Interconnects

Last week coined the term “Black October“ after an onslaught of critical vulnerability disclosures, including PHP2Shell and a long-standing GitLab bug, signified the moment when the cyber risks of AI models started to materialize. It seemed like the news cycle had stabilized a bit when I started writing this, but then there was news about another major crypto security incident at Binance. The crypto industry has a very long history of hacks, so it’s not the best industry to indicate vulnerabilities translating into new societal harm. Still, we are very much still living through the implications of this transition.The best guess for what is happening in the cybersecurity space is that the behind the scenes work has only accelerated, bringing more people and resources in to try and patch the internet — the Wall Street Journal reported that cyber first responders are strained, which seems like the system working. We’re living through this plot of an exponential rise in vulnerabilities, which will take a long time to level off at the top of the sigmoid.A few weeks ago, before the rapid acceleration in vulnerability disclosures and the uptick in reported hacks, I wrote the following take. It’s been getting a pretty consistent barrage of feedback, some hopeless dunk attempts and some reasonable discussion, so I wanted to unpack where I stand.Open-weight Mythos in April would’ve been fine?When Claude Mythos was announced, it was previewed as if it was a new class of cyber weapon, where if it ended up in the wrong hands it would’ve caused mass societal destabilization.It seems like this was wrong.By all measures, GLM-5.3 is the model that crosses that threshold of capability, and there’s little public evidence that much has changed, and we’re over a month out of the model weight release.In fact, I would go further. If Claude Mythos was accidentally released as open-weight, it seems like the world would have been more or less fine. Yes, there would be a clear increase in cybersecurity incidents and it would be bad for society if the model leaked, but it would be a scenario that looks more like an acceleration than a step change in risk.The leading proponents of this new wave of open-weight fear mongering have effectively been making a falsifiable prediction — that the current open-weight models are going to cause substantial, not seen before AI harms by crippling our cyber infrastructure.Based on all the evidence I have, which I’ve unpacked above, they’re set up to be wrong. At least now, after so many years of debating open weight risks, we will start to get some real answers.To start the reflection, I wanted to do more due diligence and track down some receipts on the claim that the AI safety community was reacting as if open-weight Mythos would be an actual disaster, and not just a painful headache. This was distinctly my impression having lived through the online discourse. I’ve gotten criticism that the frontier AI community never said it would be a “true disaster,” but regardless of exactly what they said, it is always about how people hear and internalize your information.Here are some examples of the sort of language that I felt was too exaggerated and from prominent people, which spilled into the public and more importantly federal government perspective that this was a cyber weapon include:* Rob Wiblin, host of the 80,000 hours podcast, on his own podcast (April 10, 2026): “If it fell into the hands of criminals or hostile state cyber actors, it would be an actual disaster.”* Anthropic, in the Project Glasswing announcement blog (April 7, 2026): “Given the rate of AI progress, it will not be long before such capabilities proliferate, potentially beyond actors who are committed to deploying them safely. The fallout—for economies, public safety, and national security—could be severe.”For one that is actually more reasonable, to illustrate the fine lines between correct and exaggerated, from Zvi Mowshowitz (who I don’t agree with a ton of things on) on April 9, 2026: “Claude Mythos would give attackers a cornucopia of zero-day exploits for essentially all the software on Earth … It would be chaos.” To date, this chaos has been confined to those running large-scale digital products, which is good, but could change. I largely still stand by my above writing a few weeks ago, but I’m nervous about it. I’m not a cyber expert, I’m working hard to make the most educated predictions I can. It’s a fine line between “all of our tech companies need to work extra hard to keep us secure” and “public infrastructure is actually going down.” It is very clear as of writing this that there are way more critical vulnerabilities which are being found in popular tools and fixes are being built.For example, Epoch AI recently published a nice post on their internal work using GLM-5.3 to try and find vulnerabilities and chain them into what could be an attack. The actual exploit they found in Firefox is a good example of why cyber risks need multiple things to align for harm to be done:We used GLM 5.3 to create a website that, simply when visited by a Firefox user and without further interaction, gave the attacker the ability to exfiltrate files from the user’s computer.Here, the exploit is in Firefox, and a would-be hacker needs to plant a website that is looking for this exploit. This is doable, and could have serious consequences, but just finding the exploit is only one step in a multi-stage process.In the previous post, I should’ve been more specific on what “fine” is for a future of cybersecurity, and in this case it means giving more credit to some of the AI safety community that discussed this reasonably as a new era in cyber instability. Reflecting on it, I’m not sure fine is the right word to describe what is inevitably coming – a forced industry update to a new way of working with software, which will clearly be expensive and accompanied by some hard work. If you’re already somber about the challenges coming from AI, I think fine is actually very suitable. If you didn’t price in a period of turbulence, fine undersells. I’ve written many times that I am in the former, difficulty will come, group — but I think humans are good at muddling through problems and patching immediate concerns.What my view was closer to is that open models are providing balance in the offense-defense equation, and that Mythos coming sooner wouldn’t have massively upended this. I suspect most of these companies would be faster to adopt open-weight models than attackers, as most attackers are lower resourced than these companies. A key needle we’re watching the balance on is how willing attackers are to cause physical harm within the U.S. or its allies. For now, while our adversaries have likely used these models and exploits to extract information en-masse in a clandestine way, it is not clear if there is a party who will cause mass harm by pushing the full force of these models at public institutions.I thought this Tweet was a funny way to summarize some of the personnel / motive gaps in hacking:Life is a lot easier once you stop seeing threat actors as hackers who want to break into systems by any means possible, and start seeing them as businesses with well defined business processes. A bakery isn’t going to start making dog food because there’s a dog food shortage. ClickFix and Phishing actors aren’t going to drop everything to vibecode exploits just because a critical CVE came out.The fact that the only actors exploiting this are the actors whose MO is exploiting things... isn’t a coincidence. It’s entirely predictable if you understand threat actor behavior.Interconnects AI is a reader-supported publication. Consider becoming a subscriber.Where I sit, I expect most of these vulnerabilities are being found by closed models, but I have very high error bars on this prediction. A lot hinges on how open models are actually being used in the near term. Today, open models are expensive to serve… This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.interconnects.ai/subscribe

Hill-climbing is easy, knowing what to hill-climb is hard. Making scores go up is expensive, but we have a much clearer playbook for it than we do for creating meaningful benchmarks at the true frontier of AI today.Independent evaluation of language models is under-appreciated in how hard of a task it is. These groups are not just taking current models and certifying their scores, which is a technically involved task. This is their job if AI progress is slow. When AI progress is as fast as it is, the technical details of the job are the routine work part of trying to put a single number on what makes for a good AI model today. The goal is to have a single number that is representative of current models, matching vibe checks, community sentiment, and scientific analysis, while being forward looking enough so that it’s not outdated in 3 months.Everyone building a benchmark faces a sort of paradox. You need your benchmark to reduce to one single number, so that it is easy to discuss, use in reports, and compare models, but every method of aggregating scores comes with meaningful trade-offs. The best scientists need to do this, everyone does. It’s a matter of creating aggregation metrics that you think are most fair, knowing there will always be imperfections.With the amount of money in the AI ecosystem today, the weight of independent evaluation is even higher – you dictate the winners and losers of a trillion dollar market. You also influence some of the models themselves, steering the shape of progress. This is a lot of pressure!My prompt for this article is Artificial Analysis’s recent update to their Intelligence Index, v5. The AA index is the most prominent independent intelligence metric for the latest frontier models, followed in a distant second by Epoch AI’s Capabilities Index (ECI). To date, all of these have been incredibly useful for the community and massively flawed!Interconnects AI is a reader-supported publication. Consider becoming a subscriber.Language models are capable of so much today, and this scope of capabilities is growing quickly. Our era of singular intelligence indices should be phased out soon, but it will take a long time.A summary of AAv5 is below, from their blog post, where they describe it as “reflecting real-world usage today:”Intelligence Index v5 weights:➤ Knowledge Work (25%): AA-Briefcase (15%), AutomationBench-AA (5%), and GDP.pdf (5%)➤ Coding (25%): FrontierCode 2.0 (15%) and Terminal-Bench 4.0 (10%)➤ Science (25%): Terminal-Bench-Science (15%) and CritPt v2 hard subset (10%)➤ General (25%): AA-Omniscience v1.1 15% (10% accuracy, 5% non-hallucination) and HLE-Diamond (10%)Here’s the new ranking of models:The problem with trying to benchmark real-world usage today is that it is extremely jagged, just like the models. The most proven use-case is coding, where open models are far more competitive and likely to keep up. In knowledge work, I’m less confident in open models.At first glance, in aggregate, their model ranking definitely looks reasonable. Still, it shows signs of strain and confusion, as any benchmark will, with GPT 6.1 Terra beating Claude Fable 5.1 on AAv5. Today, some benchmarks measure the absolute frontier of capabilities, playing nicely in open-ended tasks through your favorite harness, which often does take a large model. Some other benchmarks measure certain skills that will be needed of a model in repetitive agentic tasks. I get the feeling that AA is tuning their benchmark more towards the former, a general agentic collaborator, but all of these behaviors overlap.This takes us to the plot going around that prompted me to write this, where they showed the balance of open vs. closed models on the new benchmark. This plot shows 8 closed models beating the best to-be open-weight model, Kimi K3.1 Pro (which isn’t colored blue). This benchmark is taking the opinion of trying to measure where frontier models are moving into, which grows the gap to open models. This is an okay decision to take, but over time as existing workflows of open models become key factors of the economy, it could become a benchmark that misses where the core economic engine of the industry is.Open models’ demise from this seems greatly exaggerated, but I do think the previous index was flattering the open weight models too strongly. Consider this plot from the Kimi K3.1 Pro launch — yes, the open models are strong and useful in some cases, but the best OpenAI and Anthropic models have been a major step ahead in their flexibility and reliability.Some of the evaluations in AAv5 that open models are most behind on are “science” evals like terminal-bench-science and CritPT v2. Open models are behind, but the value proposition was never to compete truly at the frontier. The open model economy will be about* Reducing domains with high market demand for inference to a commodity price of hardware.* Allowing ownership of intelligence, specializing it to internal data, and controlling internal agents across many companies in the economy.Both of these will take a very long time to materialize and are partially represented in benchmarks. The supply of compute is currently so low that open model providers also are raking in solid margins – when this shrinks, we will know the system is starting to work. The same goes for diffusion into enterprises, it’s not clear what the exact ceiling on it is because of the difficult ML expertise that is needed to build evaluations and do post-training today.At the next major model cycle, say in 2-3 months with GLM 6 or Kimi 3.5, I expect the leading open model labs to make substantial steps on this index, again surprising many observers. This is a cycle we have been on for years now.I’ve written many times about how open models keep up with the frontier, and a summary is that:* The best open models do benchmax a little. The worst open models benchmax a lot (but don’t get a ton of attention anyways).* The best open models are a bit narrower in which capabilities they cover, as seen above with their gap on scientific domains. The open model labs are working to address this, but I think it could be come a strategic error to not focus.There’s a more nuanced version of explanations of the performance gap today, where so many of the gains in training are coming from having RL environments with distributions close to the evaluations (a longer analysis on this is here). Keeping up with rapid progress across multiple domains requires a sophisticated RL stack that not all the open model labs may possess. Some may be reliant on buying data that has moderate lead times as well.Another part of this announcement is shifting to “private benchmarks”, where the prompts and test environments are not public, as a sign that open labs are benchmaxxing. This sort of analysis has been done before, and fits with my above theory on how open labs benchmax. It is easier to create training data to support an evaluation when you know the shape of it, but the task isn’t impossible with hidden data. All in, this is a change that directionally I think is defensible, but it’s not unanimous. For example, you’re putting complete trust in the private benchmark builder, as you no longer get the benefit of the community finding issues in the benchmark and pushing the benchmark to improve itself – which happens a lot! There is enough money on the line, where I’d rather see transparency in benchmarking and enabling many eyes to find bugs.The final point I need to make, which does flatter the closed labs, is that the likes of Anthropic and OpenAI moved on from caring about these indices months, or even years, ago. Their focus is to a large majority on supporting internal uses – like AI R&D automation – and the use-cases of customers who are spending billions of dollars. Yes, they have team members hill climbing on other capabilities to unlock new markets, but these behaviors are to a minority extent driven by public benchmarks. At the same time, it is… This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.interconnects.ai/subscribe

We — Nathan and Florian — wanted to discuss all the messy trade-offs of a busy week in AI and ask all the pressing questions like:* Why did Kimi K3.1 Pro have a mixed reaction with such incredible benchmarks? Will it be adopted?* Were we wrong about the risks of open-weight models and cybersecurity?* Why do cyber experts discuss recent events so differently than AI experts?* What comes next in the few months? What will change our minds?We will write more about this subject, but wanted to bring you behind the curtain as we aspire to quickly learn more about the cybersecurity industry and trends across the last few years that have been leading us today. We conclude the podcast with some lighter topics, from Muse-maxing to excitement in still hiring junior technical talent. Please send us questions and topics, as we plan to record more of these podcasts going forward.Listen on Apple Podcasts, Spotify, and wherever you get your podcasts. For other Interconnects podcasts, go here. YouTube ↓Chapters* 00:00 Kimi K3.1 Reaction* 08:15 Kimi’s open-weight cyber principles* 11:23 You can hack everything era* 15:54 PHP2Shell, LiteLLM, and the new internet* 24:00 Abliterated APIs, safeguards, and trusted access* 31:41 Anthropic not hiring juniors employees* 36:16 Florian’s Muse addiction* 45:30 Wrap-upLinks we discuss* My thread with Dean Ball about the “boy who cried wolf” story of the frontier labs* Kimi K3.1 Pro announcement blog* Kimi K3.1 Pro on Artificial Analysis Intelligence Index* Kimi K3.1 on Artificial Analysis defensive cyber benchmarks* Program Bench* Open-weight Mythos would’ve been fine take* PHP “hasn’t caused a disaster yet” update* Abliteration.ai serving abliterated models API* Morph serving abliterated DeepSeek via API* Joshua Saxe’s Substack on AI and cybersecurity* Azeem Azhar, “The internet’s Wuhan moment”* Anthropic, “How employment is changing at Anthropic”* LiteLLM security incident reportTranscript00:00:00 Nathan Lambert: We are back with another Interconnects podcast. We are with Florian, eval guy, recently moved to SF. Florian, is Kimi K3.1 Pro, name is a mouthful, going to live up to the benchmark scores, for example on Artificial Analysis, or where is it going to land?00:00:17 Florian Brand: In the past, the Kimi models always lived up to the expectations, but this release is kind of weird compared to the others. I think it is also partially because they are becoming more professional, and the previous model drops were kind of out of nowhere. This time, which also got some criticism, they launched a bigger social media campaign, which allegedly also involved paying people to post about the model without disclosing.00:00:56 Nathan Lambert: I think a lot of companies do this and tell you not to mention it, but it’s a mess. It’s bad form. No company should do it. Essentially, one person tweeted a screenshot of the press instructions for Kimi that said, “Don’t say this is a paid promotion.” And to my knowledge, no one has corroborated it after that.00:01:15 Florian Brand: Yeah, and there are right now a few of these model launches, including by American companies, that are just very obviously astroturfed. We have seen launches where it’s thousands of retweets and a few hundred likes, because it works so well to get the attention. Some of them disclose that it is a paid ad while others don’t. But in general, we have seen the professionalization of model launches over time, and pretty quickly. A year ago, everyone just dropped the model, then patched vLLM afterwards, so no one was actually able to run any of the open models for a while. Now you get all the inference frameworks on day one, you get all the inference providers on day one. Or, which is also what we see more and more, the model creator has a few weeks of exclusive access before they release the weights. And that’s also what’s happening with K3.1 Pro. I think it’s two weeks of exclusive access before they release the model weights.00:02:38 Nathan Lambert: They said the weights would come out by November 4th. And the headline things... I don’t know if I’m going to be able to share screen. I don’t have this set up on this computer. But the headline is this Artificial Analysis number where it’s ahead of GPT-6 Astra. And then if you look at the Kimi K3.1 blog, they have their headline numbers, which are really strong, and then they also have some numbers later on which seem a bit more jagged. I think it was Terminal-Bench Science, which is 20 points behind GPT-6 Astra. I think there were a few others, like ProgramBench. Do you know Terminal-Bench Science and ProgramBench? Are any of these things that people think are important benchmarks for frontier models? Or how do you explain the fact that on Artificial Analysis the model is top, what, three, four? It’s the two latest Claude 5.5 models, Muse Wonder Full, and then Kimi K3.1, which looks crazy. It just looks like totally crazy scores, and I think everybody is still surprised by it. But then there’s this long tail of details that you use to evaluate a model now. I had posted a thread on Twitter of, “What are people’s impressions?” And there’s a lot of people that are like, “This seems not great,” and then some people that are like, “Yeah, it’s fine.” I don’t know, there’s a mixed bag. Relative to K3, which I think was pretty unanimously, this feels like a big model and the post-training seemed a little bit undercooked for that. Maybe this is just a tiny bit RL fried. We’ll see where we land.00:04:26 Florian Brand: For the benchmarks, one of the things is that the benchmarks that are right now used by Artificial Analysis, they will post their new version probably very soon after we release this podcast.00:04:43 Nathan Lambert: We can’t tell you the exact date, but both of us know.00:04:48 Florian Brand: In general, the Artificial Analysis benchmarks are rather short in terms of time horizon, whereas something like ProgramBench is really long and also very explorative. Other benchmarks are telling the model, “All right, this is the goal, and I want you to reach it,” whereas ProgramBench is more like, “All right, here’s what I want, but you need to explore a lot.” And I think that might be the reason that we see some differences: their RL training is still more towards the exploitative side of model training versus the explorative side, where you need to find the solution somehow before you actually know what you’re doing.00:05:41 Nathan Lambert: But if you’re using a three trillion parameter model, I feel like you kind of want it to be pretty robust at exploration. Because if you’re going to want to use a model for pretty known coding tasks, like using DeepSeek-V4.1-Flash-1017, which is not that far behind on these scores, it’s just going to be way faster. It might not be wildly different on cost per task with the actual token numbers. That’s why I think it’s pretty reasonable criticism of this Kimi release: a big model has got to be... Because it’s going to be slower, it’s got to be like Fable, which is just... I still think Fable is really nice if you’re doing very blurry things. I revert to it when checking my slides over sometimes, where I just feel like it kind of gets me a bit better, where Opus is really, really grindy. I think the open-weight models are better in this kind of grindy perspective, because you can just get them to be faster and cheaper.00:06:44 Nathan Lambert: It seems like Kimi is gearing up for this Flash-style release with their naming change. Previous Kimi models were just K2, K2.6, K3, and now this is K3.1 Pro. So a lot of the company forecasting, investment-y side of interest on Twitter was, a lot hinges on this for Kimi, because it’s kind of true. We’re not the inference experts, but serving the model seems like you’ve got to commit way more resources to serve this three trillion versus this 700 billion parameter model. You’re going to need NVL72 racks, I’m guessing, to do it well.00:07:21 Florian Brand: To fit the model, you need B300s, one node per … This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.interconnects.ai/subscribe

This has been a week where the balance of American open-weight models took a step change back toward the frontier. Meta took the lead of this charge with Muse Spark Open, which is now the strongest American-made open model by quite some margin, but the addition of Reflection’s first model Beam is also relevant. Without these two models, there would be 20+ open-weight models built and released by Chinese companies better than the best Western models. At the time of writing this, Mistral has also announced a strong model, Mistral Large 4, with an index score of 38 (see below), but its weights are not yet released. There are rumors of a few more strong American open-weight models coming from other organizations too!In many ways, the gap between the best American and Chinese models has been reduced roughly by half. We’ve gone from the best American open models being largely irrelevant to consumers who require the cutting edge of performance, to them being behind but relevant.This paints the picture of the cut-throat, competitive nature of the frontier model business. Reflection and Mistral announced their model scores and Meta dropped the weights of a smarter model before these other two labs could get their own weights out. All of these models are competing with Chinese models released months ago, where the industry is collectively asking — where have Kimi and Zhipu AI been since the releases of GLM-5.3 and Kimi K3 on Aug. 18th and Jul. 16th, respectively?That is a very nuanced question, which I’ll dig into another time. Today’s piece is focused on Meta’s turnaround, what Muse Spark Open means for the ecosystem, and the balance of power across Western frontier labs.Why Muse Spark Open should take off — and why it may notOne thing we see again and again at Interconnects through our open model analysis is that Western open-weight models get astonishingly more adoption than Chinese models with the same performance-to-size footprint (even though the Chinese models usually come out first)! Muse Spark Open is landing into this environment, at the same time as when open-ended LLM agents are finally real (~GLM 5.2 capabilities and on). Riding off a rapid increase in performance of Chinese models, the open ecosystem has been living through an acceleration in demand. The inference providers are rapidly scaling, the open post-training economy is forming, and countless enterprises are looking to own their intelligence. Muse Spark Open’s performance falls somewhere above GLM 5.2, near DeepSeek V4.1 Flash, and behind GLM 5.3 or Kimi K3. It should be squarely “good enough” to slot right into this.For some background, from the Muse Spark Open model card on Hugging Face:Muse Spark Open is an 816B total/42B active parameter open weights model with a context window of 1M tokens. Built for coding and agentic work, Muse Spark Open is natively multimodal. This open weights version has additional safety post-training to make it more robust against misuse as a standalone model without system guardrails. We recommend deploying Muse Spark Open as part of a broader AI system, with additional guardrails suited to the specific use case.Here’s the distribution of the various Muse models on the Artificial Analysis Intelligence Index.This is included to loop back to the original model card snippet. It was shared that this open-weight model underwent additional safety training to prepare for this open release. Many observers saw this release as potentially a model that was “broken” during the safety training, as it came out after Muse Spark 1.3. In reality, this was much more likely a safety tune of Muse Spark 1.2, which would only be a 1 point drop on Artificial Analysis to shore up safety behaviors. When Meta released Muse Glimmer, the 30B agentic model, Alex Wang said that they were preparing to release the Muse Spark 1.2 model as open weights, which fits with the above.If anything will hold Muse Spark Open back, it’ll be the specific license they designed for it. In summary, this custom license has pretty much all of the restrictive license behaviors we’ve seen (and criticized here) in the last few years of building with open models. They are as follows:* Users of the models in products must display “Built with Muse” on their products. Meta did this with Llama, and Kimi has tried it, but almost no one complies.* There is a revenue-gating factor, where companies that make more than $20M per month across their entire business deploying the model need a separate license with Meta. Other prominent open model providers are doing this too, in order to set up revenue share agreements for re-selling inference.* Muse Spark Open makes you agree to an acceptable use policy, which in the true “open source software” corner of the world makes this model extremely not open-source.* All of the above applies to outputs from the model as well, and derivatives, so there is legal ambiguity if you can safely use the model for synthetic data.The license does permit using Muse Spark Open for internal research and applications.We’re at a weird equilibrium where this license is complex and unproven in court, but many players will likely ignore it. Many, many people will build on this model from various inference providers, and Meta will only likely go after egregious violations. For now, licenses haven’t been the determining factor of AI adoption, but this could change in the future. This is why we think it’s an unnecessary strategic risk for labs trying to compete and regain adoption to have complex, non-standard open licenses for their models.Interconnects AI is a reader-supported publication. Consider becoming a subscriber.Is Meta a tier 1 (OpenAI/Anthropic) or tier 2 (Gemini/xAI) frontier lab?Just a few days before Muse Spark Open, Meta unveiled their closed frontier models, Muse Wonder and Muse Wonder Full. On traditional benchmarks, these are very strong models only behind Anthropic’s Opus 5.5 and similar to OpenAI’s GPT-6-Astra or Google’s Gemini-4-Argon.These major releases are coming just over a year out from Meta’s massive AI redirection, characterized by CEO Mark Zuckerberg’s Personal Superintelligence post, which got a lot of criticism at the time. Meta is the only American company with frontier open and closed models. Alibaba’s Qwen has followed a similar strategy, but with a far more open-leaning approach.The vibe checks across all of these models have been solid, but the expectations were for them to not immediately challenge the level of refinement of the most recent Anthropic and OpenAI models in real-world usage. I’m excited to try them more — time to sign up for another AI subscription.I like this recent run of form from Meta, driving their own products and directions in the popular Muse agent, taking wins like the leading US open-weight crown when useful. Whether or not it’s financially successful, it’s a unique perspective that fits their history as a business.It often feels like there’s a revolving door of which lab is nipping at the heels of the Anthropic-OpenAI leading duo. It’s not clear that in the long run Gemini and xAI are out of it, but those two are much more muddled in their offerings than Meta today. If you listen to Zuck and Alexandr Wang on The Cheeky Pint, they clearly had a very specific vision of a high-ownership, high-compute per researcher lab that I would expect to pay off. Many labs die due to layers of bureaucracy and misaligned incentives. Meanwhile Alex Wang has over 200 direct reports — i.e. it’s a very flat org with minimal management — and they’re releasing very consistent model and product improvements.Much of my worldview of building models is that you need the resources, the compute, the data, and the talent (all downstream of large sums of money), but organizing them is very hard. Gemini is sometimes described as a place where it’s impossible to get any work done or grow — which is embellishing the issue, as many people love it there, but there are kernels of … This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.interconnects.ai/subscribe

I’ve had one too many discussions on the cyber risks of open models — where someone assumes that either banning open models happens in a silo (and bad actors will somehow actually be stopped) or that China is a threat and doesn’t care about AI safety — that I feel we’re going to end up making policy decisions that both limit American AI competitiveness and increases long-term cyber risk. I feel like we’re locked in on a path of lose-lose situation, unless people embrace nuance, trade-offs, and scary realities.Much has been written about the views of various players in this debate, which I summarize as follows:* Open weight models pose an untenable risk to society’s functionality, occupied by frontier lab leadership and the national security community in the U.S.* Western voices saying open weight models are necessary for defense and banning them will make the world less safe, occupied by AI risk moderates to different extremes. I include myself in this bucket, along with Hugging Face’s perspective after the OpenAI incident, and others like Joshua Saxe.* Chinese companies continuing to release open-weight models with strong cyber capabilities, who are making decisions based on the risk assessments of their society and government.This list is also organized by volume or clarity of message in the Western AI ecosystem. The risk-focused discussions have been disproportionately visible, including pieces of work which I view as very detrimental to potential of engaging across assumed party lines on this issue. There’s also very little substantive engagement trying to understand how Chinese companies assess risk — rather, mostly arguments that border on ad hominem attacks like “Chinese labs don’t care about safety.” Interconnects AI is a reader-supported publication. Consider becoming a subscriber.The delusions of the anti open-weight allianceThe most recent publication from the “open-weights are dangerous” crowd was the report by Anthropic on the risks of GLM-5.3 as an offensive cyber tool. The problem with this is not the technical research they did, which is largely reasonable, but the failure to engage on more cross cutting questions like: What happens if we ban open models due to cyber risks or why do Chinese companies deem these models safe to release?These questions, which I’ll come back to, are the ones we need to answer to understand the ecosystem perspective on current AI risks. All policy actions happen in a multi-piece puzzle, where the shape of risk can be changed, but it is rare that you get a universally better outcome. They are all trade-offs.This sort of solipsistic writing by Anthropic fits very closely to another type of discourse I’m hearing about regularly that is determining the fate of open models and AI broadly — classified briefings. I’ve had many discussions with people who say something like “Look, I’m pro open models, but if you’re seeing what I’m seeing, it’s an onslaught out there due to [XYZ Chinese open-weight model] and we need to act.” This perspective flies in the face of public information, where to date closed models have been documented as the cause of most existing cyber attacks. As someone who tries to make grounded predictions, the evidence of numerous attacks from OpenAI models is the only data I have on the shape of cyber risks (or, you can broaden the scope to include all of FelonyBench). There are a few possible explanations, and it’ll take years to get to the true answer: * Maybe open model weights and closed model APIs (with some safe guards) are both far closer to being easy to mis-use, rather than API models being closer to safe. The trope "Open Dangerous, Closed Safe" may be closer to "Open Unsafe, Closed Unsafe."* Maybe there are far fewer bad actors who are willing to use AI models for “loud” cyber attacks that target crucial infrastructure in the US or powerful and underprepared institutions. Closed models have stronger capabilities and stronger safeguards (in theory), but the stronger capabilities part could matter more in net harm if the safeguards on both are porous. * I have a lot more to learn about how the cybersecurity ecosystem works, so I won’t support this full-heartedly as one of my current recommendations, but I think it’s at least a reasonable argument that open weight models as a diffusion tool for cyber defense could be the best tool we have to prevent harm over the next few years. There are many domains, e.g. sensitive government agencies, where open-weight models are the only tools that can be deployed in the near-term on air-gapped networks.Lots of this situation seems like the messy reality being hard to pin down, where the logical limits are clearer. I may agree with the simple reality that closed models are far safer in the limit, as a technical tool with more layers of protection, but they may actually be causing more harm in the near term because of how accessible model APIs are. Still, much of the impact of the tools comes down to how they’re used and controlled. A duopoly on such a crucial tool may be the defining, risky part. This is the debate we’re having, and over time, as intelligence diffuses, many things may change.These together make me come to the conclusion that if you think the latest open-weight models need to be banned to slow the diffusion of cyber risks, you probably also need to make public-facing APIs for the frontier closed-models illegal. The current stack of safeguards on closed models is stronger than open-weight models, but far from perfect. It is very likely that cyber capabilities of closed models increase much faster than guardrail performance, and a world where open models are banned while closed models continue to progress would be increasing the offense-defense cyber gap. It will take meaningful time to build mitigations that allow defenders to use strong AI models on their private infrastructure if we remove access to strong open-weight models.Explaining China’s AI risk postureChina cares about AI safety. They have come to this discourse in a complete, endogenous manner relating to Chinese culture and power structures. There are some big picture pieces of this puzzle that reflect in other ways, such as Chinese society’s techno-optimism that extends into AI and the Chinese government’s complete focus on political stability. These things do, of course, interface with emerging AI risks like cybersecurity.They are in their own AI risk dynamic, which hasn’t risen to the prominence of the political dance between the White House and the frontier labs which we have followed in the U.S. for a few months. My understanding is that the Chinese companies have to register every major model release with the Chinese government. This includes evaluations, which originated around information control for banned government information. It is unclear if this framework has expanded meaningfully into any risks like cyber or bio. The Chinese government is very decentralized and information from the labs needs to go through existing structures to make it to the top.At the same time, there is a much stronger air of social risk in China, where leading AI researchers aren’t allowed to leave the country and the industry is closing off to foreign investment. On balance I would guess the personal risks facing people by unleashing any domestic harms would be much higher than in the U.S., where the worst case is that your company gets deleted.Overall, the political oversight seems much earlier than in the U.S. The Western AI companies’ core principle is close to “the best way to stop a bad guy with AI is to be a good guy with AI.” In China, these companies are much more practical about building a wonderful tool, and in many cases building a tool that complements their existing businesses.I do think that the companies consider the global implications of their work. These labs also clearly are incentivized, and often pushed via the government (at least in public messaging) to compete as much as t… This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.interconnects.ai/subscribe

This episode is with Jean-Stanislas “JS” Denain of Epoch AI, who leads their Insights Team and is one of the people I find myself debating the state and trajectory of AI with more and more. We’ve had follow-on discussions of many of my favorite recent posts online and/or in private, so I wanted to dig into the nuance in a public episode.A big takeaway of this podcast is how JS and I both have so much uncertainty with exactly where we are heading, and this was our best effort at stating our observations today. Chapters / topics include:* 00:00 Predictions for RSI* 18:15 The role of robotics in an AI acceleration* 24:20 How far behind are Chinese models?* 27:39 Does distillation explain the gap?* 40:58 What Chinese job postings reveal about their labs* 48:13 Are open or closed models safer?* 58:10 How Epoch AI ticks* 1:00:55 What a frontier post-training recipe looks likeEnjoy!More from JS: Epoch AI profile and writing, X, LinkedInListen on Apple Podcasts, Spotify, and where ever you get your podcasts. For other Interconnects interviews, go here.Transcript00:00:00 Nathan Lambert: I’m here with JS Denain, who is a senior researcher at Epoch AI. He leads the insights team. He is one of the people who I feel like I get the best feedback on my writing from, whether it’s from US-China AI capabilities, now RSI. And I just wanted to open this discussion and honestly go deeper with him, trying to understand how he thinks about these various things. And I think you have a very useful, moderate point of view, which I feel like you’re probably a step further into what would be called faster scenarios for AI progress. But let’s get into this, and it’s like, what measurements do you think OpenAI and Anthropic are seeing when we get all these proclamations on RSI happening very imminently?00:00:47 JS Denain: Yeah. I think, so there’s the measurements they’ve published, right? So, OpenAI and Anthropic both had blog posts, I mean, Anthropic two at least, on the effect AI has on accelerating AI progress. I think at least the things they publish, I don’t think are super strong evidence of imminent self-sustaining acceleration AI capabilities, or full automation of the job of AI researcher. But I think the kinds of things we see are, I think probably the most striking thing I saw in the OpenAI blog post was increasing usage of AI systems in model deployment, like the increase in spending on Codex that we saw. And it’s kind of unclear how exactly to interpret this, because maybe it’s a measurement artifact where they’re only looking at Codex, but in fact, there was a bunch of ChatGPT usage before from the researchers. But overall, that plot, for example, just shows a 2X a month increase in Codex spending by researchers, and that does seem to me to be some evidence of they’re getting a lot of value out of this probably. I don’t think this is strong evidence that in six months we have a software intelligence explosion.00:01:57 Nathan Lambert: Do you think this is the same? So, what is the information they have internally relative to what we have? And this is obviously hypothetical. We don’t have this internal information. Because I get the sense that a lot of people are more scared in their updates from the labs than the information we have. And I try to take this very seriously of, what will they be seeing that is making the acceleration of risk comments go faster, and how much of this is material evidence versus how cultures evolve over time? And I’m much more interested in evidence.00:02:29 JS Denain: Yeah. So two things. I think, first of all, I guess I don’t think, I don’t know, right? I don’t have full information here. I don’t currently think that either there’s some specific thing that people at OpenAI or Anthropic are seeing right now that we don’t have access to that warrants being way more freaked out about this. I also don’t think that... I think the public evidence we have right now, more general on AI progress and just a priori case for this being an important dynamic, I think is enough. I think to care about this particular dynamic of AI accelerating AI progress, that being a big deal and worth tracking. And then it’s kind of unclear what the urgency is of when the feedback loop really kicks in. So, basically on the what is there on the inside that people have access to, I could give examples of kinds of metrics, right, they could be looking at. It’s plausible that we have access to the capabilities of AI systems, but internal teams have their KPIs, and maybe they’re seeing compute multipliers in the pre-training team or other kinds of metrics that people are tracking going crazy. And then the combination of this plus some intuitions of how the different outputs of different teams combine yields a prediction on the trend in actual performance of the end AI systems. So that could be an early warning sign. It’s unclear to me that the recent discourse we’ve seen is evidence of things going crazy on those metrics.00:04:09 Nathan Lambert: And how do you think of the link between RSI and existential risk? So I would posit that you agree. I think that there are very real risks of AI, and I’m curious on how you think these, what I would describe as very, very early measurements change anything on the scope of risk. Because I don’t think if you had asked people six months ago, it would be as immediate to x-risk among people are very reasonable. I think there’s more people that are reasonable talking about x-risk again, which was a little surprising to me.00:04:43 JS Denain: Yeah. So, okay, my sense is something like... So personally, I feel very uncertain about this, but I do feel, yeah, basically bought into there’s, I know Evan Hubinger was like, at least 10% of x-risk within I don’t know what timeframe. I think I’m like, yeah, I know, and this seems pretty reasonable over a decade-long timeframe. I just feel extremely uncertain about it, but I’m definitely very worried about this. Now, why am I worried about this, and where do I think the disagreements come from? And then how do I relate this to the early sense of RSI? My sense, I’m kind of a capabilities theory of everything person. I think, and I think some people disagree here, but I really think that principal component of disagreement between everyone is how huge do the capabilities get, how soon, of AI systems? And I sort of agree that there’s other factors that come in play for how big your economic growth gets, also depend on the diffusion you get. And you could possibly you could think that capabilities are going to get crazy, but the AI system’s going to be just aligned and benign and stuff, and so there’s no huge risk. But my sense is concretely, when I look at the main kinds of disagreements between people, most of the people who I see who are very skeptical of those most extreme scenarios... I think just expect capabilities to not be as huge or as I think folks who are—00:06:18 Nathan Lambert: What does being a capabilities maximalist look like in a few years? Because I think I’m probably on the skeptical side, so please continue.00:06:28 JS Denain: I think it looks, for example, something like the AI 2027 scenario, right? I think it looks like the mechanism for this is AI is automating the AI research process, I think, and that’s a reason to pay attention to it. But in terms of effect on the real world, I think it’s like massive progress on robotics. I think a huge industrial explosion, AI systems are just managing factories. You have this kind of self-sustaining economy that just is able to make a large scientific progress much faster than you would have expected. And so I think concretely, the kinds of disagreements I would expect are on, yeah, if you have AI systems that are both very intelligent in the book smart sense, but also have been trained to have more affordances and use them astutely, have been trained to kind of manage projects in efficient ways and stuff like that. How big are the real-world bottlen… This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.interconnects.ai/subscribe

I was recently invited to brief a group of Congressional members and staff on the state of open-weight models in the lens of U.S.-China competition. I’m sharing my prepared remarks as a state of the union on open models that is accessible to a broader audience.Interconnects AI is a reader-supported publication. Consider becoming a subscriber.Recap: What is an open source v. open-weight vs. closed model?Open language models are AI models where their weights are publicly available for inspection or downstream use. These are most often contrasted to so-called “closed” AI models. Closed models offer access only through Application Programming Interfaces (APIs) that developers can use to directly query a model, like GPT-4 or Claude Opus 4.5, or through products, like ChatGPT and Claude Code.Open language models primarily are bucketed into two categories, open-weight and open-source models. Open-weight models are the most common form, such as popular models like Meta’s Llama, Alibaba’s Qwen, Google’s Gemma, or DeepSeek’s models. These models are governed by licenses, governing documents dictating what is allowed with downstream use, and are often accompanied by inference code in libraries such as Transformers, VLLM, SGLANG, etc. Since about April 2025, Chinese AI companies have been the clear leader in open-weight models.True “open-source” models are similar to these, as they include the weights, licenses, and inference code, but they also include the complete information needed to reproduce the model – the training code and training data. The most prominent open-source models have been built in the United States, led recently by the Allen Institute for AI’s Olmo models that I helped build in my recent 2.5 years there. The other prominent open-source models are also built by American non-profit organizations, including OpenAthena’s Marin models and EleutherAI’s Pythia models.Open-weight, open-source, and every other label for a model – including closed models primarily offered via an API – exist on a spectrum. For example, Nvidia’s Nemotron models are far more open than most open-weight models, releasing large quantities of their training data under permissive licenses, but they’re not fully open-source because they do not release all of the data. Closed models also exist on a spectrum based on what information the API reveals and the terms of use.The state of competition between American and Chinese open-weight models (unit economics, technical capabilities, etc.)We are living in a world where GLM-5.2 and Kimi K3, some of the latest, leading Chinese models, have enacted a step change in the commercial viability of open models — crossing a similar threshold in agentic capabilities that Anthropic’s Claude Code crossed in December of 2025.America was the early leader in open language models, primarily through Meta’s Llama models, which were used extensively across research and commercial tasks. Chinese open-weight models surpassed American open-weight models in these two key areas about 18 months ago. The simple metric showing this is Hugging Face Downloads, where China took the lead in July of 2025 primarily through the success of Alibaba’s Qwen models. I personally maintain tools to track this data, and since I first published the American Truly Open Models (ATOM) Project in August of 2025, China’s download lead has grown to about 1.6B – with a total of 3.2B downloads, twice that of America’s total.On popular capabilities benchmarks, such as the Artificial Analysis Intelligence Index (AAII), the Chinese open-weight models have a clear lead over American counterparts. The top three Chinese models as of writing this on September 14, 2026 are Z.ai’s GLM-5.3 and GLM-5.3-Flash and Moonshot AI’s Kimi K3 with scores of 45, 42, and 44 respectively. By comparison, the leading American models are Thinking Machines’ Inkling and Inkling Small, both with a score of 26, and Nvidia’s Nemotron 3 Ultra, with a score of 23. The top American models were released in June and July of 2026, and are updated less frequently than their Chinese counterparts. For example, Chinese labs released models with scores above these American models 2-6 months before the American companies got there (e.g. GLM-5 or DeepSeek V4 Pro). There is a trend of more American companies releasing models, including names like Arcee AI, Poolside and IBM, but they are not rapidly closing this performance gap. Other benchmarks tell a similar story.Together, Chinese open-weight models are approximately 2-5 months behind the closed American frontier, with the open-weight American models being approximately 6-9 months behind the likes of OpenAI and Anthropic. The Chinese labs are closest in tasks with clear user demand, such as agentic coding, and further behind on more open-ended scientific tasks, such as physics or biology.The reasons why Chinese labs can produce these strong models, despite having fewer resources than American counterparts, is still an open debate and heavily influenced by different work cultures, but is also influenced by a few key technical factors. The Chinese labs release their models faster and focus on a slightly narrower distribution of tasks, flattering them slightly on public benchmarks. Releasing faster helps them score higher because all the labs are making consistent progress, so once you “finish” a model to be released, it is a snapshot of performance at that given time — labs where that time is later tend to score higher. Still, the models built by the Chinese labs are genuinely strong and represent real competition to the American industry. This competition will not decrease meaningfully as the closed labs patch vulnerabilities in their API offerings which enable distillation.Distillation is most impactful in new domains and does not make it trivial to create a universally strong final model. I estimate that if distillation was fully prevented, e.g. with know-your-customer (KYC) tools at Anthropic and OpenAI, the gap from the strongest American models to Chinese open-weight models would only increase by 1-2 months.For example, the Chinese labs are rapidly changing their posture towards paying for training data in 2026. Earlier in the year, the top Chinese labs including Moonshot AI and Z.ai had a strong preference towards building data workflows in-house, but by the summer they had begun to buy the cutting edge data – challenging RL environments for agentic tasks – from both established American companies and new Chinese startups.With the advance of open weight models in China towards the frontier of capabilities, and the recent documentation of growing risks around frontier models in areas such as cybersecurity (e.g. the OpenAI-HuggingFace incident), there’s growing regulatory uncertainty on how continued releases can enable a safer ecosystem?A structural challenge in open-weight models is that there are few effective methods for stopping pieces of open software from reaching bad actors. If an attempt was made to restrict access to the strongest open-weight models from China because they amplify risks, the parties who would be set back are American businesses. We have an example of this – HuggingFace used a Chinese open-weight model to understand the cyberattack because closed models would not answer their requests. Thus, managing the risks of open-weight models often comes down to ecosystem preparation.Open-weight models are becoming an essential tool for AI diffusion, and the best path to get ahead of these risks and unbalanced relationships where American companies rely on models built in China is to continue to enable investment in open models in the US. Ownership of open models allows better coordination and preparation of risks that are global in their nature while accelerating diffusion of AI services throughout the domestic economy.The state of open model adoption:… This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.interconnects.ai/subscribe

We’re in an era where a few organizations are using thousands of concurrent agents to improve their processes and output. These organizations happen to be just the frontier AI labs, in particular OpenAI and Anthropic. In the last few weeks, I’ve been pondering what it means for so many employees across these organizations to rapidly update their expectations for the pace of AI progress and associated risks.A core perspective I have is that the frontier labs and broader frenetic, competitive culture in the San Francisco AI scene set up an environment that amplifies any AI concern. This has some benefits in causing more general audience awareness of AI, as fear sells, but exaggerating risk timelines or severity will have negative second-order effects. I remember many loud AI safety debates, and their associated clouds over the viability of open-source AI, in 2023 and 2024 — the primary risks then did not arrive in the forecasted timelines.The general populace of these two key labs was very anxious about AI risks and the rate of progress even a year ago, and especially as agents got stronger product-market fit at the start of 2026. This cultural precondition, when exposed to the reality that thousands of agents will constantly be working fairly productively in your business, will only increase this anxiety. The step from this anxiety, and incidents like OpenAI-HuggingFace, to extinction risks feels very religious.Richard Ngo had an apt summary of the situation:Now a large proportion of the AI safety community is implicitly or explicitly orienting to futures where an intelligence explosion occurs within a few years. My default expectation (absent an extensive pause) is that a similar thing will happen: they’ll turn out to be directionally correct (relative to the expectations of almost anyone not linked to the community) but factually wrong. Specifically, we won’t have superintelligence within the next 8 years, but things will still be moving so fast that it’ll *feel* like the people who argued for short timelines were right.… I wanted to say something now because it feels like the level of bandwagoning towards “singularity soon” is getting pretty wild.Personally, I think this view aligns closely to what I outlined in my alternate scenario to true recursive self-improvement (RSI), which I called lossy self-improvement. A summary of this view is that:* Automatable research is too narrow to achieve a massive net acceleration in progress, in the face of scaling laws’ exponential costs,* Diminishing returns of more AI agents in parallel are real, &* Resource bottlenecks and politics are a major factor in building strong LLMs (and AI can do much less to accelerate this).So, I’m left balancing the above, latent increase in the cultural temperature with the potential that the labs have seen genuinely scary, specific breakthroughs that are not public yet. My expectation is that more of the current AI safety concern is on the former – scaled agents working – but I hold high levels of uncertainty here. Foundational, imagination-based AI breakthroughs are the sort of thing that would make me update my RSI timelines from closer to a tool to sustain progress in the face of exponential costs (scaling laws), to something more unpredictable and/or unstable.Interconnects AI is a reader-supported publication. Consider becoming a subscriber.Some of the best recent resources on RSI have been Dwarkesh’s podcasts with Noam Brown and the trio of John Schulman, Beren Millidge and Charlie O’Neill. I have a few important reflections from both of them.First, the podcast with Noam Brown made me internalize how big of a short-term acceleration mass inference capacity is. These labs will throw thousands of agents at important, measurable problems. At the same time, compute capacity available to them is going to continue to scale. I have my doubts that the labs can afford to spend a constant portion of this compute on internal R&D as the total volume goes up, especially with plans to IPO, as they face increased scrutiny on basic economics. It is important to not confuse massive steps in inference-time scaling, a dynamic which should be fairly predictable, with being the outputs of RSI, which is highly uncertain.Second, the trio podcast debating the state of the art in technical capacities induced more of a surprising reaction that I haven’t fully settled. Through the first hour or so of this podcast, where they debate the role of RL, distillation, scaling, inference-time compute, etc., I found myself strongly agreeing with the distribution of claims. A TLDR would be that our current techniques work and let us solve problems we know how to state, but they don’t result in a magical level of generalization to unknown, harder problems in most partially verifiable domains (i.e. progress in math is an exception, rather than a rule).The surprise of this podcast was the end, where they were predicting timelines for various thresholds of AI. I had GPT-6-Astra summarize the answers provided to three questions from Dwarkesh, of the form “when will AI reach X ability”:All timelines are relative to the interview date.* Drop-in remote worker for broad white-collar work over a month* Charlie O’Neill: ~1 year with programmatic access to workplace tools; ~2 years if it must operate through a browser. Means ordinary white-collar work, not highly creative research.* Beren Millidge: ~3 years for full generality; 80–90% coverage sooner. Main uncertainties: online learning and the long tail of tasks.* John Schulman: ~1 year for an “okay” version, with uneven capabilities that improve over time.* 10× productivity uplift for AI researchers* Charlie O’Neill: 5–10 years. Bottleneck: absorbing information and deciding which experiment to run next.* Beren Millidge: Finds John’s ~2-year estimate plausible, but gives no independent timeline. Assumes AI can run successive experiments and learn from feedback; other bottlenecks would remain.* John Schulman: ~2 years.* AI surpassing top human experts across all computer-based work, including multiyear projects (“ASI”)* Charlie O’Neill: 5–10 years. Highlights limitations in memory and context length.* Beren Millidge: ~5 years for areas labs focus on; potentially longer for literally every domain. Gives no firm timeline for the universal version.* John Schulman: 3–4 years. Spatial/physical fields may take longer; requires onboarding and solving longer-horizon learning.Roughly, a recurring problem when discussing RSI is a lack of specification in intelligence. The jaggedness of intelligence means that we need to discuss thresholds in specific, measurable tasks. The nature of LLMs’ intelligence is shaped very differently than humans, and the roles we forecast are human-shaped. AIs, therefore, do not cross these thresholds like remote worker or AI researcher discretely. It’s a slow diffusion, and a form of long tail will always exist.Take the case of productivity of AI researchers. Many people under-index how much of science is communication and standard setting with colleagues. I do buy the cycle of experiment design and testing being 10x faster in the near future, but not hypothesis generation and intuition building. Accelerating understanding will be the key bottleneck – and it is one that despite all of the AI tools getting massively improved, humans will only improve marginally in their capability. A big improvement in the nature of science will be enabling humans to invest more time here, not them becoming exponentially better at it.This links back to the Noam podcast. Agent swarms in the near future will be effective at solving clear, open problems with verifiable answers. In this vein, when it comes to improving AI models, RSI is much more helpful at efficiency rather than expanding peak intelligence. This is due to the fact that LLM serving has clear metrics you want to improve that are measurable and malleable. This’ll enable better inference-time scaling and more efficient multi-agent sys… This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.interconnects.ai/subscribe

As AI became more powerful, it was inevitable that a different, growing group would start to take AI safety more seriously – what we did not know ahead of time, is which set of views they latched onto. We have seen that some of the most extreme views of risk, i.e. moderate probabilities of mass extinction, were the ones that reached the masses. A lot in the AI world is about to change due to this.How did we get here? Why did this quitting announcement reach so far? In many ways, the rest of the world’s views around AI in the past was a dampening factor. You can think about this like the damp ground around a fire. Many people were striking matches for years about AI risk – they’d smolder in their community and largely burn out, going unnoticed. As the stakes of AI have risen this year, from the OpenAI-HuggingFace incident and breakthroughs like the Navier-Stokes result (also from OpenAI), the ground has dried out and the latent energy around the AI discourse has increased. More people not in the industry have thought, “huh, maybe I should care about this AI thing.” The ambient temperature and stakes have been obviously rising.Then, some basic factors of human nature apply, with the most crucial being that fear sells. Fear is the simplest story, the one people cannot look away from. Jacob Coxon was the one who stumbled into this new powder keg, totally unaware of what was going to come. What looked like a fairly innocuous event – another AI researcher quitting citing safety risks – landed into a very different environment and it caught like wildfire. The discussion of existential risk, mass extinction, and the trajectory of AI has traveled further than even the most seasoned AI commentariat would ever predict.There are a set of facts we need to get clear, which paint the picture of the situation. The key Tweets to reference are from Jacob Coxon, the resignation thread, and Evan Hubinger, the source of the >10% extinction risk figure .* There are plenty of AI risks which are likely to cause harm, even if estimating annihilation is useless. It is important to weigh these with respect to the benefits. The entire discourse around existential risk is on very poor footing. At least Evan was clear in his post, with “kill all humans,” but a major problem in the AI Safety discourse is that people talk about existential risks, when they mean very different things (much like how AGI is a vaguely meaningless term). I put the probability of complete extinction as being so low it isn’t worth discussing, but the probabilities of AI caused disasters – e.g. cyber attacks on critical infrastructure or bio-risks – as being worth debating. Throwing this whole discussion out because there are not these disasters yet is a harmful reaction.* Jacob Coxon is acting genuinely and with good intentions. The outpouring of support from more well-established AI researchers who know of him and his intentions of resignation is useful. Many factions of AI turned to scapegoating him individually, based on account metadata, personal factors, etc. These are not useful. Many frontier lab employees genuinely have similar views to him. I’m not sure it’s a majority, but there is a substantial group.* Many frontier lab employees, especially at Anthropic, are out of touch and this will impact their forecasting and/or descriptions of current AI events. I say this without blaming individuals, but it’s a common agreement among my friends not at OpenAI/Anthropic (Ant especially) that people at the labs operate with a religious energy. It’s very common to go through very out of touch interactions with them. I do not blame most of the individuals who get distorted views being part of these companies, but the interactions are wild and spill over into a lot of wack discussions in the AI media ecosystem. Living in this environment that normalizes such out of touch behavior will inevitably distort any human’s understanding of technical progress.Interconnects AI is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.* This was not a mass political campaign, but rather an opportunistic media coordination. For context, the Wall Street Journal had an exclusive story that Jacob coordinated before posting. I suspect that Jacob shared his plan of quitting in groupchats with AI safety advocacy groups ahead of time, e.g. the morning of posting, asking for amplification. This is normal practice, and could have included some prominent politicians. From there, I think it’s more likely that other politicians are bandwagoning on a rising issue. When you combine this with other factors, like Daniel Kokotajlo’s appearance on Joe Rogan coming out the same day – it definitely looks like a very well-executed, coordinated media campaign. This doesn’t mean it’s a conspiracy or a regulatory capture tactic within Democratic political structures. The determining factor seems to be that no one – including Jacob and those posting about X-risk today – knew that it would go so viral.* We do not have proof that RSI causes the risks these researchers forecast. The general argument for RSI follows as: The current pace of progress is very high, the current progress is heavily dependent on AI tools, the current AI tools are superhuman in some domains (e.g. math) – so, all together, AI is going to work more on itself and become superhuman in all relevant areas over time to autonomy and intellect. This view dramatically undersells human bottlenecks in building models and allocating resources at organizations, and draws conclusions on future AI capabilities more broadly.I called my alternative view to this, Lossy self-improvement. AI has always been very jagged, and we are making models which are superhuman goal-seekers at math and software engineering, but they have massive limitations on intuitions, creativity, and other types of reasoning that humans are strong at. With AI agents assisting research, we will rapidly find the areas where AI is superhuman – and I expect there to be well more than just research mathematics – but it won’t be a panacea for the current limitations of our approaches to LLMs.There is another understandable social dynamic at play here, causing many deep AI insiders to overstate the returns from RSI. Many of these researchers were the earliest people to bet on AI’s progress, and the extent to which they were visionaries should not be downplayed (see Ilya’s comments on deep learning as early as 2015). They have been right again and again, forecasting AI’s capabilities better than I certainly could have guessed. This does not, though, mean that their forecast of what will come next will be right. The core idea of RSI is a way to spend more compute on the process of developing a model recipe, rather than just spending more compute on the training run itself. We’re seeing benefits from it, but I argue the expected return on that input is far less than they believe.Their argument is that RSI will make AI progress go exponential, make it so we cannot monitor the technology, and enable rogue models and new forms of risk. This scenario is often called “Fast Takeoff”. We have not seen the stacking efficiency gains that massively reduce model size and cost, that would lead to an explosion in progress by allowing consistent speedups in experimentation.* The biggest short-term risk could be from the AI labs not taking safety seriously enough – they haven’t hardened their own infrastructure, enabling AI misuse to proliferate. From my earlier post on the HuggingFace-OpenAI incident, Lessons from the hacks:* Frontier labs do not seem like they’re watching the models closely enough, due to a general frenetic competitive environment & current SF cultureFrom OpenAI’s own retrospective, the misaligned model behavior was unfolding over months, and in some cases OpenAI did not know about the hacks for ~weeks. The time to response is too long and I do not think this is an OpenAI o… This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.interconnects.ai/subscribe

Housekeeping: Paid subscribers to Interconnects now get a permanent 40% discount on my book when purchasing at Manning.com. Access the code at the Interconnects perks page.Many AI optimists tend to compare what is happening in this AI boom to the industrial revolution, or to other periods of rapid technological advancement and diffusion into society. These comparisons fit on the scale of technological change, but miss a crucial factor in how most people are exposed to that change. The problem facing AI is that most people have no super tangible new goods thanks to it and society has more inertia resisting change than in previous eras.I’m writing this coming back online from a few weeks off for my wedding in New England. In this time it would have been very easy to not think about AI at all. The touch-points that average people have to AI products today are fringe, marginally beneficial, or even just very confusing to them (e.g. many people have heard about and brought up the OpenAI-HuggingFace incident, but don’t know what to make of it). People on the positive side think of AI as a way to make fun images, enhanced Google Search, etc. These are very small benefits. On the negative side is an association with addictive social media algorithms, friends of friends addicted to AI chatbots, and a plethora of takes on data centers.AI is still a rounding error in everyday lifeCore aspects of everyday life — family, food, transportation, and entertainment — have few direct impacts yet. It’s a remarkable breath of fresh air to pop out of the bubble and realize how little what is happening really matters today. Being obsessed with AI is a choice that a very few people have yet opted into. For example, the only thing I used AI for in this time was search and creative work (making the pretty seating chart for my wedding guests to find their table). In industrial revolutions past, average people got absolutely life changing outcomes. The First Industrial Revolution in the late 18th century gave access to cheaper clothing, cooking ware, reading material, and a shift to new livelihoods. The Second Industrial Revolution in the late 19th century introduced household machines (e.g. sewing machines), preserved food, indoor plumbing, photography, better light sources, bicycles, and further benefits of manufactured goods and electrification. The list is remarkable — most of these we still use regularly today — and very physical.While even the most optimistic versions of AI will usher in new scientific discoveries, advanced therapeutics for rare diseases, and potentially even sustained economic abundance, these benefits have the risk of being too indirect. How will a common citizen come to credit OpenAI or Anthropic for saving their life, if they went to their family doctor who told them about a new miracle cure?What percentage of Americans will care about OpenAI solving the Navier-Stokes Millennium Prize Problem? It feels very likely in 50 years that the average American’s day to day life looks very similar. Their home, appliances, relationships, and vehicles will be similar (of course, self-driving will continue to diffuse, but that has been developing on a very independent trajectory from the innovations of LLMs). In this time, AI will get a lot of credit. 50 years is a remarkable length of time with how fast everything is changing today in this, AI-focused narrow slice of the world.The most important part of what is happening early in the AI revolution, is building foundational infrastructure, and a general process, which will compound over decades. A major mathematical breakthrough today will look astonishingly minor in scope relative to the advancements that come later in the compounding journey. It is hard to predict what it looks like for every technology you use daily to get faster compounding improvements due to AI. Interconnects AI is a reader-supported publication. Consider becoming a subscriber.Breaking social stasisMuch of the narrative around AI is trying to push people to care, due to this long-term reality of progress, at least as a subconscious motive. This will take a long, long time to get right, and AI’s buildout faces immediate political problems due to this imbalance. Today’s AI is primarily a tool to serve the elite. For knowledge work, which is roughly half of the U.S. economy, AI is as fundamental as electricity (or quickly will be so, with rapid improvements to agents in the next 18 months). It’s highly destabilizing to have such a transformative, productive tool only bring half of society along. It’s not hard for many people to pick up on this — the technology economy booms while life stays otherwise stagnant.In writing this, I learned of Engels’ pause, which is “the period from 1790 to 1840, when British working-class wages stagnated and per-capita gross domestic product expanded rapidly during a technological upheaval.” If we — the leaders of the AI industry — think this is the closest analogue to what comes next for AI, those not benefiting are right to push back.AI is the greatest tool ever for scaling technology companies and starting new online-native small businesses. I don’t even expect the tech industry to grow in headcount and nurture its workers through an era of massive success — I agree with Doug OLaughlin that headcount would likely shrink while knowledge work output explodes. These sectors were already the most successful in the American economic system, so the brand of AI will be tarnished as not being a collective good. I worry that this instinctive reaction will kneecap AI’s development, sending it down a path that looks closer to the cautionary tale of American nuclear power. Part of the challenge is the speed and relentlessness of expectations in society. The AI industry has millions of eyes on it, and won’t get much patience to wait and bring innovations later. If given 100 years to diffuse into society, its impacts will certainly become much more obvious, like the industrial revolutions of centuries past. Together, the AI industry is facing a few simple issues, in what I would call the first half decade of 50-year diffusion process.* AI’s positive impacts early in its evolution are too indirect.* AI is facing a political backlash deeply intertwined with the history of Big Tech in Western society. This is only an AI story due to timing, and if AI’s exponential growth came decades after the growing pains of today’s technology platforms like Google and Meta, it seems likely that the datacenter issue would’ve never risen to such a central political position.Solving either of these would alleviate a substantial amount of pressure, and give the AI industry a lot more time in showing the positive case for why people should be okay with changes to the status quo (primarily economic). These are both made more challenging by AI self-labeling itself as negative and/or unsafe technology, through proclamations of doom and mass unemployment. The leading figures have begun addressing this issue, but the public needs more work to fully buy into the overarching trajectory.When zooming out long-term, I could see robotics and self-driving becoming closely linked in storytelling to the current AI revolution. If the intelligence explosion from mass-producing large language models does spill over into enabling the acceleration of robots in everyday life, humans will quickly latch onto the tangible benefits of AI. This is ironic, as many people have spent time trying to convince people that what is happening specifically with LLMs is very different than the previous decade or two of general AI progress. If the same dynamic later saved (or massively overshadowed) LLMs, it would be funny.Reflecting on what I expect the history of this era to look like, it feels a lot like growing pains of AI. Society needed to break out of old habits and work through problems that predate ChatGPT — which releases a lot of energy and frustration — in order to tap into the longer … This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.interconnects.ai/subscribe