METR AECI Shows Anthropic Capabilities Progress Doubled October 2025 to October 2026

Anthropic disclosed in its S1 filing that it may imminently meet the Responsible Scaling Policy threshold of capabilities progress twice the pre-AI baseline. METR investigators accessed internal systems and measured the Epoch-derived AECI index directly. Redacted evidence, described as the majority of their informative data, supported AI-driven automation of R&D as the dominant driver rather than compute scaling alone. The observed 2x overall rate implies a 2-3x acceleration in compute multipliers. Under a Cobb-Douglas model treating researcher labor and experiment compute as equal inputs, this requires a 4-9x effective speedup in human researcher output to sustain the measured pace after accounting for diminishing returns. External timeline forecasts that assumed linear human progress therefore understate the feedback loop now active. Once AI systems contribute the majority of novel algorithmic improvements, each subsequent multiplier discovery shortens the interval to the next. Anthropic's July 2026 RSP update formalizes the threshold; the METR data indicate it has already been passed. Frontier labs must now recalibrate deployment schedules and safety evaluations against compressed iteration cycles rather than historical human-only rates.