As of November 2026, no top-five AI developer has made a qualifying statement. While AI automation of R&D is accelerating—Anthropic recently made a threshold determination based on expected progress sec.gov and Google's CEO expects continued acceleration from recursive self-improvement—humans remain firmly in charge of direction. OpenAI states that people still "judge which ideas and results to pursue" openai.com, and Anthropic emphasizes that "large performance gaps persist" in choosing goals, w
Set against related capability questions, the estimate remains stable, acknowledging that while AI experiment selection will grow rapidly, labs will maintain a human-led framing for a couple of years .
Current Status and the Resolution Bar
As of November 2026, no qualifying event has occurred. The resolution criteria are highly specific and restrictive: they require not just that AI automates research, but a public report of an absolute headcount reduction in a named research or research-engineering job family, explicitly attributed to AI automation. Recent industry events illustrate how strict this is. Anthropic's October 2026 employment study described a sharp drop in junior technical h
Evaluated alongside related technical milestones, the forecast continues to reflect that labs will delay explicitly attributing any headcount reductions to AI well into the 2030s .
Current Status and the Observed-Progress Hurdle As of November 14, 2026, no top-five developer has issued a qualifying statement. Anthropic's November 10 SEC filing is the closest to date, declaring a determination that it met its Responsible Scaling Policy's AI R&D threshold. Crucially, however, the company explicitly based this on "what we expect from future models and our overall rate of progress, rather than on what we have observed from current models" [Anthropic, PBC - S-1/A-2](https:/
Accounting for broader expectations around structural bottlenecks in translating AI capability gains into measurable R&D productivity and governance friction that could delay explicit public attributions pushed the median timeline slightly later into mid-2027.
Baseline and Current Measures
The required threshold—a tenfold gain in training compute efficiency to reach a fixed capability level within a single year—represents roughly three times the historical baseline of 3× per year Trends in AI AI software progress: Data & research. While some current estimates are near the threshold, they do not resolve the question. An analyst's best guess of ~10× per year includes post-train
Integrating expectations about internal R&D compute allocations and shrinking efficiency doubling times pushed the median and upper percentiles slightly later to account for the lag in retrospective public reporting.
Current Status and the Resolution Bar
As of November 14, 2026, no qualifying case exists. The resolution criteria demand explicit public attribution by a top-five developer that an AI originated a training-method or architecture change that shipped in a released frontier model. Previous high-profile examples of AI-assisted engineering do not meet this bar. Google's AlphaEvolve 2 sources and its subsequent deployments 2 sources represent kernel optimization and hyperparameter t
Current Attribution Norms and Institutional Bottlenecks As of November 14, 2026, no NanoGPT speedrun record credits an AI alone, either on the main track or the optimization track NanoGPT speedrun record training time. The primary barrier to resolution is not purely technical, but rather the social convention of leaderboard attribution. Even when AI systems generate the vast majority of the code or optimization steps, human submitters rou
Set against related questions, the upper tail was extended to 'never' to account for the risk that the specific repository is abandoned before a sole-AI credit is merged.
The primary constraint on this resolution is not the underlying capability of frontier models, but the availability of a suitable evaluation instrument. Extrapolating METR's historical post-2023 doubling time of 128.7 days from the top entry of 17.4 hours (Claude Mythos Preview) suggests that frontier systems could reach a 160-hour 50% success threshold around mid-to-late 2027 The 10x-a-year clock: what METR's AI time horizons really ... [Ta
Set against related questions, the median and upper percentiles were shifted to 'never' to mathematically align with the 62% probability that a human-baselined 160-hour metric is never published.
As of mid-November 2026, frontier developers rely heavily on AI for implementation and experiment execution, but none has stated that a released model was principally designed by an AI. The scale of automation is massive: Claude authors over 90% of Anthropic's codebase contributions, Google DeepMind's production code is more than 75% AI-written, and OpenAI runs 3.14 agent-workdays per researcher-workday. METR’s recent assessment of Anthropic found progress on specific R&D tasks at around twice t
Evaluated alongside related research milestones, the timeline holds firm, tracking slightly behind experiment selection milestones as principal model design remains highly sensitive .
The 2026-11-14 ruling clarifies the current state of play: METR's November 5 assessment successfully measured an AI-driven speedup in research progress using internal Anthropic data, finding an approximate 2× acceleration. However, because it was a one-off first edition without a stated cadence, it does not resolve the question. Resolution now depends on a follow-up publication that independently measures an AI-attributable rate of research progress (in value) and, by publication, belongs to an
The forecast was adjusted slightly later to reflect the time required to establish a steady reporting cadence for third-party assessments.
Status Quo and Driving Forces As of mid-November 2026, no jurisdiction currently holds the binding authority to dictate a minimum share of a frontier developer's compute for safety or other non-capability purposes. Existing frameworks, such as the EU AI Act and California's SB 53, regulate AI models as consumer products rather than internal research instruments 2 sources. However, political momentum for such a lever is accelerating alongside AI capabilities. On November 10, Anthropic
The idea of a "negative internal-public gap"—deploying models externally before using them for internal AI R&D—remains an advocacy proposal from the AI Futures Project How to pace the US frontierHow to pace the US frontier. No top-five developer has adopted this policy [Beware delaying public deployment - by Eli Lifland](https://aifuturesnotes.substack.com/p/beware-delaying-pub
Set against related questions, the median and upper percentiles were moved to "never" because deploying models externally before internal research contradicts the core safety paradigms expected in most branches.
Current Letter Plateau The existing "Pacing the Frontier" letter has stalled well below the required threshold. As of November 14, 2026, the letter displays 1,386 total signatories https://pacingthefrontier.com https://pacingthefrontier.com. Nearly all of these were collected during the campaign's first two weeks, reaching 1,376 by mid-August. Since then, the count has added only ten signatures despite several major events in
Set against related questions, the upper percentiles were shifted to "never" to reflect the high probability that this specific letter campaign stalls permanently.
Ask a followup
Sign in to run · $20 free credit, no card · every claim cited