# What Anthropic told investors about the accelerating pace of AI research > Source: > Published: 2026-11-10 19:19:46+00:00 # What Anthropic told investors about the accelerating pace of AI research Something interesting this way comes Anthropic's journey to the public markets has generated three S-1 filings thus far. The first did not include full Q3 numbers. The second did. The third included more detail on the company's (quick) growth in the first month of Q4. The financial notes were the headline news. But buried in the filings is a series of word changes worth our time. In its first filing, Anthropic warned that it might only recognize a speedup in its research progress after it occurred. A week later, the company said it had missed an acceleration in the past and was nearing its own safety threshold for AI-driven acceleration. Eight days later, Anthropic said in a fresh filing that it had met the threshold thanks to what it expects from future models and its current pace of progress. This matters. AI building AI, what we often refer to loosely as recursive self-improvement, or RSI, could represent an incredible breakthrough in how we build smarter synthetic intelligence. RSI is also a prize; we may reach artificial *general* intelligence before we reach RSI, but that's not guaranteed. The first lab, or nation, to achieve real RSI could pull ahead of rivals, granting it an edge in future AI development and deployment. Anthropic, an AI lab famous for its concern about the same AI systems it is building, has a safety policy in place to catch the onset of recursive self-improvement, which forms part of its "[Responsible Scaling Policy](https://www.anthropic.com/responsible-scaling-policy)." All that in hand, here's how Anthropic discussed progress in using AI to build AI in its [first S-1 filing](https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm) (emphasis added): Our ability to maintain our position at the frontier of AI depends in part on our use of AI systems to accelerate our research processes, a recursive dynamic we refer to as “closing the loop,” in which successive versions of our models assist in building the next, with the goal of compounding capability gains over time. Our models have been and are increasingly becoming critical tools in designing, training, testing, and optimizing next-generation AI systems. [...] In addition, **our increasing use of AI systems to accelerate our own research and development may cause the capabilities of our models to advance more rapidly than our ability to develop, validate, and implement corresponding safety, alignment, and security measures. Much of our evidence regarding the pace of such AI-driven acceleration is derived from lagging indicators, and our ability to forecast or measure the rate at which our capabilities are advancing, including any acceleration attributable to our most recent or unreleased models, is limited.** As a result, we may not identify a significant acceleration in our rate of development until after it has occurred. The text says that progress may outstrip safety, and acceleration may only be detectable in reverse. About what you expected? Agreed. Now, fast forward to the [company's first IPO update](https://www.sec.gov/Archives/edgar/data/2130357/000162828026068676/anthropicpbc-sx1ax1.htm), its initial S-1/A filing (emphasis added): [W]e have in the past not identified, and may in the future not identify, a significant acceleration in our rate of development until after it has occurred. **For example, our internal estimates and our expectations regarding future models, rather than our current models, indicate that we may imminently meet our Responsible Scaling Policy’s definition of “dramatically accelerated progress attributable to AI.”** We have not designated any of our current models as exceeding this threshold, and we continue to assess the underlying data. We are also evaluating the implications of meeting the AI research and development capability threshold and will release further reports in the future as we set forth in our Responsible Scaling Policy. Here Anthropic said that it might *soon* meet its threshold for AI-driven acceleration, and that it missed a previous speed-up of its AI research ("we have in the past not identified"). Notably, the time between Anthropic's first S-1 filing (October 26th) and its second (November 2nd) was minimal. I don't know precisely why the company said one thing in October and another in November, apart from allowing that a filing date is not a *writing* date. Then, the [second S-1/A filing dropped November 10th](https://www.sec.gov/Archives/edgar/data/2130357/000162828026072484/anthropicpbc-sx1ax2.htm), with the following text (emphasis added), underscoring that Anthropic shifted from saying it was near to meeting the threshold to saying that it had met it: For example, **our use of AI systems has already been accelerating our research and development, and our internal estimates indicate that we may experience a further increase in the rate of progress in AI capabilities.** Because the AI research and development capability threshold under our Responsible Scaling Policy applies where we “observe or expect” a specified degree of accelerated progress, we made a determination that we have met this threshold based on what we expect from future models and our overall rate of progress, rather than on what we have observed from current models. We continue to assess the underlying data. We are also evaluating the implications of meeting the AI research and development capability threshold and will release further reports in the future. Hot damn. This is not 'sparks of AGI,' or whatever other hype term you've become burned out on in recent quarters. No, this is Anthropic telling investors in an SEC filing that AI has accelerated its research capabilities, and that it expects ("from future models and our overall rate of progress") even more acceleration in the future. Why does this trip the Responsible Scaling Policy? The 'threshold' listed includes the following riff (lightly edited): We will consider this threshold to be met if we determine that either our models would be able to fully substitute for our entire set of Research Scientists and Research Engineers, at competitive costs (i.e., within a factor of 5); or there is “dramatic acceleration” of the pace of AI progress for reasons that likely relate to the automation of AI R&D [where] we observe or expect double the rate of progress in AI aggregate capabilities compared to both the rate we’d expect and the fastest rate of extended progress we’ve observed in the absence of significant AI contributions to AI R&D and it is plausible that this doubling is substantially attributable to the automation of research and/or engineering (as opposed to other factors, such as increased headcount, compute, or general productivity), such that continuation of the trend in AI progress seems likely to lead to even greater acceleration. So, Anthropic says it has met its policy threshold. What has Anthropic promised in response? Investing in "moonshot" safety R&D projects, reaching an "eyes on everything" internal state, undertaking "systematic alignment assessments to examine Claude’s behavioral patterns and propensities," bulking its internal red teams until they outperform external bug bounty hunters, and publishing reports. Seems reasonable enough. It's also nice that we can simply pull up a doc, read it, and infer from the text what the company will do! *Real talk: If AI research does indeed accelerate from here, worries about AI-driven job loss could return to the fore. For example, the* *Remote Labor Index* *found that Anthropic's latest model (Mythos/Fable 5.2) scored 41.3% on the benchmark, nearly doubling the preceding record set by OpenAI's GPT-6 Astra.*