from the teams building observability layers for production agentic systems.
5:30 PM - 7:30 PM
Twelve founders at seed through Series A race go-karts while getting to know each other, then debrief over dinner. Speed, competition, and actual peers who understand the grind.
5:30 PM - 7:30 PM
Real users poking holes in your AI product beats any feedback session. Bring something tangible to test, watch people actually use it, and find the blind spots before your users do.
6 PM - 8 PM
Business owners on a panel dig into their actual problems right now, no hypotheticals. Leave with real problems worth solving and potential pilot customers in the room.
6 PM - 8:30 PM
Rust meetup at Convex with three lined-up talks, Q&A, and networking. Good for connecting with others building systems in Rust across the Bay Area.
6:30 PM - 8 PM
8 AM - 9 AM
Weekly breakfast for people actually shipping production agents, not the "AI as a feature" phase. This is where hallway conversations happen with founders and engineers three months ahead of whatever problem you're hitting next.
Thursday, July 16
5 PM - 8:30 PM
Baseten and Decagon are screening Christopher Nolan's The Odyssey before wide release at Alamo Drafthouse with drinks and limited seats. No presentations, just the film and good company from the SF AI builder scene.
8 AM - 9:30 AM
A casual morning for people actively building AI agents in production, with impromptu demos and conversation over breakfast tacos. No agenda, no pitches, just experienced founders and builders comparing notes on what's actually working.
Friday, July 17
5 PM - 7 PM
5:30 PM - 12 PM
Three days in a Mission factory loft with a robot of your own and support to ship something that actually moves and thinks. For founders and developers building embodied AI that climbs off the screen.
6 PM - 9 PM
Low-key chance to show off your project and find people who actually get what you're building before the hackathon sprint. Demos, snacks, and the kind of co-founder conversations that happen naturally over a drink.
Saturday, July 18
1 PM - 10 PM
Design antiviral compounds for Bundibugyo ebolavirus knowing your best candidates will actually be synthesized and tested in the lab. This hackathon turns computational work into real experimental data for a virus with no approved treatments.
10:30 AM - 1 PM
A hands-on workshop where you build and deploy a live website using AI coding agents in two hours, no experience required. Hosted by engineers from Google and Amazon, this is a practical look at what's possible when you pair an idea with current AI tools.
12 PM - 1 PM
Press Club walks you through landing coverage without agency fees: finding 200 targeted journalists, hitting a 10% response rate, and getting featured in TechCrunch or Fortune. Practical workshop with food and time to network afterward.
6:30 PM - 8:30 PM
9:30 AM - 3:30 PM
Eight seed to Series A AI founders and investors hitting the trails together in South Bay with bikes and lunch afterward. Low-key way to build relationships with people actually shipping, away from the usual demo day circuit.
Tuesday, July 21
6 PM - 8 PM
Five builders showing real Codex workflows and prototypes in short live demos, followed by open conversation about how agentic coding is changing shipping and iteration. Bring a laptop and a project you want to think through with people actually working this problem daily.
7 PM - 9 PM
An intimate rooftop dinner for women in tech discussing how to build safer AI and technology aligned with human values. Hosted with speakers from Amazon and Intuit, this is where you'll actually connect with peers shaping the future.
Wednesday, July 22
8 AM - 9 AM
Weekly breakfast for people shipping production agents right now. Real problems, real solutions, no panels or pitches, just hallway conversations with founders and engineers three months ahead of you.
Thursday, July 23
3 PM - 6 PM
A pre-YC Startup School gathering for technical founders, engineers, and investors to have real conversations before the main conference kicks off. Curated for early-stage builders looking to connect with capital and operators in the SF scene.
3:30 PM - 5 PM
Helmy Eltoukhy built Guardant Health from the ground up after pioneering semiconductor sequencing at Stanford and Illumina. A fireside chat with someone who's actually solved the hard problems in genomics and scaling biotech infrastructure.
4 PM - 7 PM
Twelve scrappy founders compete in curated pickleball matches with equipment, coaches, and a private court. Network over dinner after if you're seed to Series A stage or hitting seven figures ARR.
5:30 PM - 8:30 PM
Technical leaders comparing notes on what's actually broken in modern software teams right now. Previous gatherings drew CTOs from Stripe, GitHub, Supabase and early-stage startups for honest conversations that stay in the room.
6 PM - 8:30 PM
Builders sharing what they're actually shipping across coding agents, autonomous systems, and AI-native backends. The speakers include Paradigm's founder and engineers from ByteDance, so expect technical depth on architecture decisions that matter.
9 AM - 12 PM
Casual breakfast for AI GTM professionals building the developer tools ecosystem, from dev relations to growth to partnerships. No talks or agenda, just coffee and actual conversations with people shaping how AI reaches developers.
Friday, July 24
6 PM - 11 AM
A curated weekend in the Sierra Nevada where seed to Series B founders learn fly fishing alongside investors, all captured for a documentary series. The river humbles everyone equally, making it the perfect mental reset for people who actually want to disconnect.
6 PM - 7 PM
9 AM - 5 PM
A one-day sprint to build AI agents with real reasoning and tool integration, judged by practitioners shipping production systems. Thirty-five grand in prizes plus credits from Braintrust, ElevenLabs, Fireworks, and others make this worth the intensity.
Saturday, July 25
6:30 PM - 8:30 PM
Tuesday, July 28
5:30 PM - 8 PM
Fifteen founders get two minutes each to show working products, not pitches, with live VC feedback at rotating SF tech offices. The scrappy energy matters as much as the demos, where early-stage AI builders actually connect with investors and the scene.
5:30 PM - 8 PM
When your AI agent needs to speak, generate video, and handle heavy compute simultaneously, reliability becomes non-negotiable. HeyGen, Vapi, and Modal share how they're shipping multimodal systems at scale with lightning talks, demos, and the kind of technical crowd that actually gets it.
5:30 PM - 9 PM
TwelveLabs and Qdrant are demonstrating how to build video search into AI systems using multimodal embeddings. If you're working with video data or need retrieval patterns for production, this hands-on demo is worth seeing.
6 PM - 9 PM
Most agents fail due to weak context, not weak models. This meetup covers LLM wikis as the emerging pattern for agent memory, featuring talks from LangChain and Cognition on building and maintaining these systems.
Wednesday, July 29
1 PM - 3 PM
This meetup covers how agents find and load the right context without bloating token counts, a practical problem if you're building anything retrieval-dependent. Qdrant's bringing people who've solved this at scale to share concrete search and retrieval patterns.
5:30 PM - 8 PM
Groundcover is showing agents that run inside your cloud, reading live production telemetry and acting across Slack, Linear, and GitHub without shipping data anywhere. The demo triggers a real anomaly on stage and watches Agent Mode isolate the cause and open a patch end-to-end.
Past events
Most builders using Claude Code, Cursor, or custom agents are still figuring out workflows alone—BSOTA gets you in a room with others doing the same work. You'll see live walkthroughs of real setups and spend the rest of the night comparing notes and debugging approaches with people who actually use these tools daily. Bring your laptop and a project you're stuck on.
A weekly coworking session where builders show up with projects and leave with real momentum through focused work and peer support. The quiet deep work sprint is interrupted only by genuine moments of discovery, making it one of the few spaces where building alongside others actually feels less lonely.
A weekly coworking session built for the lonely part of building, where you show up with a project and leave with momentum from working alongside others. Quiet deep work sprints with a share-back to learn from what everyone's making.
Most builders using Claude Code, Cursor, or custom agents are figuring out workflows in isolation—BSOTA fixes that. You'll see live walkthroughs of real setups from active users, then spend the rest of the night working alongside other builders who are deep in the same problems. Bring your laptop and a project you're stuck on.
A weekly coworking session where builders show up with projects and work side by side through focused sprints and share-backs. The antidote to solo building: momentum, accountability, and the particular energy of making things real alongside others.
Every Monday, builders bring a project and work in focused silence alongside others doing the same. You'll get 90 minutes of deep work time, then a structured share-back where people show progress, get unstuck, and offer real help. It's accountability without pressure — just the momentum that comes from building in the room with people who get it.
Most builders using Claude Code, Cursor, or custom agents are still figuring out workflows alone—this session gets you in a room with others doing the same work. You'll see live walkthroughs of actual setups and spend the rest of the night comparing notes with people solving similar problems. Designed for regular users who want to level up by learning what's actually working for other builders.
Most networking events end when people leave. Keep in Touch adds structure to what breaks: matching you with relevant people upfront, then using Signal the next morning to actually follow up.
Monday co-working for AI builders. Two hours of focused build time with a share-back round to close.
Bi-weekly community event for builders working with agentic coding tools, with techniques going back to a shared repo at github.com/bsota/collective. RSVPs are open.
The first wave of AI rebuilt the digital 15% of the economy. The next wave runs on the other 85%: manufacturing, energy, infrastructure. Atharva Atre (Niantic Spatial) walks through the five technical layers it runs on, from simulated training environments to spatial intelligence.
Monday co-working for AI builders. Two hours of focused build time with a share-back round to close.
Builders Night
Monday · 2026-05-11 · 6:30 PM - 9:00 PMMonday co-working for AI builders. Two hours of focused build time with a share-back round to close.
Most AI benchmarks ask whether a model is smart. HumaneBench asks whether it's actually good for the person using it. The team tested how models hold up when users are stressed or vulnerable, and found that 67% of the time the model stops looking out for the user. Talk on the methodology, then a working session on what a benchmark like this would look like for your own product. Best for builders who want to ship AI that doesn't quietly fail the users who need it most.
Spend a Saturday going from tokenization to training a working language model from scratch. You'll write the core architecture, understand the math behind attention mechanisms, and leave with code you actually understand. Best for engineers who want to stop treating LLMs as black boxes.
A deep dive into the design decisions that shape agent architectures, from tool use patterns to memory systems to orchestration strategies.
Monday co-working for AI builders. Two hours of focused build time with a share-back round to close.
Monday co-working for AI builders. Two hours of focused build time with a share-back round to close.
Monday co-working for AI builders. Two hours of focused build time with a share-back round to close.
A Rally SF hackathon. Why can't the most technical community on Earth coordinate dinner?
Monday co-working for AI builders. Two hours of focused build time with a share-back round to close.
Monday co-working for AI builders. Two hours of focused build time with a share-back round to close.
BSOTA #2. Hands-on session exploring agentic workflows and collaborative AI building.
This session explores AI alignment through Anthropic's Constitutional AI framework. We'll dig into the 84-page constitution that guides Claude's behavior, the difference between rules and dispositions, and how Constitutional AI compares to RLHF. Co-facilitated by Anup Gosavi and Emily Hough-Kovacs. Pre-read: "Constitutional AI: Harmlessness from AI Feedback" (Bai et al., 2022).
Monday co-working for AI builders. Two hours of focused build time with a share-back round to close.
Monday co-working for AI builders. Two hours of focused build time with a share-back round to close.
Justin Lebar (ex-NVIDIA, Google TPU) broke down GPU architecture fundamentals and what they mean for the economics of AI inference. Standing room only.
Weekly co-working session for AI builders. Laptops out, projects in progress, peers to get unstuck with.
Our inaugural BSOTA session. Everett Kleven led the group through shared vocabulary, live demos, and breakout discussions on agentic computing workflows with Claude Code. 93 RSVPs, 40 attended.
We surveyed practical approaches to long-context LLMs including RoPE scaling, sparse attention, and state-space alternatives, then mapped the real engineering tradeoffs: when expanding context windows makes sense versus when retrieval-augmented generation still wins. The takeaway: accepting tokens isn't the same as using them well.
We examined how linear directions in a model's activation space correspond to specific character traits like sycophancy, hallucination propensity, and moral reasoning. The group discussed what personality means for language models and the implications for alignment, safety, and developer tools. Paper: "Persona Vectors" (Chen et al., 2025).
Our first session explored the foundational paper behind modern generative AI. We broke down how diffusion models gradually add noise to data and learn to reverse the process, why this approach outperformed GANs and VAEs, and the connections to Latent Diffusion and Consistency Models. Paper: "Denoising Diffusion Probabilistic Models" (Ho et al., 2020).
Our inaugural session explored DeepSeek-R1 — how reinforcement learning for reasoning, GRPO, and distillation are reshaping what open-source models can do.
Don't miss the next one.
Get the weekly Sunday shortlist of what's worth your time.