The Heuristic Scientist: Open-Ended Algorithm Discovery with LLMs

Name: The Heuristic Scientist: Open-Ended Algorithm Discovery with LLMs
Start: 2026-06-03T18:00:00+02:00
End: 2026-06-03T21:00:00+02:00
Location: Berlin

Date: 2026-06-03
Location: Berlin, BE, Germany
Host: BLISS Calendar

About this event

Large language models are changing more than how we write code or summarize papers. They are starting to change how we search for algorithms themselves. The Heuristic Scientist: Open-Ended Algorithm Discovery with LLMs is an in-person Berlin gathering for people who want to think seriously about what happens when language models move from assistant to research partner. This event is built for curious technical minds, researchers, builders, and thoughtful newcomers who want to explore a fast-moving topic with others in the room. If you care about AI systems that can generate, test, refine, and extend heuristics in open-ended ways, this is a chance to discuss the ideas in person and meet people asking similar questions. What Is This? At its core, this event is about a big shift in AI: using LLMs not just to explain known methods, but to help discover new ones. The phrase "heuristic scientist" points to a style of system that can propose strategies, iterate on them, and search through possibilities in ways that look more like exploratory science than straightforward prediction. Rather than treating algorithm design as something humans do alone and models merely document, this session focuses on the emerging idea that LLMs can participate in open-ended discovery. That raises exciting technical questions: how do you evaluate generated heuristics, structure feedback loops, preserve novelty, and know when a promising idea is actually useful? This is an in-person event in Berlin designed around shared learning and real conversation. Expect a format that combines focused discussion of the topic with opportunities to connect with other attendees interested in LLMs, AI research, experimentation, and the broader community forming around these questions. What to Expect The evening will center on the theme of open-ended algorithm discovery with LLMs and the broader implications of systems that can generate and refine heuristics. You should expect a program that gives enough structure to ground the topic while leaving room for discussion, questions, and exchange among people with different backgrounds. Likely areas of focus include: What heuristic discovery means in practice and how it differs from standard prompting or code generation How LLMs might support search and experimentation across possible algorithmic approaches Evaluation challenges, including how to test whether a generated heuristic is robust, novel, or merely plausible-sounding Research and product implications for people building tools, studying model behavior, or exploring agentic workflows Open questions around reliability, iteration, human oversight, and the limits of current systems Because this is an in-person community-oriented event, a meaningful part of the value will come from the room itself. You can expect conversations with people who approach AI from different angles: some more theoretical, some more applied, some deeply technical, and some primarily interested in where the field is heading. There is also a practical networking dimension here. If you are looking to meet others in Berlin who care about LLMs and AI beyond surface-level hype, this event creates a setting where the conversation can get specific. The topic naturally invites discussion about experimentation, model capabilities, research workflows, and the future of algorithmic creativity. Why Attend If you have been following LLMs closely, you have probably noticed that the most interesting developments are no longer just about better outputs. They are about systems that can explore, revise, and search. This event gives you a focused way to engage with one of those developments: the use of language models in open-ended algorithm discovery. You should attend if you want sharper mental models for where LLM-based research systems may be going next. The subject sits at the intersection of AI reasoning, automation, scientific process, heuristics, and tool-building, which makes it especially relevant if you work across research and application. A few concrete reasons this event may be worth your evening: You will leave with a clearer frame for understanding how LLMs can be used beyond content generation and coding assistance You will hear how others think about open-ended discovery, including what seems promising and what still feels unresolved You will meet people in Berlin who are actively interested in AI, LLMs, and the research community around them You will have a chance to test your own ideas in conversation, whether you are skeptical, optimistic, or somewhere in between Just as importantly, the event offers something many online discussions do not: context-rich, live exchange. Topics like heuristic search, emergent strategies, evaluation, and algorithm design benefit from back-and-forth discussion, not just hot takes or isolated demos. Practical Details Location: In person in Berlin, Germany. This is a physical gathering, so plan to attend on-site and take advantage of the face-to-face discussion and networking. Date and time: Wednesday, June 3 at 6:00 PM GMT+2. Since this is an evening event, it is well suited for people joining after work, research, or study. This event is tagged across LLM, AI, community, and networking, which is a useful signal for what kind of room to expect. The audience is likely to include people who want both substance and connection: thoughtful discussion of a specific topic, plus the chance to meet others who are paying close attention to where AI is going. If this topic intersects with your work or curiosity, the simplest reason to come is that it is easier to think well in a room with other serious people. Berlin has no shortage of AI interest, but not every event gets specific. This one does. Show up ready to listen, ask good questions, and talk with others about what algorithm discovery with LLMs could actually become.

Who should attend

This event is for you if you want a serious, in-person conversation about where LLMs may be pushing beyond assistance into discovery. - **You work with LLMs or AI systems** and want to better understand how models might help generate, test, or refine heuristics rather than just produce answers. - **You are a researcher, engineer, or technical builder** interested in algorithm design, evaluation, experimentation, or agent-like workflows. - **You are exploring the frontier between research and application** and want to discuss what open-ended discovery could mean in practice, not just in theory. - **You are intellectually curious about AI progress** and want a more grounded conversation than generic "future of AI" panels usually offer. - **You value thoughtful community and networking** and would like to meet other Berlin-based people paying close attention to LLMs, AI, and emerging research directions. - **You do not need to be an expert on this exact topic** to get value from attending, but you should be genuinely interested in how language models may participate in scientific or algorithmic exploration.