Open Source AI Summit

September 10-11, 2026

The Presidio • San Francisco

AI does not have to be controlled by corporations, opaque models, and black-box infrastructure.

The Open Source AI Summit explores an alternative ecosystem rooted in independence and decentralization, where individuals build, run, and own their intelligence.

For two days, we'll explore the technical, economic, philosophical, and geopolitical dimensions of open-source AI. Topics will span open models and local inference to verifiable hardware, AI economics and safety, and the right to compute.

Topics We'll Explore

The technical, economic, philosophical, and geopolitical dimensions of open source AI.

  • Local AI for Business

    Real-world enterprise deployments with local and open AI.

  • Open Model Use Cases

    Lessons from deploying open models in production.

  • Inference & Compute

    How modern AI inference works, from scaling and efficiency to infrastructure.

  • Agentic Frameworks

    Model interoperability, orchestration, and dev tooling.

  • Economics & Markets

    Open protocols, AI commerce, and agent economies.

  • AI for Science

    Accelerating scientific discovery through open models and research.

  • AI Safety & Guardrails

    Balancing capability, transparency, and responsible deployment.

  • Verifiable Hardware

    Confidential computing, TEEs, and trusted AI infrastructure.

  • Private AI

    Building AI systems that protect sensitive data by design.

  • The Right to Compute

    The Computing Manifesto, AI regulation, and the First Amendment.

  • Open vs. Closed AI

    Why the debate is not just technical, but philosophical as well.

  • The Global AI Race

    Geopolitics, national competitiveness, and AI sovereignty.

Speakers

More coming soon

Showing all 47 speakers

Agenda

Thursday, September 10

9:00–9:30
Breakfast
9:15–9:25
Kickoff
9:30–9:45

Ramez Naam

Investor & Award-Winning Author

Many fear AI will concentrate power and fuel authoritarianism. However, an alternative future is possible. One where open source, plural AI boosts human freedom and empowers democracy.

Ventana Room

9:50–10:20

Vincent Weisser · Era Qian

Prime Intellect · Edge

This conversation will cover Vincent’s founder’s journey, the evolution of the open-source AI ecosystem, and what it takes to get to open superintelligence.

Ventana Room

10:30–11:05

David Sacks · Steve Lee

Craft Ventures · Spiral

This discussion will cover public sentiment toward AI, the risks of distributing vs. centralizing powerful models, the importance of protecting open-source AI and America’s leadership in it, and the future of autonomous agents and payments.

Ventana Room

11:05–11:20
Break
11:20–11:40

Ahmad M. Osman

Osmantic

Ahmad M. Osman is an AI researcher who spends much of his time growing the Local AI community and creating educational resources. Today, he’ll share the state of Local AI and why open source and Local AI are relevant to everyone, not just organizations and businesses.

Ventana Room

11:45–12:00

Jesse Posner · Erik Cason · Nick George

Vora

Vora is building the world's most loyal AI. It knows your life, it works for you alone, and it stays yours forever. The Vora team will reveal the machine that runs it for the first time: open-source, private by design, in your home on hardware you own.

Ventana Room

12:15–12:45
Ventana Room

Michael Millerman · Jesse Posner

Millerman School · Vora

Jesse Posner interviews political philosopher Michael Millerman about how we can use philosophy to better understand the arguments now running through AI. Open against closed, centralization, safety, acceleration: each carries assumptions about power, freedom, and technology that political philosophy has worked on for centuries.

Ventana Room

Cypress Room

Zain Hasan

Together AI

We'll follow one request end to end, from the HTTP POST to the tokens streamed back out, and see what the engine is actually doing at each step. Tokenization and the input pipeline. How the scheduler decides what runs next, and how continuous batching lets a single forward pass serve hundreds of users at once. How chunked prefill keeps one giant prompt from ruining latency for everyone else. How the KV cache is managed like virtual memory with PagedAttention, and how prefix caching quietly saves you work you didn't know you were repeating. Then sampling and detokenization on the way back out.

Cypress Room

12:45–1:35
Lunch · Hawthorn Room
1:35–2:00
Ventana Room

Noah Smith

AI2

Modern AI systems are often cast as products: a model, a chatbot, an API. But for science, that framing is too narrow. Scientific communities need more than access to finished systems: they need the ability to understand, modify, reproduce, and extend them. That requires open AI infrastructure spanning the model-development lifecycle, from data and training methods to intermediate artifacts, evaluations, and documentation. In this talk, I will use Ai2’s Olmo project as a case study, highlighting examples of what becomes possible when researchers can study and intervene in model construction rather than only consume its outputs. I will argue that this kind of openness is essential if universities, nonprofits, governments, and scientific communities are to build AI systems that serve their own purposes. The goal is not to reproduce commercial AI in the open, but to create an ecosystem that gives scientific communities greater agency and accelerates discovery.

Ventana Room

Cypress Room

Dustin Dannenhauer

Delegance AI

This talk examines modern AI-assisted coding approaches (one-shot, /goal, dynamic workflows, loops, and graphs), how they apply to different kinds of work, and where local models fit in. Dustin will introduce Playbooks, a new approach that coordinates multiple models, including local models, to tackle complex work.

Cypress Room

2:10–2:40
Ventana Room

Max Webster · John Komkov · Christian Catalini

Hivemind Ventures · Fathom Capital · MIT Cryptoeconomics Lab

A discussion on the current and future split of tokens and economic value between open-weights models and the closed frontier, as well as pathways toward agent autonomy and visions for capitalizing autonomous agents.

Ventana Room

Cypress Room

Ravin Kumar

Google DeepMind

Open models have one often overlooked quality, that is they can be customized. In addition to open weights these days there’s a plethora of open tools to help you update those weights and specialize models to frontier performance, even over the usual suspects. And importantly let you create a model that is uniquely your own. In this talk we’ll cover some notable variants of Gemma and how you can make your own.

Cypress Room

2:50–3:20
Ventana Room

Greg Lukianoff · Conner Brown · Kyle Olney

FIRE · Bitcoin Policy Institute

An exploration of whether model weights and outputs qualify as protected speech, how AI regulation intersects with the First Amendment, and whether you have a right to run intelligence on hardware you own.

Ventana Room

Cypress Room

Jasper Hugo

Spiral · goose

This talk will show a live LM Inspector and introduce Eredu, the library behind it. Eredu is a Rust engine for running language models locally, built not just to produce text but to make model generation inspectable. LM Inspector is a desktop app that follows generation token by token, shows the candidates a model scored for each token, and lets users pause, rewind, branch, and compare results. The talk will end with a live walkthrough of inspecting generated text.

Cypress Room

3:20–3:35
Break
3:35–4:05
Ventana Room

Max Webster · Tom Brow · Ilan Strauss · Roland

Hivemind Ventures · Block · AI Disclosures Project · taskfuel.ai

A discussion around the importance of open protocols for AI systems in creating markets and new forms of emergent intelligence. In particular, we’ll focus on existing and potential standards for communication, identity, and commerce.

Ventana Room

Cypress Room

Douwe Osinga · Dustin Dannenhauer

Spiral · Delegance AI

A look at different agentic harness design approaches, where the landscape is headed, and how harnesses can improve agent capabilities.

Cypress Room

4:15–4:45

Steve Lee · Elias Rohrer · Jordan Mecom · Rob Hamilton

Spiral · Block · AnchorWatch

This panel will explore how AI is reshaping open source software security, including the growing role of AI-powered security scanning in both identifying and exploiting vulnerabilities. Leaders from the bitcoin space will speak about lessons learned, the race between attackers and defenders, and why access to open-weight models may be essential to protecting open source code.

Ventana Room

4:45–5:45
Happy Hour · Golden Gate Club

Sponsors

BuzzBuilt byVoraOpenTeams

Apply

Apply to Attend

If accepted, reserve your spot for $150. A limited number of complimentary passes are available for open source AI builders and academics.

Apply to Speak

Propose a talk, panel, workshop, or discussion. We're shaping the program now and want your ideas. Accepted speakers attend free.

In the Meantime

Join the community at Builder, a free monthly AI meetup in the Presidio that explores many of the same topics as the summit.

Meetup

About the Host

The Open Source AI Summit is organized by Presidio Bitcoin, a hub for builders and startups working toward a more private and decentralized future across AI and bitcoin.

September 10-11, 2026

Specific venue, detailed schedule, and speakers announced closer to the event.

Interested in sponsoring the summit? Email hello@opensourceaisummit.org

© 2026 Presidio Bitcoin