Foresight Institute

AI is starting to rewrite what's possible in science. Breakthroughs that once took decades of manual effort are now aided by frontier models that can search vast solution spaces, simulate complex systems, and find patterns no human could. The capability is here. Yet much of science still runs on slow, siloed infrastructure that wasn’t built for AI-native discovery. This September, some of the most exciting minds working at this frontier are joining us in San Francisco for a two day workshop to unhobble our engines of discovery, translation and progress: @tomoxl, Synchron @alexteng101, 50 Years @nabihasaklayen, Cellino @MartinBJensen, Gordian @vivnat, Google DeepMind @SeanEscola, Protocol Labs @mikhailshapiro, Merge Labs @SGRodriques, FutureHouse @elidourado, Astera Institute @alex_pentland, MIT Media Lab @soniaarrison, 100 Plus Capital @beccajcarlson, Deliverome Bio @dileeplearning, Astera Institute @draparente, Civilization Ventures @ejjavorsky, Future of Life Institute @Thane_ac, Deep Science Ventures @MollyminHe, Element Biosciences @Ben_Reinhardt, Speculative Technologies Chelsea Fries, Institute for Protein Design and @madelynheart_ Powered by @protocollabs . Link in first response.

Request to join: luma.com/qa37cydx

Can we trace an LLM's thoughts to detect deception and misalignment? This Thursday, @_georg_lange joins our Berlin AI Node Salon. An independent researcher working on mechanistic interpretability for LLMs, Georg is focusing on circuit tracing, dictionary learning, and automated interpretability to create technology broadly useful for detecting deception, intent, and misalignment. July 16 at CIC Berlin | Link in first comment.

Ask any grad student at an academic conference why they're doing that research, and they'll have an answer. Ask them again: is this the highest-impact thing you could be working on? Most won't say yes. During our conversation last year, @AGamick, co-founder of @Convergent_FROs, pointed out that scientists simply aren't incentivized to ask themselves that question. They're rewarded for getting their next grant, hiring their next grad student, publishing their next paper. That's part of why she and her team help launch Focused Research Organizations: time-bound projects built around one clear, high-impact goal. And they’re pointing these organizations towards the most important problems that, if solved, will allow us to thrive in the age of AI. Learn more about what those are in the full conversation: link in the reply ↓

More about Genoa: linkedin.com/company/genoah…

Every Vision Weekend, we host a $10K pitch contest for projects in frontier science and tech. At Vision Weekend UK, Yasmeen Hmaidan took the win. Her startup Genoa is building a sweat sensor that tracks electrolytes and inflammatory biomarkers in athletes in real time. The prize money will fund prototyping and an early pilot with professional footballers, where recovery monitoring directly affects performance and injury risk. The long-term ambition: bioengineering athletic performance and extending human healthspan. Congratulations Yasmeen, we're excited to follow your work.

If you're researching or building at the frontier of AI-first science, apply to attend. Applications are reviewed on a rolling basis: foresight.org/events/2026-ai…

AI is starting to rewrite what's possible in science. Breakthroughs that once took decades of manual effort are now aided by frontier models that can search vast solution spaces, simulate complex systems, and find patterns no human could. The capability is here. Yet much of science still runs on slow, siloed infrastructure that wasn’t built for AI-native discovery. This September, some of the most exciting minds working at this frontier are joining us in San Francisco for a two day workshop to unhobble our engines of discovery, translation and progress: @tomoxl, Synchron @alexteng101, 50 Years @nabihasaklayen, Cellino @MartinBJensen, Gordian @vivnat, Google DeepMind @SeanEscola, Protocol Labs @mikhailshapiro, Merge Labs @SGRodriques, FutureHouse @elidourado, Astera Institute @alex_pentland, MIT Media Lab @soniaarrison, 100 Plus Capital @beccajcarlson, Deliverome Bio @dileeplearning, Astera Institute @draparente, Civilization Ventures @ejjavorsky, Future of Life Institute @Thane_ac, Deep Science Ventures @MollyminHe, Element Biosciences @Ben_Reinhardt, Speculative Technologies Chelsea Fries, Institute for Protein Design and @madelynheart_ Powered by @protocollabs . Link in first response.

Secure & Sovereign AI Workshop: foresight.org/events/2026-se…

AI development is outpacing our defenses, privacy infrastructure, and coordination mechanisms. AI serving humanity broadly – rather than concentrating power – requires capabilities that strengthen security, protect privacy, and enable peaceful cooperation. That's what this room will be working on. This month, we're gathering some of the best minds working on these questions in Berlin for a two-day workshop on Secure & Sovereign AI: • @BenHagag20, Carnegie Mellon University • @juanbenet, Protocol Labs • @sudo_ml, Project Sovereign • @jesseposner, Vora • @MorganLvng, TechCongress Fellow • @ObadiaAlex , ARIA • @FazlBarez , University of Oxford • @jsotterbach, SPRIND • @elliot_olds, Atstake • @socrates1024, ZCash Foundation • @IvanVendrov, Midjourney • @DavideCrapis, Ethereum Foundation • @hodthoughts, Weizenbaum Institute • @galmasha, University of Connecticut • @JanabelXia, Privacy Residency • Evan Harris, Independent Researcher • Rob Sison, UNSW • Or Zamir, Tel Aviv University • Daniel Juhl, Heron AI • NitzanShulman, Heron AI • Netzer Epstein, Heron AI Sponsored by @protocollabs. Link to event in comments.

Newsletter: foresightinstitute.substack.com/p/vision-weeke…

Vision Weekend USA Early Bird ends today! Plus: a recap from London, new grantees, and prize applications close July 31. Our monthly newsletter is out with the latest from across the Foresight ecosystem. Link in first reply.

Submissions close tomorrow. Last chance to win up to $5K for building a grounded 2035 scenario where AI goes well. The AI Futures Challenge 2026 is open until June 30, any time on Earth. Just take our free 1.5hr course and share your most hopeful yet realistic vision for 2035. $5K Grand Prize + 5 × $1K Bounties. There’s still time → #AIFutures" target="_blank" rel="nofollow noopener">worlds.existentialhope.com/#AIFutures

Request to join: luma.com/ujoakb1l

This Wednesday, Paul Schneider joins our Berlin AI Node Salon for a session on designing AI interfaces for researchers – from coding agents to research collaboration. Originally trained as a physician and health economist, Paul has worked across academia, consulting, and entrepreneurship before turning his focus to AI research tooling and building Shoulders: an AI workspace for researchers. Location: CIC Berlin | Chatham House Rule applies. Link in first reply.

We’re excited to share the results of another collaboration with the team at @figbrains! GUI-Perturbed is part of GOLEM, Manifold’s long-term research program focused on building AI systems that can perceive, reason, adapt, and act across any input and output modality.

My AAAI 26 talk on the first formal guarantees on corrigibility and the limits of safety filters is now online (link in next tweet below 👇)

It’s 2035, and AI has gone well. What could a normal day in your life look like? Most popular AI futures are dystopias. Almost no one takes the time to seriously and vividly imagine a future that's actually worth building. So we made a short film following one day in the life of an AI auditor in 2035, in a world radically transformed by AI for the better. Digital twins model how people’s unique physiologies respond to illness. Sensors stop pandemics before they start. AI helps governments actually listen to citizens by synthesizing vast amounts of data into actionable policy suggestions, but humans stay in control and make the final call. If this resonates, please share to spread the word about positive AI futures!

Early bird tickets: foresight.org/events/vision-…

One week left to get early bird tickets for Vision Weekend USA. Ticket prices increase after June 30. November 13-15 in San Francisco, join us for our main 40th anniversary celebration in the US: talks on frontier science and tech, expert office hours, 1:1 connections, tech demos, and a $10K pitch competition. The All Access pass includes a pre-party on the USS Hornet, lab tours at @E11BIO and @LongshotSpace plus a Cypherpunk party at our AI Node. Speakers include: @v_maini @ilblackdragon @SumnerLN @woj_zaremba @seemaychou @hannu @PracheeAC @sebkrier @robinhanson and @AdamMarblestone. Powered by @protocollabs. Link in first reply.

Can AI agents be made to jailbreak each other, automatically? Tomorrow, AI researcher @SamuelNellessen joins our AI Node in Berlin to present on Slingshot: Automating Agent-to-Agent Jailbreaks with RL. Specializing in empirical alignment and agentic security, Samuel’s work focuses on transitioning safety evaluations from subjective natural language processing tasks into objective control problems. June 24 at CIC Berlin | Chatham House Rule applies. Request to join: luma.com/85yfklbx