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

According to Beating Monitoring, Google Docs co-founder and AI think tank Golden Gate Institute chairman Steve Newman released an analysis of the AI current situation in Q2 2026, referencing the widely discussed “AI 2027” scenario forecast from last year, and comparing each point to actual progress.

Core conclusion: The overall completion of the quantitative forecast for AI 2027 is about 65%, but the most critical “AI software development acceleration factor” (the degree to which AI accelerates its own development) has only reached 17% of the predicted value. The 65% figure is mainly supported by benchmark scores and data center construction progress, rather than truly reflecting AI’s capability leap in the real world. Based on this, Newman judges that AI is still a “normal technology,” constrained by traditional friction such as slow deployment and enterprise integration difficulties.

But he also lists evidence pointing in the opposite direction. Anthropic’s annual revenue in 2025 nearly multiplied by 10, and in the first three months of 2026, it tripled again to reach $30 billion, with SemiAnalysis estimating it at $40 billion in April. Claude Mythos became the first model restricted from release due to being “too smart,” capable of finding high-risk vulnerabilities in mainstream operating systems and browsers and chaining them into executable attacks. OpenAI researchers revealed that GPT-5.5 can operate independently for several days.

Newman presents two scenarios. Conservative scenario: models begin accelerating their own development in 2026, making AI’s impact on the economy and science unavoidable. Aggressive scenario: the script for AI 2027 is delayed by a few years, AI fully takes over its own development, then quickly solves issues like robotics, with barriers being eliminated by competitive pressure and AI’s own flexibility. The author of “AI 2027” also updated the timeline in April, moving the median prediction for “automated programmers” from late 2029 to mid-2028, while still believing that superintelligence could emerge within 18 months.

Newman’s summary: He increasingly believes that the singularity will arrive within the next few decades, and the dizzying pace of current changes is very likely the slowest we will experience in our lifetime.