From mid-September to late October, my calendar was full of AI events across the Bay Area and Los Angeles. I hosted, co-hosted, spoke at and supported many AI events in San Francisco and LA. Almost every week, I was on a stage, at a booth or at a hackathon. This post is a recap of those weeks: what each event was about, what I shared on stage, and what I learned from running so many developer events in a short time.
I lead Developer Relations (DevRel) at GMI Cloud, which means I help developers learn and build with our AI cloud. I also run Bayhaus Creative, a video production studio in San Francisco. Both teams worked on these events. When you sit in the audience, a good event looks easy. But behind each event, there is a team that handles the venue, partners, speakers, the event schedule, livestream, photo, video and the content we publish after. In this post, I want to show you what that work looks like.
GMI Summer Signal ’26 at the Exploratorium
The first event was the one I care about most. GMI Summer Signal ’26, with the theme The Modern Renaissance, brought together teams that build multimodal AI (AI that works with image, video and voice) and AI agents, all in one room at the Exploratorium, one of my favorite venues for AI events in San Francisco. I hosted the event. It ran from 3 PM into the evening, with live product demos, technical talks and product launches.
The question on stage was simple: what happens when image, video, voice, world models (AI that simulates 3D worlds) and agents work together as one system? Builders from GMI Cloud, MiniMax, Inworld AI, Luma AI, Utopai Studios, Reactor, Tripo AI, Resemble AI and LlamaIndex showed what already works in real products today.
When you host a launch event in a museum, every detail matters, from the order of the stage program to the welcome signs. My team ran the program and the partner demos. We also filmed and photographed the event, so the recap video was ready soon after it ended. Here it is:
Hack for Humanity at Entrepreneurs First
Five days later I hosted Hack for Humanity: San Francisco at Entrepreneurs First. It was part of The AI Collective‘s global Hack for Humanity program, powered by Google Gemini. The challenge was personal: pick a real problem in your own community and build something to help fix it.
I love this format because builders work on problems that real people have. Teams came with a problem from their own neighborhood and left with a working prototype. As host, my job was to keep the energy high, help teams get past problems quickly, and give every team fair time to show their demo on stage.
JEVATHON at CodeRabbit
The next Saturday I co-hosted JEVATHON at CodeRabbit, the first community hackathon from TypeSafe AI, organized with The AI Collective. The day had an opening keynote, talks from partners, many hours of building, demos and judging, and a party on the roof at the end.
I gave the Jev DevRel presentation. I showed teams, step by step, how to set up Jev with GMI Cloud, so they could add image, video and voice to their agents early in the day. A good setup session decides how far teams get before the demos start. So I always build it around a finished project that people can copy and build on.
Meeting builders at The AI Conference
Between hackathons I spent time at the GMI Cloud booth at The AI Conference in San Francisco. Working at the booth is some of the most useful DevRel work I do. You hear the real questions people have: how much it costs to run AI models for many users, which models to use for video, and how to turn a demo into a real product. Many of those conversations turned into content ideas and follow-up calls.
Agent Memory Night with AWS during SF Tech Week
For SF Tech Week, The Multimodal Society hosted Agent Memory Night with AWS. The topic is one I know well: AI agents that work great in a demo, but then forget what they learned, lose their place in the middle of a task, or waste money reading the same screenshots and PDFs again and again.
I gave the first talk, called “Making memory work for multimodal agents.” I explained the feedback loops we use in video production and in computer use (AI agents that use a computer like a person does). In a feedback loop, each run leaves notes that help the next run do better. In video, an agent remembers which edits a director approved. In computer use, the agent saves its progress, so after a crash it can continue right where it stopped.
I shared the stage with speakers from FalkorDB, TwelveLabs and Cognee. They talked about how to build memory for agents that run for a long time, how AI can understand video, and how to decide what an agent should keep, forget and look up later.
Here is the full video of Agent Memory Night, with every talk:
Scaling long-form AI video with MCP in Santa Monica
Then we headed south for LA Tech Week. GMI Cloud hosted Scaling Long-Form AI Video with MCP, a beachside brunch workshop in Santa Monica with SHÙ Studio, Real Reel and NVIDIA. The main panel, with creators and the Luma AI team, was about how to go from short AI video clips to full stories that are 30 to 90 minutes long.
The workshop showed how AI studios use the Model Context Protocol (MCP), an open standard that lets AI agents connect to tools and data. With MCP, studios connect video, voice and image models, editing tools, media libraries and agents into one production system. With that system in place, a studio can grow from short clips to full-length shows. When you run a hands-on workshop in another city, you have to set up everything again: the venue, partners, equipment and a team that can help each person get their setup working.
Hybrid Production Night with Luma AI at Dream Lab LA
The next evening I joined Hybrid Production Night with Luma AI at Dream Lab LA. The room was full of filmmakers and creative people who want to use AI in their work. There were demos, one-on-one coaching from Dream Lab’s creative technologists, music, and a live performance capture demo, where cameras record an actor’s movements and face and turn them into digital animation.
I talked about hybrid filmmaking, where AI tools and real cameras work together in one workflow. This is the work my team does every week at Bayhaus Creative, so it felt great to share ideas with creative people who are building the same future. I wrote about a similar crowd in my recap of the AI Filmmaking Masterclass and Hackathon.
Fireworks AI and GMI Cloud podcast at the Bayhaus studio
Back in San Francisco, on Sunday, October 18, Fireworks AI and GMI Cloud recorded a podcast together. I hosted the recording at the Bayhaus Creative podcast studio. We set up the cameras, lights and microphones, so both teams could focus on the conversation.
Bay Area Frontier Research Club at Array Ventures
A few days later, I joined the Bay Area Frontier Research Club at Array Ventures, with General Creative as the sponsor. The topic was how AI learns what people like: reward models (models that score AI answers), how to test model quality, and how personal opinions can become data that a model learns from.
I came to support the research community, and I left with pages of notes for my own work. Creative taste is the hardest thing to teach a model, and it is at the center of AI video. Talking about research papers over dinner keeps the conversation relaxed and deep at the same time.
SCALE at GMI: Spooky Demo Day at AGI House
The last event of the season was SCALE at GMI: Spooky Demo Day for Cohort 2 (the second group of startups in the program), at AGI House in Hillsborough. I hosted the day and spoke on stage. Every team built their product on GMI Cloud GPUs with NVIDIA. Their work covers agents, security, healthcare, infrastructure, robotics and 3D, and AI for specific industries.
Founders came from around the world. Many of them already make money from their products, and many have support from Y Combinator, Antler, Cornell Tech and NVIDIA Inception. Demo day is when months of work in the program finally pay off. It was a joy to give the mic to each team and watch them do a great job.
What hosting AI events in San Francisco and LA taught me
A great team makes multiple cities possible
I host, speak and plan the programs. A team makes each event happen. At GMI Cloud, my DevRel team handles partners, speakers, community and content. At Bayhaus Creative, our team handles photo, video, livestream and editing. With these two teams, we could run events in San Francisco and Los Angeles in the same week, and still publish recap content while the event was fresh in people’s minds.
Every event should leave something behind
A recap video, workshop code on GitHub, a set of photos, a blog post. Content turns a single evening into something people can find months later. I plan the photos and video from the first day of event planning. That habit comes from my years in video production.
The same question shows up in every room
Engineers and filmmakers kept asking me the same thing: how do I get AI to learn from yesterday’s work? The answer has three parts: memory, workflows and human feedback. The teams that move fastest treat all three as one system. I write more about how video shapes my DevRel work in AI developer relations through the lens of 1,600+ videos.
A quick map of my fall AI events in San Francisco and LA
| Date | Event | Location | My role |
|---|---|---|---|
| Sep 14 | GMI Summer Signal ’26 | Exploratorium, San Francisco | Host |
| Sep 19 | Hack for Humanity: San Francisco | Entrepreneurs First, San Francisco | Host |
| Sep 26 | JEVATHON | CodeRabbit, San Francisco | Co-host and speaker |
| Late Sep | The AI Conference | San Francisco | GMI Cloud booth |
| Oct 8 | Agent Memory Night (SF Tech Week) | AWS, San Francisco | Speaker |
| Oct 14 | Scaling Long-Form AI Video with MCP (LA Tech Week) | Santa Monica | Host and workshop lead |
| Oct 15 | Hybrid Production Night with Luma AI | Dream Lab LA, Los Angeles | Speaker |
| Oct 18 | Fireworks AI and GMI Cloud podcast | Bayhaus Creative studio, San Francisco | Studio host |
| Oct 22 | Bay Area Frontier Research Club | Array Ventures, San Francisco | Community support |
| Oct 23 | SCALE at GMI: Spooky Demo Day | AGI House, Hillsborough | Host and speaker |
Let’s build your next AI event together
If you are planning AI events in San Francisco, Los Angeles or beyond, from a developer workshop to a hackathon or a product launch, I would love to talk. I plan the program, create the technical content and host on stage. My teams handle video production, community and the content after the event. Reach out through roanweigert.com, see how we produce events at Bayhaus Creative, or follow my work on YouTube.
FAQ
Which AI events in San Francisco did you host this fall?
I hosted GMI Summer Signal ’26 at the Exploratorium, Hack for Humanity: San Francisco at Entrepreneurs First, and SCALE at GMI: Spooky Demo Day at AGI House. I also co-hosted JEVATHON at CodeRabbit.
What did you talk about at SF Tech Week and LA Tech Week?
At SF Tech Week I spoke about memory and feedback loops for multimodal agents. At LA Tech Week I led a workshop on scaling long-form AI video with MCP and spoke about hybrid filmmaking with Luma AI.
Can you host or speak at my AI event?
Yes. I host hackathons, workshops, demo days and launch events, and I speak on agent memory, MCP, multimodal AI and AI video production. My team can also handle photo, video and livestream coverage.
