OpenAI o1, every mention
99 scenes · ← back to OpenAI o1
tap a year for its mentions
every year anyone Shawn Wang 49Alessio Fanelli 39Ben Hillock 17Nathan Lambert 12Kevin Weil 7Itamar Friedman 7NotebookLM Host 2 6Florent Crivello 5Charlie Snell 5Noam Brown 4
Verbatim, from the transcripts: the passages where OpenAI o1 comes up
🔬 RL with Verifiable Rewards, but the Verifier is a Lab — Lila Sciences
Cooking with OpenAI’s Research Chief: AGI, o1, Evals, and Scaling Laws — Mark Chen
Scaling Past Informal AI - Carina Hong, Axiom Math
🔬How GPT‑5 derived new results in theoretical physics and quantum gravity — Alex Lupsasca, OpenAI
Marc Andreessen introspects on Death of the Browser, Pi + OpenClaw, and Why "This Time Is Different"
- ▶ 0:27 Marc Andreessen It's an overnight success because it's like, bam, you know, ChatGPT hits, and then, and then O-One hits, and then, you know, OpenClaw hits, and like, you know, these are open, these are, these are like overnight, like radical overnight…
- ▶ 8:36 Marc Andreessen Cause it's like, bam, you know, chat GPT hits and then, and then O-one hits and then, you know, open call hits. 2 times in the scene
- ▶ 29:41 Marc Andreessen So open AI comes out with a one and it's an amazing technical breakthrough and it's just like absolutely fantastic.
⚡️ Reverse Engineering OpenAI's Training Data — Pratyush Maini, Datology
- ▶ 16:35 Pratyush Maini I think also what's interesting to me that the GPT-D-Fort.one series came about four months after the O-one data was available, so it's kind of interesting, like, how fast do Frontier Labs move? 2 times in the scene
Artificial Analysis: The Independent LLM Analysis House — with George Cameron and Micah Hill-Smith
- ▶ 23:40 Micah Hill-Smith If we go back to just over a year ago, before oh one, 2 times in the scene
[State of RL/Reasoning] IMO/IOI Gold, OpenAI o3/GPT-5, and Cursor Composer — Ashvin Nair, Cursor
- ▶ 0:18 unnamed speaker Before that, you were opening, I worked in O.M.O.N.O.T.E. 2 times in the scene
- ▶ 17:18 unnamed speaker The idea was, you, you started with codecs, someone else was doing instroft GPT, then we launched GPT, four, four O, I guess O-one. 3 times in the scene
- ▶ 24:48 unnamed speaker Yeah, was there an internal prototype pre-O'one that was like, okay, this is the thing, we'll fund it to scale it up, right? 2 times in the scene
How Zyphra went all-in on AMD + Why Devs feel faster with AI but are slower — with Quentin Anthony
- ▶ 37:03 Quentin Anthony Uh, oh, one's like the initial thinking models were a big deal when I was doing like core academic, like how do I create a performance model for explaining how this kernel behaves?
Why RL Won — Kyle Corbitt, OpenPipe (acq. CoreWeave)
- ▶ 14:45 Alessio Fanelli Was it when a one preview came out? 3 times in the scene
The RLVR Revolution — with Nathan Lambert (AI2, Interconnects.ai)
- ▶ 27:56 Shawn Wang There's a very good post on, just on the retrospective of Qstar, there's a very good post, uh, that you had, which was that I want to send people to is, which is that it was open, was O-one a psyop, right? 2 times in the scene
- ▶ 38:37 Nathan Lambert What I would say that we have already done with O-one and R-one, which is you do a lot of RL, you show the inference time scaling works and you get really high benchmark numbers.
- ▶ 46:57 Shawn Wang There's one case where, with O-one and sort of the, the sort of Q-star ideas, there was one case where it was sort of overhyped in some sense, but now it's coming back with O-one Pro and DeepThink. 2 times in the scene
⚡️Anthropic vs Cognition on Multi-Agents: A Breakdown with Dylan Davis
- ▶ 24:27 Shawn Wang We are in one of these situations where, you know, in a world where the cost of intelligence for a given set of intelligence, let's say GPT-IV, let's say O-one, whatever, it is literally falling a hundred X over the course of one year.
Information Theory for Language Models: Jack Morris
- ▶ 5:51 Jack Morris Like you've seen that so many times, most recently, probably with the reasoning models, like oh one came out of open AI September, 2024 last year.
- ▶ 42:58 Shawn Wang If you gave the O-one harness on top of GPT-II, you would get nothing because GPT-II didn't know enough. 2 times in the scene
Scaling Test Time Compute to Multi-Agent Civilizations — Noam Brown, OpenAI
- ▶ 5:28 Spooks (Swyx) I think the last time you did a lot of publicity, you were just launching O one, you did your Ted talk and everything. 2 times in the scene
- ▶ 28:07 Noam Brown And I remember it was interesting that I talked to somebody who left OpenAI after we had discovered the reasoning paradigm, but before we announced a one, and they ended up going to a computing lab. 4 times in the scene
⚡️Open Questions in Agentic RL — Will Brown (Prime Intellect)
- ▶ 2:55 Will Brown Single-turn RLVR world, the models like O-one and R-one, and we want to, like, have these things become more agentic, and it seems like the path is to incorporate reinforcement learning into this process.
What is an RL environment? w/ Nous Research's Roger Jin
Sleep-Time Compute — Letta AI (Charles Packer, Charlie Snell, Kevin Lin)
- ▶ 21:17 Charles Packer How many people, you know, turn off of, like, O-one and, like, go to, like, O-three mini or, like, I guess, like, whatever the, you know, the past or low latency one is just because they can't stand waiting that long for the answer to come…
- ▶ 23:18 Charlie Snell I mean, we also have some results on like the, uh, like, oh, one, oh, one mini where those ones actually will be spending minutes. 5 times in the scene
GPT 4.1: The New OpenAI Workhorse
- ▶ 27:09 unnamed speaker Should I use a one and make a plan and then use 4.1 to implement the plan? 2 times in the scene
SF Compute: Commoditizing Compute
- ▶ 19:46 Michael Swix (Swyx) So let's say GPT-IV and O-ONE both had total training costs of like a five hundred million dollars is the rough estimate.
The #1 SWE-Bench Verified Agent
- ▶ 3:17 Guy Gur-Ari Uh, I think we ended up using a one for that.
The new OpenAI Agents Platform: CUA, Web Search, Responses API, Agents SDK!!
- ▶ 9:55 Alessio Fanelli I cannot use O-one and call search as a tool.
S1: the $6 DeepSeek R1 Competitor (ft. Entropix)
- ▶ 4:19 unnamed speaker Yeah, I, I was, I was just thinking about the R one paper, or actually the S one, the O one paper, so many letters and numbers.
smol agents are all you need
- ▶ 14:10 unnamed speaker Is it just because of like O-one is better than Sonnet or, um, yeah. 6 times in the scene
The AI Architect: Bret Taylor
- ▶ 1:33:21 Bret Taylor We know one came out really interesting quality wise, but it's quite slow, quite expensive. 3 times in the scene
The Agent Reasoning Interface: Claude, ChatGPT Canvas, Tasks, Operator — with Karina Nguyen, OpenAI
- ▶ 0:39 Shawn Wang Streaming chain of thought for O-one models and more via novel synthetic model training.
- ▶ 16:21 Karina Nguyen So for example, like I think the way you prompt a one or like cloud three is going to be very different from each other.
- ▶ 17:37 Alessio Fanelli And talking about O-one prompting, we just had a O-one prompting post on the newsletter, which I think was the- 10 times in the scene
Beating OpenAI and Anthropic by Looking At Data: the new #1 on SWE-Bench w/ W&B CTO Shawn Lewis
Outlasting Noam Shazeer, Crowdsourcing Chai AI w/ 1.4m DAU — with William Beauchamp, Chai Research
- ▶ 1:10:49 William Beauchamp And that's what opening I've done with O-one and O-three. 2 times in the scene
The Unreasonable Effectiveness of Reasoning Distillation: using DeepSeek R1 to beat OpenAI o1
- ▶ 1:35 Shawn Wang Distilling DeepSeq into Quen and basically beating O-one preview. 2 times in the scene
- ▶ 6:48 unnamed speaker And then just to keep digging on that, a lot of this on the reasoning side you can do now, because of course, O-one doesn't show the reasoning traces. 4 times in the scene
- ▶ 11:05 Shawn Wang Like the only way you, you reach the frontier with the full size models of O-one and O-three is with that stuff.
- ▶ 19:45 Shawn Wang And that score was actually higher than O one when it came out.
DeepSeek V3, SGLang, and the state of Open Model Inference in 2025 (Quantization, MoEs, Pricing)
- ▶ 46:46 unnamed speaker I also noticed that with OpenAI offering fine tuning for O-One and all these things, I think people are also very interested in sort of RL trainers is what they, what you have here.
OpenAI o1 isn’t a chat model (and that’s the point)
- ▶ 0:09 Alessio Fanelli We have actually a podcast that is preceded by an essay that Ben Hillock wrote, um, on Latent Space recently about how he was wrong about O-one and how to actually prompt it to, to get good results. 4 times in the scene
- ▶ 1:14 Ben Hillock Cause like, I want, uh, I wanted O-one to be good, even when it wasn't, at least like the first time I tried it, it wasn't good, but I wanted it to be good. 6 times in the scene
- ▶ 3:05 Alessio Fanelli Are there for either of you, any use cases that you were not able to do at all in previous models that you didn't even try at first in a one, and then you found out it was actually good. 6 times in the scene
- ▶ 6:58 Dan McAteer Yeah, using O-one was, it was the first time where I would connect it to my IDE. 3 times in the scene
- ▶ 8:15 Alessio Fanelli So the, the anatomy of a one prompt.
- ▶ 12:27 Ben Hillock I think that what makes a one even trickier than other models is that, um, there, there is actually an asymmetric miss to how well open AI understands the model and how well we, for example, the fact that like reasoning tokens are hidden,…
- ▶ 14:41 Alessio Fanelli Like I should not be using a one for this, or are you just trying to upgrade most of the tasks and workflows to, to be a one compatible, so to speak? 6 times in the scene
- ▶ 18:04 Shawn Wang Like we, we use, uh, I think, uh, Ben, you and I talked a little bit about how we use O-one for AI news. 4 times in the scene
- ▶ 21:01 Ben Hillock You know, how, how good all one is and how long it takes. 4 times in the scene
- ▶ 26:21 Ben Hillock But as far as just like, a lot of times I find it getting stuck in sort of a loop that I don't see in quad.com, and so I haven't seen that for, I haven't tried it too much for O.N. specifically, uh, but, uh, that was just sort of an… 4 times in the scene
- ▶ 29:07 Dan McAteer So we'll use the Ben's template, the structure of that, um, like O one prompt, and it will generate for me just based on my like high level thoughts, my brainstorm, like a fleshed out prompt that follows that structure.
- ▶ 29:47 Ben Hillock I haven't seen a huge difference trying it either way with a one, like, and in some parts of the documentation, opening, I recommends putting it after for prompt caching. 2 times in the scene
Beating Google at Search with Neural PageRank and $5M of H200s — with Will Bryk of Exa.ai
- ▶ 13:15 Will Bryk I think it's very similar to O-one, by the way. 2 times in the scene
- ▶ 40:30 unnamed speaker Like I kind of, I think Owen is also very full self-driving.
- ▶ 44:54 unnamed speaker We never, we never talk about AGI, but you had, uh, this whole tweet about O-one being the biggest kind of like AI step function towards it. 6 times in the scene
The State of Reasoning — from Nathan Lambert, Interconnects/AI2 [LS Live @ NeurIPS 2024]
- ▶ 4:11 Nathan Lambert So this is like one of the many ways that we can kind of lead towards a one is that language models have randomness built into them and a lot of what people see as failures in. 2 times in the scene
- ▶ 5:19 Nathan Lambert What is O-one is, has been a large debate since its release. 6 times in the scene
- ▶ 8:53 Nathan Lambert I would say that this is a hard pivot in the talk where oh, one is. 3 times in the scene
2024 Year in Review: The Big Scaling Debate, the Four Wars of AI, Top Themes and the Rise of Agents
- ▶ 15:16 Shawn Wang I think that the gap is widening, uh, with O-one. 3 times in the scene
- ▶ 35:29 Shawn Wang Oh, one for AI news for a while. 6 times in the scene
- ▶ 1:05:52 Shawn Wang I was looking at the OpenAI livestream today when they introduced O-one API with structured output and everything. 4 times in the scene
- ▶ 1:20:30 Shawn Wang Um, you want to get beyond 1300, you have to pay up for the O-ones of the world, and the four O's of the world, and the Gemini 1.5 pros of the world.
- ▶ 1:23:23 Shawn Wang It's actually, we got O-one and O-three. 2 times in the scene
- ▶ 1:33:24 Alessio Fanelli And then September, we got a one. 4 times in the scene
Best of 2024: Synthetic Data / Smol Models, Loubna Ben Allal, HuggingFace [LS Live! @ NeurIPS 2024]
- ▶ 26:34 Loubna Ben Allal For example, Queen 2.5 math, everyone's trying to reproduce a one.
2024 in Post-Transformer Architectures: State Space Models, RWKV [Latent Space LIVE! @ NeurIPS 2024]
- ▶ 1:40 Dan Fu So this is one of the, the, this is the iconic image from the open AIO one release.
- ▶ 39:02 Eugene Cheah But, but, but then putting it back to another paradigm, right, is that I think O-one style reasoning
Best of 2024: Open Models [LS LIVE! at NeurIPS 2024]
- ▶ 12:11 Luca Soldani Open replication of what OpenAI's O-one is, um, you're gonna be on the 10 K spectrum of our GPUs.
0 to over $8M ARR in 2 months as a Claude Wrapper (Bolt.new, Qodo)
- ▶ 42:14 Itamar Friedman By the way, even O-one, which is supposed to be able to do system two thinking, like Greg from OpenAI, like, hinted, is doing better on these kind of problems, but still, it's very useful to break it down for O-one, despite 4 times in the scene
- ▶ 45:24 Itamar Friedman And then, and even the new results of, uh, all one, we, we published it.
- ▶ 58:31 Itamar Friedman And now, now with O One, people are talking about System One. 2 times in the scene
Why Compound AI + Open Source will beat Closed AI — with Lin Qiao, CEO of Fireworks AI
- ▶ 29:31 Lin Qiao Declarative system is going to be appear as a model that has extremely high quality, and this model is inspired by O-one announcement from OpenAI.
- ▶ 31:02 Shawn Wang When OpenAI released the one, a lot of people asked about 4 times in the scene
Agents @ Work: Lindy.ai (with live demo!)
- ▶ 34:59 Florent Crivello I think we're seeing it a little bit right now with O-one. 5 times in the scene
- ▶ 39:27 unnamed speaker O-one doesn't have API use today.
Agents @ Work: Dust.tt — with Stanislas Polu
- ▶ 42:11 Stanislas Polu We'll see what comes out of, uh, the O-one class, if it ever gets function calling.
In the Arena: How LMSys changed LLM Benchmarking Forever
- ▶ 25:56 Shawn Wang But I just wanted to briefly focus on O-One. 7 times in the scene
Singapore: the AI Engineer Nation — with Minister Josephine Teo
- ▶ 37:34 Shawn Wang That is the top of the week, top of the town this week because of OpenAI's O-One model.
[Paper Club] SWE-Bench [OpenAI Verified/Multimodal] + MLE-Bench with Jesse Hu
Production AI Engineering starts with Evals
- ▶ 1:27:35 Ankur Goyal Because OpenAI is so good at making their models so available, I think they get a lot of credit for the science behind, you know, oh, one, and wow, it's like an amazing new model. 3 times in the scene
- ▶ 1:31:29 unnamed speaker I mean, speaking of O-One, I mean, let's go there. 4 times in the scene
Building AGI in Real Time (OpenAI Dev Day 2024)
- ▶ 0:06 unnamed speaker Until dev daylights, code ignites Real-time voice streams reach new heights O-one and GPT-FOR-O in flight Fine-tune the future,
- ▶ 6:16 NotebookLM Host 2 And speaking of powerful tools, they also talked about their new O-one model. 6 times in the scene
- ▶ 20:43 Shawn Wang Is it because O-one, October first? 3 times in the scene
- ▶ 36:46 Olivier Godement And so, um, let's take all one, for instance, like, is the one small enough, like, for your problems? 3 times in the scene
- ▶ 39:51 Alessio Fanelli You had kind of like ChatGPT app to get the plan with a one, and then you had, um, Cursor to do apply some of the changes. 4 times in the scene
- ▶ 50:19 Shawn Wang I've converted to Cursor for, and O-one is so easy to just toggle on, on and off.
- ▶ 1:04:47 Shawn Wang I was, I was hoping for like an all one native thing in assistance.
- ▶ 1:07:24 unnamed speaker Now that OpenAI's O-One preview is announced, it is incredible to see the OpenAI team also obscure their chain of thought traces for competitive reasons, and still perform lower than Cozine's Genie model. 16 times in the scene
- ▶ 1:22:38 unnamed speaker Agents and AI employees beyond level three, and his projections of the intelligence of O-one, O-two, and O-three models in future. 4 times in the scene
- ▶ 1:37:06 Kevin Weil This set of models, O-one in particular, and all of its successors are going to be what makes this possible, because you finally have the ability to reason, to take hard problems, break them into simpler problems, and act on them. 7 times in the scene
- ▶ 1:53:28 Sam Altman Who feels like they, they spend a lot of time with O.I. 3 times in the scene
[Paper Club] 🍓 On Reasoning: Q-STaR and Friends!
- ▶ 19:33 unnamed speaker So, you know, I think the relevance here for, um, for O-one is that if we were to generate reasoning traces, we would have to do work like this, um, where the rationale would have to be exposed, um, into, into step-by-step thinking, and,…
- ▶ 25:23 unnamed speaker Yeah, this is a method, uh, that is general, and, uh, I, I, I would be very surprised if they did not use this for, uh, for O-one.