90/30 Club (ML reading) #26: Tiny Recursive Model
- Date
- 2025-10-28
- Location
- San Francisco, CA, USA
- Host
- Luma
About this event
Tiny Recursive Model is the kind of paper that can spark a real room: part technical reading group, part live sense-making session, and part community gathering for people who like to think carefully about where ML is headed. If you want more than a surface-level take, this 90/30 Club session gives you a place to read, discuss, challenge assumptions, and leave with a sharper understanding than you would get from skimming alone. What Is This? 90/30 Club is an in-person ML reading gathering built around a simple idea: set aside focused time for a serious paper or concept, then use the group to pressure-test your understanding. This session centers on Tiny Recursive Model, making it a good fit for people who enjoy discussing model design, reasoning, autonomy, and the broader implications of new ML approaches. The "90/30" format signals that this is not a passive talk. The emphasis is on spending meaningful time with the material and then using discussion to connect technical details with larger questions. Rather than treating the paper as something to "cover," the goal is to work through what it is actually claiming, how convincing it is, and what it might change about the way we think about machine learning systems. Because this is an in-person event in San Francisco, the format also matters socially. Reading groups are one of the best ways to meet thoughtful people in ML without the pressure of a conventional networking event. Conversation tends to be more grounded when everyone is reacting to the same object of study. What to Expect You should expect a structured but conversational evening. The session will likely begin with a quick framing of the paper and the key questions worth tracking as you discuss it: what the model is trying to do, what is novel, what assumptions it relies on, and where its strengths or limitations may be. From there, the group can dig into the material itself. That often means unpacking the core idea behind the model, clarifying terminology, and comparing interpretations with other attendees. If you've ever read a paper and felt like you understood 60% of it but wanted a room to help close the gap, this is the right format. Likely areas of discussion include: The main claim of Tiny Recursive Model and what makes it interesting How the approach works at a conceptual level What counts as evidence that the method is useful or important Where recursion, reasoning, or autonomy show up in the paper's framing What this might imply for future model design, evaluation, or deployment There is also room for disagreement, which is part of the value. Good reading groups are not about pretending every new ML result is profound. They are about separating signal from hype, identifying open questions, and learning how other technically curious people read the same work. Why Attend If you work in ML or follow it closely, you already know that the hardest part is often not finding papers. It is figuring out which ones deserve sustained attention and developing a robust view of them. This event helps with both. You get dedicated time to focus on one piece of work and a group setting that pushes the conversation beyond first impressions. This is especially useful if you care about autonomy-related questions in AI. A paper like Tiny Recursive Model can be discussed not only as a narrow technical artifact, but also as part of a larger shift in how people think about capability, efficiency, recursion, control, and system behavior. The reading group format gives space for both technical and strategic angles without flattening either one. You may leave with: A clearer understanding of the paper's actual contribution Better intuitions for how others in the ML community are interpreting it New questions worth exploring in your own work or reading Connections with people who enjoy serious, grounded technical discussion That combination is hard to get from online discourse alone. In-person discussion tends to produce more nuance, more honesty about confusion, and more useful follow-up conversations. Practical Details This event takes place in person in San Francisco, USA on Monday, October 27 at 7:00 PM PDT. The in-person format is a strong part of the appeal: you can engage deeply, ask questions in real time, and continue conversations before or after the main discussion. Because this is a reading-focused gathering, it helps to arrive ready to participate. You do not need to have a perfectly polished interpretation of the paper, but it is worth coming with at least a basic sense of the topic and a few questions or points of confusion you want to bring to the room. A few useful ways to prepare: Skim or read Tiny Recursive Model in advance if you can Note one or two claims you find compelling, unclear, or questionable Be ready to listen closely and revise your view during discussion Come prepared for both technical conversation and informal networking If you like ML events that are substantive, social, and intellectually honest, this session is a strong use of a Monday evening. It is for people who want to think with others, not just consume content.
Who should attend
This will feel especially worthwhile if you like learning through discussion and want sharper thinking, not just more information. - You follow **machine learning research** and want a better way to engage with new papers than reading alone or relying on social media summaries. - You are interested in **autonomy, reasoning, recursion, or model design** and want to examine how those ideas show up in a concrete piece of work. - You work as an **engineer, researcher, founder, student, or independent builder** and value conversations that connect technical details with broader implications. - You enjoy **reading groups, salons, and focused technical meetups** where people actually discuss the material rather than sit through a generic presentation. - You are looking to meet **thoughtful ML people in San Francisco** through a setting that makes conversation easier and more substantive than standard networking events. - You do not need to be the most advanced person in the room; this is also a good fit if you come prepared, ask good questions, and want to strengthen your ability to read and evaluate ML work carefully.