XOutlier
Download .zipAdd to Chrome

Noam Brown

@polynoamial181k followers

Researching reasoning @OpenAI | Co-created Libratus/Pluribus superhuman poker AIs, CICERO Diplomacy AI, and OpenAI o-series 🍓 reasoning models

Open on X
Top outlier
105×
Best organic post
Peak views
3M
Single post
Bookmarks
34k
Across 91 organic posts
Last scan
Oct 16
Community cache
Posts 91 posts
#DatePost↓ OutlierViewsLikesRepostsBookmarks
1Sep 12, '24Today, I’m excited to share with you all the fruit of our effort at @OpenAI to create AI models capable of truly general reasoning: OpenAI's new o1 model series! (aka 🍓) Let me ex105×3M11k1.5k4.6k
2Apr 25, '25It's deeply concerning that one of the best AI researchers I've worked with, @kaicathyc, was denied a U.S. green card today. A Canadian who's lived and contributed here for 12 year90.7×2.6M8.6k711962
3Dec 20, '24We announced @OpenAI o1 just 3 months ago. Today, we announced o3. We have every reason to believe this trajectory will continue.65.7×1.9M4.2k420720
5Jun 28, '25You don’t need a PhD to be a great AI researcher. Even @OpenAI’s Chief Research Officer doesn’t have a PhD.45.7×1.3M3.4k191831
6Jun 9Implications of Large-Scale Test-Time Compute tl;dr: As LLMs become more capable, benchmark performance is increasingly a function of test-time compute. In fact, we likely don't kn43×1.2M3.4k4283.3k
7Jul 19, '25Today, we at @OpenAI achieved a milestone that many considered years away: gold medal-level performance on the 2025 IMO with a general reasoning LLM—under the same time limits as h40.5×1.2M4.6k5011.5k
8Jul 6, '23I’m thrilled to share that I've joined @OpenAI! 🚀 For years I’ve researched AI self-play and reasoning in games like Poker and Diplomacy. I’ll now investigate how to make these me32.5×923k2.1k130654
9Jan 21, '25This is on the scale of the Apollo Program and Manhattan Project when measured as a fraction of GDP. This kind of investment only happens when the science is carefully vetted and p32.3×918k7.3k6581.2k
10Nov 9, '24I've heard people claim that Sam is just drumming up hype, but from what I've seen everything he's saying matches the ~median view of @OpenAI researchers on the ground.29.7×843k1.7k176562
11Nov 20, '23Everyone at @OpenAI is united. This is not a civil war. Unless Sam and Greg are brought back, there will be no OpenAI left to govern.20.6×585k1.6k107109
13Sep 28, '24Those of us at @OpenAI working on o1/🍓 find it strange to hear outsiders claim that OpenAI has deprioritized research. I promise you all, it's the opposite.19×541k3k150305
14Jul 21, '25Congrats to the GDM team on their IMO result! I think their parallel success highlights how fast AI progress is. Their approach was a bit different than ours, but I think that show17.7×503k2k155396
15Jul 19, '25I think it's safe to say this @OpenAI IMO gold result came as a bit of a surprise to folks16.2×462k2.6k137341
16Jan 19, '25It can be hard to “feel the AGI” until you see an AI surpass top humans in a domain you care deeply about. Competitive coders will feel it within a couple years. Paul is early but 15.5×440k2.4k260664
17Apr 16, '25We did not “solve math”. For example, our models are still not great at writing proofs. o3 and o4-mini are nowhere close to getting International Mathematics Olympiad gold medals.15×426k3k160351
18Sep 19, '24.@OpenAI is hiring ML engineers for a new multi-agent research team! We view multi-agent as a path to even better AI reasoning. Prior multi-agent experience isn't needed. If you'd 14.7×417k1.7k178895
19Jun 21, '25When this happened in 2014 almost everyone in AI thought it was outrageously overpriced13×370k3.7k107185
20Dec 5, '24My teammates and I at @OpenAI are excited to finally share the full o1 model (aka 🍓) with you all. It can do a little better than just counting how many r’s are in “strawberry”:12.7×362k2k150274
21Oct 21, '25Self play works so well in chess, go, and poker because those games are two-player zero-sum. That simplifies a lot of problems. The real world is messier, which is why we haven’t s12.3×351k97565280
22Oct 2, '25My new hobby is asking GPT-5 Thinking to find errors in every @Wikipedia page I read. Interestingly, almost every page I checked has at least one error. 🧵12×341k1.7k129333
23Jul 9, '24When I joined @OpenAI a year ago, I feared ChatGPT's success might shift focus from long-term research to incremental product tweaks. But it quickly became clear that wasn't the ca11.8×334k1.4k65214
24Sep 23, '24o1-preview is pretty good at planning11.7×333k82596317
25Sep 28, '25Julian was co-first author on AlphaGo, AlphaZero, and MuZero. He doesn't have a major twitter presence, but he's been at the forefront of AI exponential progress for more than a de11.5×326k1.7k86580
26Apr 14, '25I worked in quant trading for a year after undergrad, but didn't want my lifetime contribution to humanity to be making equity markets marginally more efficient. Taking a paycut to11.4×324k2.7k99530
27Oct 21, '25Below is a deep dive into why self play works for two-player zero-sum (2p0s) games like Go/Poker/Starcraft but is so much harder to use in "real world" domains. tl;dr: self play co11.3×321k1.6k1641.5k
28Dec 18, '24I think it's safe to say LLMs can reason10.9×310k1k94198
29Dec 5, '24Sadly, tic-tac-toe remains unconquered by LLMs.10.6×302k1.5k39183
30Apr 10, '25Memory isn't just another product feature. It signals a shift from episodic interactions (think a call center) to evolving ones (more like a colleague or friend). Still a lot of re8.8×249k1.5k125300
31Dec 20, '24This is the result I'm most excited about. Even if LLMs are dumb in some ways, saturating evals like @EpochAIResearch's Frontier Math would suggest AI is surpassing top human intel8.4×240k1.2k110201
32Dec 6, '24o1 has been out for less than 24 hours and people are already disappointed with it because it can’t generate infinite money and solve climate change8.3×236k1.5k59134
33Mar 25, '25This isn't quite true. Test-time compute helps when verification is easier than generation (e.g., sudoku), but if the task is "When was George Washington born?" and you don't know,7.9×223k1.4k111372
34Sep 12, '25When we at @OpenAI released o1-preview a year ago, it would think for seconds. Today, our best reasoning models can think for hours, browse the web, and write code. But there's a l7.8×223k1.6k130207
35Nov 10, '24I love seeing a new eval with such low pass rates for frontier models. It feels like waking up to a fresh blanket of snow outside, completely untouched.7.5×213k1.4k64217
36Apr 14, '25Our latest @OpenAI model, GPT-4.1, achieves 55% on SWE-Bench Verified *without being a reasoning model*. @michpokrass and team did an amazing job on this! (New reasoning models com7.4×211k1.2k93171
37May 3, '25I recently made this plot for a talk I gave on AI progress and it helped me appreciate how quickly AI models are improving. I know there's still a lot of benchmarks where progress 7.3×206k1.1k122315
40Jul 2, '25AI researchers will literally negotiate $100 million comp packages by themselves but they won’t play poker for more than $50 buy-ins6.9×197k2.2k48164
41Jul 25, '245 years ago we revealed Pluribus, the first superhuman multiplayer poker AI. It cost only $150 to train. Why did poker take longer than Go? And how did it end up being so cheap? Th6.9×197k1k110427
42Dec 21, '24Practice working with AI. Human+AI will be superior to human or AI alone for the foreseeable future. Those who can work most effectively with AI will be the most highly valued.6.9×195k1.8k213541
43Nov 22, '223 years ago my teammates and I set out toward a goal that seemed like science fiction: to build an AI that could strategically outnegotiate humans *in natural language* in Diplomac6.9×04.1k599986
44Mar 12, '25Seeing these creative writing outputs has been a real "feel the AGI" moment for some folks at @OpenAI. The pessimist line lately has been “only stuff like code and math will keep g6.5×184k1k84171
45Aug 17, '25I recently chatted with a VC who believed AGI was coming and would disrupt a lot of jobs, but not *their* job. Of course AIs could write code and review contracts. But making accur6.4×183k1.3k69338
46Feb 9, '25I think a lot of the misinformation about what LLMs can and cannot do comes from the fact that progress in the field is much faster than academia is used to.6.4×183k1.5k84213
48Dec 20, '24To all the AI skeptics out there, this is a good one to tune in for6.2×176k1.4k68127
49Feb 6, '25When we briefed people on 🍓 before o1-preview's release, seeing the CoT live was usually the "aha" moment for them that made it clear this was going to be a big deal. These aren't6.1×174k83946139
50Feb 6, '25There's a lot of talk of LLMs "saturating all the evals" but there's plenty of evals people could make where LLMs would do poorly: -Beat a Zelda game -Make a profit in a prediction6.1×174k1.7k97354
51Nov 11, '24With @OpenAI o1, we developed one way to scale test-time compute, but it isn't the only way and might not be the best way. I'm excited to see academic researchers explore new appro6.0×171k1k94465
52Jul 19, '25Their bet allowed for formal math AI systems (like AlphaProof). In 2022, almost nobody thought an LLM could be IMO gold level by 2025.5.9×169k1k59129
53Aug 11, '25In my opinion, the most important takeaway from this result is that our @OpenAI International Math Olympiad (IMO) gold model is also our best competitive coding model. 🧵5.9×167k1k85191
54Jan 31, '25We at @OpenAI are proud to release o3-mini, including for the FREE tier. On many evals it outperforms o1. We’re shifting the entire cost‑intelligence curve. Model intelligence will5.8×164k1.9k191236
56Mar 29, '24If you haven't disabled voice authentication for your bank account and had a conversation with your family about AI voice impersonation yet, now would be a good time.5.7×162k784122196
57Jul 23, '25It can be hard to “feel the AGI” until you see an AI master a domain you care deeply about. Everyone will have their Lee Sedol moment at a different time.5.7×162k1.3k94203
58Sep 15, '25GPT-5-Codex is 10x faster for the easiest queries, and will think 2x longer for the hardest queries that benefit most from more compute.5.7×161k1.2k90146
59Feb 21, '25A lot of folks seem surprised so I want to point out: the robot in this video is likely teleoperated, not autonomous (would be great if @1x_tech / @ericjang11 could confirm that). 5.6×160k94531137
60Jul 19, '25It’s truly a privilege to be able to wake up every morning, see where the latest intelligence frontier is, and help push it a little further.5.6×159k1.8k7490
61Jun 10, '25Input is now $2 per 1M and Output is now $8 per 1M. The cost vs intelligence curve will continue to improve rapidly.5.4×154k1.2k49118
62Nov 22, '23Most elaborate team building exercise ever5.4×154k1.2k4844
63Aug 28, '25To all undergrads interested in learning about AI: be wary of taking “Intro to AI” as your first AI course. In many programs, the class you actually want first is “Intro to Machine5.3×150k1.2k108523
64Aug 5, '25Our new @OpenAI open models5.2×149k1.4k121153
65Dec 26, '24I think people are overindexing on the @OpenAI o3 ARC-AGI results. There’s a long history in AI of people holding up a benchmark as requiring superintelligence, the benchmark being5.2×148k1.7k98193
67Feb 25, '25There’s a long history of AI doing well at games, but that’s typically involved the AI *training* on that game. What makes this result so cool and significant is that the model was5.1×145k1.7k100224
68Feb 25, '25Deep Research has been a big hit but was previously limited to pro subscribers. Now, all plus users can try it for themselves!4.8×135k1.1k4274
69Aug 24, '25GPT-5 Thinking definitely isn’t perfect, but it’s the first AI model I can trust more than many common sources of truth on the internet.4.7×134k7753983
70Jul 19, '25It takes us a few months to turn the experimental research frontier into a product. But progress is so fast that a few months can mean a big difference in capabilities.4.6×132k92546110
71Jan 17, '25Lots of vague AI hype on social media these days. There are good reasons to be optimistic about further progress, but plenty of unsolved research problems remain.4.6×130k1.3k95152
72Apr 16, '25Our new @OpenAI o3 and o4-mini models further confirm that scaling inference improves intelligence, and that scaling RL shifts up the whole compute vs. intelligence curve. There is4.6×130k1.4k155243
73Feb 27, '25Scaling pretraining and scaling thinking are two different dimensions of improvement. They are complementary, not in competition.4.6×130k1k80123
74Dec 4, '24We just have to ship faster than the goalposts move4.5×128k1.2k5682
75Jun 11, '25I'm fortunate to be able to devote my career to researching AI and building reasoning models like o3 for the world to use. If you want to join us in pushing forward the intelligenc4.1×117k1.2k44156
76Aug 7, '25I'm more optimistic than ever that we at @OpenAI can eliminate hallucinations. There's still more research to be done, but GPT-5 is solid progress.4.0×113k92876141
77Mar 6, '256 years ago AI pioneer and now Turing Award winner @RichardSSutton distilled 75 years of AI into a simple Bitter Lesson: general methods that scale with data and compute ultimately3.9×111k83765312
78Aug 6, '24When I signed with @OpenAI last year, word spread that “Noam is joining OpenAI.” Google, upon hearing this, called @NoamShazeer and tried to persuade him to not go to OpenAI and to3.9×110k4901373
79Sep 12, '24Believe it or not, the name Strawberry does not come from the “How many r’s are in strawberry” meme. We just chose a random word. As far as we know it was a complete coincidence.3.9×110k1.1k4075
80Apr 16, '25This is worth tuning in for!3.8×107k1k4235
81Jul 11, '1912 years ago I tried making my first poker AI in college and dreamed of beating the world's best pros. After seven years of a PhD, I'm excited to announce that I finally did it! It3.8×02.9k53088
82Aug 6, '25Hmm I wonder what this could be…3.7×106k1.1k2737
83Mar 21, '24I’m 36.3.7×104k8741883
84Jan 23, '25Making it a little easier for the whole world to see what we're seeing.3.6×102k1.2k3674
86Aug 3, '25Considering the technology and the pace of progress, I think this is quite sane.3.5×101k94449134
87May 13, '25People often ask me: will reasoning models ever move beyond easily verifiable tasks? I tell them we already have empirical proof that they can, and we released a product around it:3.5×98k1k53180
88Sep 6, '241 engineer working in the right direction beats 100 geniuses working in the wrong direction3.1×89k85871119
89Feb 3, '25o1 was released less than 2 months ago. o3-mini was released 2 days ago. Deep Research was released today. It’s a powerful tool and I can’t wait to see what the world does with it,3.1×88k90476131
90Feb 12, '25This meme summarizes the paper nicely3.1×88k78241188
91May 22, '25There’s an old joke in AI: as soon as machines outperform humans at something, it stops being considered AI. Glad to see poker solvers have reached that point.2.9×81k87635185
93Jul 17, '25Something to watch out for when evaluating tool-using agents: they can "cheat" by browsing the web and simply looking up the answer key. The @OpenAI ChatGPT Agent team had to take 2.6×74k78749109
94Jan 28, '25Algorithmic breakthroughs and scaling are complementary, not in competition. The former bends the performance vs compute curve, while the latter moves further along the curve.2.6×74k94064125
95Sep 17, '2512/12 problems solved, which would be equivalent to a 1st place performance. GPT-5's solutions were responsible for solving 11/12 of them.2.6×73k7666392
96Apr 16, '25Today, we're releasing @OpenAI o3/o4-mini. The eval numbers are SOTA (2700 Elo is among the top 200 competition coders) But what I'm most excited about is the stuff we can't benchm2.2×63k99881103
97Oct 12, '22After building on years of work from MILA, DeepMind, ourselves, and others, our AIs are now expert-human-level in no-press Diplomacy and Hanabi! Unlike Go and Dota, Diplomacy/Hanab2.1×01.3k234284
99Dec 6, '22I've been listening to @lexfridman's podcast for a long time, but it was truly an amazing experience to sit down with him myself and talk about our latest research in multi-agent A1.8×01.4k88220
100Mar 9, '226 years ago today, AlphaGo beat Lee Sedol in a milestone for AI. Typically deep learning gets the credit, but it's important to know that nobody has *ever* trained a raw NN that's 1.7×01.2k184187
101Sep 4, '20Best advice I got when I started grad school: "You either have imposter syndrome, or you're an asshole."1.1×09967631

Scan another creator

@
@karpathy@levelsio@paulg@naval@marclou@illyism

Run this on any account.

XOutlier builds this table for every public profile you open on x.com, with live data and every post, not just the top 50.

Add to Chrome Download .zip