AI "Apocalypse Script" has only progressed 17%, but the founder of Google Docs says the Singularity is imminent

According to Beating Monitoring, Steve Newman, co-founder of Google Docs and chairman of the AI think tank Golden Gate Institute, published a post comparing the reality of progress against last year’s widely discussed “AI 2027” predictions, item by item. The conclusion is that overall quantitative projections have been completed by about 65%, but the most critical “AI software research and development acceleration multiple” (the degree to which AI accelerates its own R&D) has only reached 17% of the predicted value.

The 65% is mainly propped up by benchmark tests and data center construction. Newman judges that AI is still a “normal technology,” and the old problems—slow deployment and difficulty integrating into enterprises—have not been resolved. On the other hand, Anthropic’s annual revenue will be nearly 10 times higher in 2025; in the first three months of 2026, it will triple again to 300 billion dollars. Claude Mythos is able to string together vulnerabilities within mainstream operating systems and browsers to carry out executable attacks, but it has been restricted from release because it is too dangerous. GPT-5.5 can work independently for several days.

Newman believes that even under conservative estimates, models will begin accelerating their own development in 2026; an aggressive estimate is that the AI 2027 script will be delayed by several years before playing out fully. Progress is only 65% complete, yet the authors of “AI 2027” have moved the timeline forward. Lead author Kokotajlo in April shifted the median forecast for “automated programmers” (AI fully replacing top software engineers) from the end of 2029 to mid-2028, while the team still believes it is entirely possible for superintelligence to emerge within 18 months. Newman’s closing remark: with the dizzying pace of change today, it is very likely the slowest we will experience in the rest of our lives.