DeepSeek, every mention
95 scenes, the whole family · ← back to DeepSeek
tap a year for its mentions
every year anyone Guido Appenzeller 41Martin Casado 21Marco Mascorro 20Joel de la Garza 20AI Security Researcher 18Olivia Moore 16Marc Andreessen 12Anish Acharya 11Amjad Masad 7Steven Sinofsky 5
Verbatim, from the transcripts: the passages where DeepSeek comes up
Why AI Demand Is Outrunning Compute Supply
- ▶ 1:10:11 Gavin Baker And if you look at the three big, you know, Chinese open source models, DeepSeq, Kimi, Quinn, they're kind of
Inside a16z’s Top 100 AI Apps Report with Olivia Moore
- ▶ 12:14 Olivia Moore So they're mostly using, like, Daobao, which is made by ByteDance, DeepSeek, Quen, Kimi, those kinds of models. 3 times in the scene
Marc Andreessen's 2026 Outlook: AI Timelines, US vs. China, and The Price of AI
- ▶ 28:05 Marc Andreessen And so the, you know, the Chinese are in the game for sure, you know, with software, um, you know, Deep Seek, you know, was kind of the big, you know, kind of, uh, fired the starting gun of the software race. 9 times in the scene
- ▶ 55:55 Marc Andreessen The deep seek, uh, this, the deep seek moment I think was in January or February of this year.
Amjad Masad & Adam D’Angelo: How Far Are We From AGI?
- ▶ 47:11 Amjad Masad Um, and the main idea there was that if we put a verify on the loop, I remember reading DeepSeq, uh, a paper from Nvidia about how they, um, used DeepSeq to write CUDA kernels and they were able to run DeepSeq for like 20 minutes. 3 times in the scene
- ▶ 55:51 Amjad Masad One thing is maybe kind of random, but like I get really fired up to see like mad science experiments, like the, uh, deep seek OCR that came out the other day.
Sacks, Andreessen & Horowitz: How America Wins the AI Race Against China
- ▶ 37:51 David Sacks Uh, it could just be, it could just be kind of a historical accident in the fact that the DeepSeq founder was, like, very committed to open source and kind of, just kind of that, like, got things started that way, or it could be part of a…
- ▶ 50:08 David Sacks And, um, you know, China, China's out there promoting, um, you know, uh, deep seek models and Huawei chips, and they're not like wringing their hands about
- ▶ 1:02:25 David Sacks And, you know, now this, I think this narrative really started to fall apart, uh, with the launch of Deep Seek, which really happened in the first, I don't know, couple of weeks of the Trump administration, because, you know, if you asked… 2 times in the scene
Marc Andreessen and Ben Horowitz on the State of AI
- ▶ 32:40 Marc Andreessen Number one, um, the, the, the researcher engineer side of things, it is striking, it is striking to the degree to which, um, there are excellent, you know, outstanding models coming out of China now, um, and, you know, and, and, and, and,…
- ▶ 35:09 Marc Andreessen and I think the current, I think the current, if you just observe currently, specifically like Deep Sea, Quinn and Kimi, and these models coming out of China, I, my, my sense basically is like, I would say the U.S. specifically in the West…
Marc Andreessen & Amjad Masad on “Good Enough” AI, AGI, and the End of Coding
- ▶ 19:12 Amjad Masad So what Nvidia did is they're trying to, uh, write, um, GPU kernels, uh, using DeepSeq. 3 times in the scene
Sam Altman on Sora, Energy, and Building an AI Empire
- ▶ 32:58 Ben Horowitz And what do you think, like, strategically, like, what's the danger of DeepSeq being the dominant open source model?
Dylan Patel on the AI Chip Race - NVIDIA, Intel & the US Government vs. China
- ▶ 5:41 unnamed speaker And this is kind of coupled with DeepSeq saying their next models are going to be on domestically produced Chinese chips, the Chinese government, uh, kind of banning
- ▶ 1:23:50 Dylan Patel Uh, if you're running DeepSeq inference, the performance difference per GPU is like north of like six, seven X, and it continues to optimize, um, you know, for, for DeepSeq inference. 2 times in the scene
The Little Tech Agenda for AI
- ▶ 19:45 Erik Torenberg And I think, I think the early concern on the open source was that we would be somehow giving it to China, but then we've seen with DeepSeek, et cetera, that they just have it anyways.
The Top 100 Most Used AI Apps in 2025
- ▶ 8:04 Olivia Moore So these are products like Deep Seek.
Dylan Patel on GPT-5’s Router Moment, GPUs vs TPUs, Monetization
- ▶ 31:39 Dylan Patel Um, so you can, you know, create the maximum efficiency and that turns out, oh, look at deep seek or, you know, go look at what the labs are doing. 2 times in the scene
Google DeepMind Lead Researchers on Genie 3 & the Future of World-Building
- ▶ 13:15 unnamed speaker Between Gini one, two, like, you know, like in, for example, NLMs, like you have, like Dipsy card one, like they saw in this paper, like the longer they keep it running, they suddenly will see like these interesting behaviors.
The Current Reality of American AI Policy: From ‘Pause AI’ to ‘Build’
- ▶ 9:28 Anjney Midha Yeah, so, so, so no empirical evidence at the time that this was, that AI was creating net new marginal risks, and just factual inaccuracies that we were ahead of China, because if you just paid attention to what's happening, DeepSeek had…
- ▶ 9:48 Anjney Midha They're not that, they're not years behind, and so when R-I, DeepSeek R 2 times in the scene
- ▶ 15:04 Erik Torenberg What, um, is it really just deep seek? 3 times in the scene
- ▶ 26:43 Anjney Midha But in the long term, if every other major nation is running their entire AI ecosystem on the back of American chips and American models and American post-training pipelines and American RL techniques, then that ecosystem win is orders of…
Steven Sinofsky & Balaji Srinivasan on the Future of M&A, AI & Tech
- ▶ 1:05:23 Balaji Srinivasan You've got Kimi, you've got Quinn, you've got deep seek. 3 times in the scene
Balaji Srinivasan: How AI Will Change Politics, War, and Money
- ▶ 6:09 Balaji Srinivasan It's a big thing, like, all the DeepSeq models.
Sovereign AI: Why Nations Are Building Their Own Models
- ▶ 11:25 Anish Acharya A stable equilibrium is certainly not one where we just turn our back on a bunch of allies, because China's definitely has, has enough of the compute resources to try to export great models like DeepSeek to the rest of the world. 2 times in the scene
- ▶ 14:36 Anish Acharya Just a year before DeepSea came out, you had a number of, like, tech leaders in Washington testifying that China was, like, five to six years behind the US, with confidence, on the record. 2 times in the scene
Building AI Systems You Can Trust
- ▶ 23:06 Scott Clark It's like, we're going to let you use OpenAI, but we're not going to let you use hosted DeepSeq or something like that.
What Comes After Mobile? Meta’s Andrew Bosworth on AI and Consumer Tech
- ▶ 26:12 Andrew Bosworth And we've seen this already with DeepSeq in China, which was put in a tough spot and then innovated incredibly in the memory architectures and a couple other places to, to really get amazing results.
Can the Military Move at Startup Speed? How the Army and Navy Are Rebuilding
- ▶ 29:21 Alex Miller Um, I had a chance last week to sit with all of the big LLM providers, and I, I loved the conversation because they were ruthlessly honest with, with me and with each other in that, hey, all of these things are available, whether it's…
The Top 100 GenAI Products, Ranked and Explained
- ▶ 4:30 Olivia Moore And then of course the deep seek launch earlier this year.
- ▶ 4:32 Anish Acharya Deep seek was so interesting because I think it, it sort of had become settled wisdom that it would be very hard for a horizontal model to get to mass consumer scale quickly again. 6 times in the scene
- ▶ 6:54 Olivia Moore Maybe people thought that was a solved problem, but then again, Deepsea came in and upended all of that.
- ▶ 12:31 Olivia Moore But maybe they got just edged out by, like, Deep Seek launching this time, and so they, they lost their spot for this ranking, but might be on there the next one, and they still have massive usage.
- ▶ 23:34 Olivia Moore We only had 10 days of data for Deep Seek for January, because it launched at the end of the month, and it shot up from literally nothing to number two on the list, you know, 10% of ChatGPT scale, uh, on web within a week, a little bit… 9 times in the scene
- ▶ 25:24 Anish Acharya Next time we should add a different cut for DeepSeq USA.
The Future of Digital Workers
- ▶ 11:06 Joe Schmidt You mentioned how many different model providers there's the deep seek news, you know, a month ago.
Jensen Huang & Arthur Mensch: Why Every Nation Needs Its Own AI Strategy
- ▶ 34:17 Arthur Mensch Uh, and then we had Chinese company, uh, like DeepSeq release, uh, stronger models, and everybody benefit from it.
Building the Next Generation of Conversational AI
- ▶ 18:37 Anjney Midha Now you got R one, Gwen, so on.
- ▶ 52:05 Ankit Kumar You know, in the RL world, like DeepSeq and so on, there's sort of a, you know, you can get a lot out of verifiable rewards, where you can get the model to generate something, you can check whether it's right in some relatively stable way.
DeepSeek, Reasoning Models, and the Future of LLMs
- ▶ 0:05 Guido Appenzeller So, DeepSeek R-One last month really captured the zeitgeist. 2 times in the scene
- ▶ 0:19 Marco Mascorro So, one of the good things about DeepSeek is basically they open source their weights, their techniques, and how they build these models, and they've been doing that for a while.
- ▶ 0:49 Marco Mascorro So here's our breakdown on Deep Seek. 2 times in the scene
- ▶ 0:51 Guido Appenzeller Some of the models they've been building, like Deep Seek Math, B-III and B-II, and now R-I. 2 times in the scene
- ▶ 0:51 Guido Appenzeller Some of the models they've been building, like Deep Seek Math, B-III and B-II, and now R-I.
- ▶ 4:51 Guido Appenzeller And we look specifically here at DeepSeq R-one. 6 times in the scene
- ▶ 4:53 Guido Appenzeller That's not necessarily because they were the first, we're not sure they are, but they currently have probably been the most open about the methodology. 6 times in the scene
- ▶ 7:05 Marco Mascorro And, and the idea, so, so DeepSafe came with B three in December, and then they, they released, uh, just a few weeks later, uh, R one.
- ▶ 7:05 Marco Mascorro And, and the idea, so, so DeepSafe came with B three in December, and then they, they released, uh, just a few weeks later, uh, R one. 3 times in the scene
- ▶ 7:05 Marco Mascorro And, and the idea, so, so DeepSafe came with B three in December, and then they, they released, uh, just a few weeks later, uh, R one. 2 times in the scene
- ▶ 8:27 Guido Appenzeller Let's talk about DeepSeq v-three. 5 times in the scene
- ▶ 8:33 Guido Appenzeller It's a classic LLM, although I think they took a little bit of reasoning from a precursor of DeepSeq R-One and applied it back. 2 times in the scene
- ▶ 8:33 Guido Appenzeller It's a classic LLM, although I think they took a little bit of reasoning from a precursor of DeepSeq R-One and applied it back. 3 times in the scene
- ▶ 9:49 Guido Appenzeller So here's a rough flowchart to see what the different steps are for, for training R one. 5 times in the scene
- ▶ 10:09 Guido Appenzeller To Deep Seek V three, so that classic model, and figuring out ways how they can leverage the, the better reasoning capabilities in Deep Seek R one zero, but at the same time, make that model better behaved and easier for humans to consume. 3 times in the scene
- ▶ 13:10 Marco Mascorro Some of these traces actually in R-one, when they decided to basically fix these R-one zero models, like, how do we improve now, or how do we make it usable for humans, right? 2 times in the scene
- ▶ 13:52 Marco Mascorro Yeah, this is, this is the aha moment, and the DIPSEC team just called it the aha moment in the model, but also the aha moment for them, at least.
- ▶ 14:42 Marco Mascorro And the way they did it is they took P. 4 times in the scene
- ▶ 17:56 Guido Appenzeller And, well, it's interesting, they said they spent, I think, five and a half million dollars on, on training V-three. 3 times in the scene
- ▶ 18:01 Guido Appenzeller So going from V-three to R-one is probably pretty cheap, is my guess, right?
- ▶ 19:55 Guido Appenzeller So DeepSeq also did a number of other interesting innovations as part of how they built the model.
- ▶ 25:06 Marco Mascorro Another, another interesting thing that, that came out of the paper from R-one was for these model models, the LAMAs, the seven Bs, the seven Bs, the 13 Bs. 4 times in the scene
- ▶ 26:11 Guido Appenzeller You know, there's a lot of controversy around the, you know, deep seek.
How to spot an AI Deepfake
- ▶ 1:11 Joel de la Garza The problem we've been talking about is the, the deep seek model came out, set the world on fire. 4 times in the scene
- ▶ 2:48 Brian C. Long When you see models like DeepSea come out, I think as you, you allude to, you know, you bring the ability for attackers to use those models in
Avoiding vulnerabilities in AI code
- ▶ 0:57 Joel de la Garza We had the release of deep seek open source model reasoning model two weeks ago, and it seems like another chat GPT moment where this crazy thing drops from the sky and everyone's kind of running off and doing interesting things.
- ▶ 11:12 AI Security Researcher If you go to Deep Seek,
How to use DeepSeek safely
- ▶ 0:17 AI Security Researcher If I had to deploy DeepSeq, I would probably focus on use cases that were not end user facing because DeepSeq is like, especially susceptible to basic jailbreaks. 2 times in the scene
- ▶ 0:42 Joel de la Garza You know, a lot of, a lot of the news of the last two weeks has been deep seek and sort of these new reasoning models that have been open source coming from China. 3 times in the scene
- ▶ 3:02 AI Security Researcher Um, for Deep Seek specifically, there was the, the, the part that limited speech about, uh, politically sensitive topics in China. 2 times in the scene
- ▶ 4:25 AI Security Researcher The short answer there is that DeepSeq has these very hard limits on things like politically sensitive speech. 6 times in the scene
- ▶ 5:22 Joel de la Garza And so this is sort of the start of that process for the deep seek folks. 2 times in the scene
- ▶ 6:09 Joel de la Garza And maybe, maybe it helps for folks to understand sort of like the, the way most people were interacting with DeepSeq was through a hosted model that was in China. 5 times in the scene
- ▶ 9:15 Joel de la Garza How do they compare to the controls that you see on like deep sea? 5 times in the scene
- ▶ 13:42 AI Security Researcher What I've been telling most people who ask is like, let's just wait a few weeks and there will be an open source model that, that implements this reinforcement learning technique and you, you'll, you'll get great reasoning on par with,… 7 times in the scene
Reasoning Models Are Remaking Professional Services
- ▶ 1:56 George Sivulka At the same time, you're starting to see models like, kind of, oh one, oh three, uh, doing this, like, reasoning at inference, scaling at a deep seek as well. 3 times in the scene
- ▶ 3:06 George Sivulka And, ah, you know, regardless of the truth with DeepSeek, I generally also think it's, it's really an American technology that's cheaper.
- ▶ 12:27 Alex Immerman And then with Hebbia, I think if you were to compare it to chat GPT as you just did, but many of the most commonly used AI applications today, perplexity, Claude, maybe deep seek recently, they're all chat interfaces.
What DeepSeek Means For The Future Of AI | Tech Veterans Weigh In
- ▶ 0:24 unnamed speaker I don't know about you guys, my Twitter feed, all the podcasts, everything, DeepSeek, everywhere. 3 times in the scene
- ▶ 2:43 Martin Casado You have, like, all this excitement about O-one and how that's gonna drive compute costs and NVIDIA, and then, you know, R-one comes out, and it looks pretty good, and then all of a sudden they're saying, well, you know, 2 times in the scene
- ▶ 3:52 unnamed speaker And so, we're, like, 10 days or so out from this release, which, by the way, as both of you said, it was, that was the R-one release.
- ▶ 3:59 unnamed speaker There was the V-three release, what, two months ago or so, which was the base model. 2 times in the scene
- ▶ 4:17 Martin Casado DeepSeq has actually released a number of soda models, including V three, which is actually probably a more impressive feed is it's almost like a chat GPD four and oh, by the way, to create, uh, one of these, you know, chain of thought… 3 times in the scene
- ▶ 7:49 Martin Casado So I've heard, I've heard two theories of why they were able to do this. 6 times in the scene
- ▶ 9:47 Martin Casado Yeah, so there's, there's two things that are pretty remarkable about DeepSeq R-One that have implications on adoption. 2 times in the scene
- ▶ 11:31 Steven Sinofsky And I think because of where they came from and that they don't have a business model, that was part of what was unique about this was it was a hedge fund, like almost like this side project, but not really a side project, just a whole… 2 times in the scene
- ▶ 14:50 Martin Casado And in that case, DeepSeek is less relevant because they're not building apps. 2 times in the scene
- ▶ 18:59 Martin Casado And to bring it back to DeepSeq, I mean, that means like, this is very important to realize that, you know, the capabilities of China, it's a very credible player, but I don't think that R-One itself as a standalone is going to have that…
- ▶ 18:59 Martin Casado And to bring it back to DeepSeq, I mean, that means like, this is very important to realize that, you know, the capabilities of China, it's a very credible player, but I don't think that R-One itself as a standalone is going to have that…
- ▶ 27:51 Martin Casado And I want to tie this back to Deep Seek, because, um, I mean, the reason we're getting so excited about this is because we've actually seen these types of things before, and we've seen things like Deep Seek come out before, and it's not… 3 times in the scene
- ▶ 35:25 unnamed speaker But I do want to touch on, you know, a lot of people are talking about Deep Seek as the Sputnik moment. 3 times in the scene
- ▶ 37:41 Steven Sinofsky And so what gave us this deep seek moment was the strength of the worldwide technology community. 3 times in the scene
- ▶ 41:37 Martin Casado So again, like there's this view of deep seek as a crisis moment for Nvidia crisis moment for open AI and anthropic.
The Quest for Community-Trained Open Source AI Models
- ▶ 12:27 Jeff Schmidt It is now used by every, every model you use nowadays, everything, everything Chachipiti, Lama, DeepSeq, all of them, they all use the, the methods that we pioneered in, in the YARN paper.
- ▶ 25:07 Jeff Schmidt OpenA, Anthropic, Meta, X, Google, and then you have a few Mistral, and then Deep Seek, um, and a couple of other places that are maybe like a specific country are like backing it, but certainly we would not, we could name them all and not…