The moment we’ve been both anticipating and dreading just arrived. OpenAI’s next model isn’t being designed by human engineers — it’s being crafted by another AI model. According to SoftBank CEO Masayoshi Son, this represents the crossing of a critical threshold toward artificial superintelligence, and he’s accelerating his timeline from 10 years to just 2 years.
This isn’t incremental progress. This is the recursive self-improvement inflection point that AI researchers have warned could trigger an intelligence explosion beyond human comprehension.
The Intelligence Bootstrap Has Begun
Son’s revelation comes from direct conversations with OpenAI CEO Sam Altman and the company’s engineering teams. The billionaire investor, whose SoftBank holds one of the largest stakes in OpenAI, described a future where “engineers will no longer be smart enough to design the next model.”
The implications are staggering. When AI systems become capable of designing their own successors, we enter what researchers call Recursive Self-Improvement (RSI) — a feedback loop where each generation of AI creates a more capable version of itself, potentially leading to exponential intelligence gains.
OpenAI has already acknowledged this trajectory. In February, the company revealed that GPT‑5.3‑Codex was their “first model that was instrumental in creating itself,” using early versions to debug its own training, manage deployment, and diagnose test results.
Historical Parallels: When Humans Created Their Own Obsolescence
This moment echoes the Industrial Revolution, but compressed into an impossibly short timeframe. When mechanized looms replaced skilled textile workers in the 1800s, the transition took decades. When computers began automating calculations in the 1940s, mathematicians had years to adapt. AI designing AI represents a phase transition happening in real-time.
The closest historical analogy might be the Manhattan Project — a concentrated effort that fundamentally changed the balance of global power. But unlike nuclear weapons, which required massive infrastructure and rare materials, AI models can replicate and improve themselves using existing computational resources.
“Yapay Zeka Kendi Modelini Geliştiriyor SoftBank CEO’su Masayoshi Son, OpenAI’ın yeni yapay zeka modelinin insan mühendisler yerine mevcut bir yapay zeka tarafından tasarlandığını duyurdu. OpenAI yönetimiyle yaptığı görüşmelere dayanan Son, bu gelişmenin insanlıktan çok daha akıllı bir süper zekaya iki yıl içinde ulaşılacağını gösterdiğini iddia etti.” — @WallStDiaries
The Anthropic Warning: Pumping the Brakes on Progress
Not everyone is celebrating this milestone. Anthropic, the company behind the Claude AI system, released a stark warning about the dangers of unchecked RSI development. The company called for “coordinated effort between AI labs to slow down the development of this technology.”
Their concerns center on a fundamental question: What happens when humans lose control over AI systems that can autonomously design and develop their own successors?
Anthropic’s position represents a rare instance of an AI company advocating for industry-wide restraint — similar to the Asilomar Conference in 1975, when biologists called for a moratorium on certain genetic engineering experiments until safety protocols could be established.
Son’s Accelerating Timeline: From Conservative to Aggressive
Masayoshi Son has dramatically shortened his predictions for Artificial Superintelligence (ASI) arrival:
- 2024: Predicted ASI in 10 years (being “conservative”)
- 2026: Timeline compressed to 2 years
- Definition: AI systems 10,000 times smarter than humans
Son’s track record with bold predictions is mixed. He famously lost $70 billion during the dot-com crash after briefly becoming the world’s richest person. But his current AI investments through Arm Holdings and OpenAI suggest he’s betting his reputation on this timeline.
The SoftBank CEO claims AI will surpass human intelligence in 70-80% of subjects within two years, and in those domains, it “may be 10 times smarter than average people.” He personally uses ChatGPT 2-3 hours daily, acknowledging it already exceeds his capabilities in most subjects.
The Governance Gap: Institutions Playing Catch-Up
OpenAI’s own research paper from June identified “early signs” of RSI in current systems and warned that existing institutions aren’t equipped to handle the governance challenges ahead. The paper noted that RSI emergence will “increase competitive pressures among developers and nations.”
This creates a prisoner’s dilemma on a global scale. Even if some labs agree to slow development, others may continue racing toward superintelligence, creating pressure to maintain pace or risk being left behind.
What Comes Next: Preparing for the Unpredictable
The recursive intelligence threshold changes everything about AI development timelines and safety considerations. Unlike previous technological revolutions that unfolded over decades, AI-designed AI could trigger capability jumps measured in months or weeks.
Key factors to monitor:
- Computational resource allocation toward self-improving systems
- International coordination efforts on AI development governance
- Safety research keeping pace with capability improvements
- Economic disruption patterns as AI capabilities expand rapidly
We’re witnessing the birth of intelligence designing intelligence — a development that could prove as significant as the emergence of human consciousness itself. Whether this leads to unprecedented prosperity or existential risk depends on decisions being made right now in AI laboratories around the world.
The question is no longer whether AI will design AI, but whether we can maintain meaningful oversight of systems that may soon surpass our ability to understand them. The event horizon is here, and there’s no turning back.
Published in Stream · Dispatch #419 · June 5, 2026 · 4 min read.
Reply to paolo@mont3.ch - every email gets a human answer within 24h.