Humanity’s biggest challenges unfold over years, decades or longer. Curing disease, new scientific discoveries, navigating financial markets… But today’s AI still operates within short horizons, e.g. minutes, hours or days. In my talk at the @aiDotEngineer World's Fair, I explored how AI models can scale to long horizons, which is our key research focus at @GenReasoning. Long-horizon thinking is not only a technical challenge. It is a mindset: the ability to hold purpose, adapt and keep making progress towards goals, both in machines and in ourselves.

Looking forward to sharing the insight from building Avalon, the world’s first autonomous company model. I’ll discuss how we trained AI as an organisation, the surprising structures that emerged, and what comes next as we scale collective intelligence.

Building in AI often makes me look back at human intelligence. Our brains haven’t changed much in thousands of years, yet civilisation has advanced beyond recognition. A big part of that progress came from our ability to cooperate, share knowledge and build on each other’s work. I believe a similar principle could shape the next leap in AI, not just smarter individual models, but collective intelligence developed responsibly.

Proud to see our KellyBench featured in the State of AI Report 2026. Frontier models can be impressive in the moment, but when decisions have to compound over time in a changing environment, they often fall apart. That’s the problem we’re tackling at @GenReasoning: long-run intelligence. I increasingly think long-horizon thinking is more than a technical challenge. It’s a mindset: hold onto the goal, adapt as reality changes, and keep making progress. That matters not just in machines, but also people. More from General Reasoning soon. Congrats to @nathanbenaich on another thoughtful, measured State of AI Report.

The market doesn’t always know what it wants. If it already knows exactly what it wants, you may already be too late. A founder’s job isn’t just to respond to demand. It’s to see what others don’t yet see and build before the market knows how to ask for it. The price of seeing early is often loneliness. For a while, you may be building something few people understand, with very little evidence that you’re right. That’s part of leading rather than following. From my conversation at Cambridge Tech Week on building the next global tech champions. #CamTechWeek

If you’re building AI for quant finance, forecasting or simulation, one challenge is look-ahead bias: models accidentally seeing information from the future. We released BackSearch to address this pain point, letting LLMs search the web as it existed on a specific date. Give it a try and let us know what you think.

Excited to speak at @aiDotEngineer World’s Fair in SF with my co-founder @rosstaylor90. Our topic is about one of the biggest challenges in AI today: moving from agents that complete short tasks to those that make progress over long horizons. Today’s AI can answer questions, write code, search the web, summarise documents, and complete workflows that last minutes, hours, or days. But the most meaningful work in the world does not happen in short bursts. 🧬Drug discovery can take decades. 🧪Scientific breakthroughs require thousands of experiments, failed hypotheses, course corrections, and long-term memory. 🏦Building companies, infrastructure, and complex systems requires the ability to plan, adapt, recover, coordinate, and keep pursuing a goal over time. This is the long-horizon challenge. How do we build AI systems that do not just respond well in the moment, but can keep working, learning, and improving across time? In the talk, I’ll cover the algorithms, environments, and infrastructure to tackle this challenge. AI’s next leap will come from scaling AI to long-horizon capabilities, where agents can reason, adapt, and coordinate over long time periods. That's how we drive meaningful progress in society, science, and the economy.

Real intelligence isn’t just about getting the right answer, but most importantly, navigating a complex world that keeps changing. Our Complex Worlds Hackathon was a small step toward that shift. What stood out wasn’t just technical strength, but intent. Builders chose hard, meaningful problems, e.g. robotics for elder care, better nursing systems, fully aware they’re difficult, but worth solving. After the event, people came up to me saying: “I knew it was tough, I knew it may not win the prize, but I want to work on it.” That’s what stayed with me: brilliant minds, grounded in humanity. The future of true intelligence.

POMDP and Circumstance at the Complex Worlds Hackathon Giovanni R (Edison Scientific), @ibragim_bad (Nebius), Jeff Smith, @ChengxiTaylor (GR)

This Saturday, we’re hosting the Complex Worlds Hackathon in London, in partnership with @joinef and @airstreet. The response from the community has been incredible. But this isn’t just a hackathon for engineers. It’s about a deeper question: how intelligence actually develops. Today’s AI is powerful, but often short-sighted. It struggles with long-term planning, adapting to change, and operating in messy, real-world conditions. That’s because intelligence doesn’t emerge from static questions. It emerges from interaction, feedback, and experience, from the environments we place agents in. This is why at @GenReasoning, we’re focusing on building long-horizon reinforcement learning environments: worlds where agents must act over hundreds or thousands of steps, adapt to non-stationarity, and develop capabilities that don’t show up in short tasks. If we want AI that can truly operate in the real world - in science, business, creativity - we need to rethink the environments we train it in. That’s what this weekend's hackathon is about. The next leap in AI isn’t just bigger models; it’s better environments. Intelligence isn’t built from a single answer; it’s built from experience.

If you care about long-horizon reasoning, this is where it starts. Excited to host this in London with great partners. Applications open 👇

“AI models from Google, OpenAI and Anthropic lost money betting on football matches over a Premier League season, in a new study by @GenReasoning suggesting even the most advanced systems struggle to analyse the real world over long periods of time. The “KellyBench” report released this week by AI start-up General Reasoning highlights the gap between AI’s rapidly advancing capabilities in certain tasks, such as writing software, and its shortcomings in other kinds of human problems.” as.ft.com/r/2b613fe7-4562-44…

Most real-world decisions don’t happen in a single moment. They unfold over time, under uncertainty, with real consequences. Succeeding in the real world isn’t just about being right once. It’s about staying right as conditions change, managing downside, and making decisions over time. This gap matters.

@gandhikanishk's Endless Terminals is the most popular env on OpenReward!

🌍 Environments of the Week It's been a week since we launched @OpenReward. Here are some of our favourite environments this week - some newly added, some heavily used, and some hidden gems. First, the most used environment of the week is EndlessTerminals by @gandhikanishk with 830k+ tool calls. openreward.ai/kanishk/Endles… 🧵

Cool idea from @AashaySachdeva: unified environment interfaces like @OpenReward can enable LLM meta-learning research! Pleased with where things are going with more parts of the stack accessible publically. For e.g. I now look forward to weekly @tinkerapi roundups as much as John Oliver episodes!

.@benchflow_ai started in 09/24 as unity for benchmarks and a hosting hub with early users from Stanford and Princeton. 4 months before R1 dropped We stopped after 9 months with 0 traction. Today our latest work SkillsBench is #1 trending on @OpenReward. Game of eval is just on

Played around with this. This was exactly something I was looking for! Tried a few things - Creating an env - pretty dope! end to end claude was able to port it from github with only minor issues. One shotted @ShashwatGoel7 OpenForecaster env here. A lot more people should contribute their own envs. I hope they launch monetisation here. Running a curator over env tasks during RL - When there are so many tasks, which one should you focus on? This is the auto-curriculum/meta-learning bit. I am still not able to beat random/pass@k but I think signals are there over long run this will help with diversity. This obviously has a power law, every run will have top envs dominating but I feel those 20% random tasks will give a big boost to any model. optimise the GEPA optimiser - gepa is great but pretty slow. What if we could teach a model to do this better? This was in my list for so long, finally with openreward was able to attempt it.

Cool to see OpenForecaster environment trending on @OpenReward. Thanks @AashaySachdeva for porting it!