The Washington Post · Policy & Technology Investigation Analysis Horizon: 2026–2031 (5-Year Window)

AI Safety Pace and Capability Simulator

Interactive dynamic model translating Anthropic CEO Dario Amodei’s call for AI companies to slow the pace of development. Simulate recursive capability runaway, alignment verification latency, and control bandwidth limits across multi-decade frontier compute clusters.

Simulation Parameters

Live Model
0.40
Pace of frontier model iteration and training cadence per year.
0.80
Fraction of compute & research bandwidth reserved strictly for red-teaming & oversight.
0.60
Level of automated code generation & recursive self-improvement permitted before manual gate.
100,000
Frontier training hardware envelope dedicated to the next generation model run.
Control Stability Score
88.5
Out of 100.0 index
Safety Buffer Months
14.2
Oversight margin before divergence
Low Risk (Controlled Horizon)
Pacing buffer is sufficient to maintain oversight over recursive self-improvement loops.
5-Year Trajectory: Capability vs. Oversight Bandwidth
Effective Capability
Control Bandwidth
Recursive Self-Improvement Audit Log Simulated 60-Month Run
Milestone Cap Index Oversight Margin Self-Loop Gate State

AI Safety Pace and Capability Simulator Explainer

Read the explanation

The simulator balances capability growth against verification bandwidth across a five-year AI deployment trajectory. Accelerating development speed steepens capability growth. Once autonomy crosses the recursion threshold, self-improvement multiplies capability output. Selecting Aggressive Scaling slashes oversight to twenty-five percent. Capability rapidly overtakes control bandwidth, triggering a critical runaway point.

Super generates helpful tools and automates fact-checking across the internet proactively. If you enjoyed this tool, build your own with Super and share it with a friend.