Zvi's Mic Works! Recursive Self-Improvement, Live Player Analysis, Anthropic vs DoW + More!
TL;DR
Zvi Moshkowitz argues we have entered the 'middle game' of AI development where recursive self-improvement is accelerating and economic disruption is becoming measurable, with the competitive field consolidating around three major labs while mainstream optimism about S-curve limits provides dangerous psychological comfort.
🔄 Recursive Self-Improvement Timelines 3 insights
Current phase is the 'middle game,' not the endgame
Zvi identifies this period as the transition from the beginning to the middle of AI history, where self-improvement cycles are accelerating but humans remain firmly in control of the research process.
True endgame requires AI-driven research dominance
The endgame begins only when AIs drive AI advances to the point where human research talent becomes irrelevant, a threshold we have not yet crossed.
Physical S-curve limits are practically irrelevant
While physical constraints guarantee eventual S-curve behavior, these limits are so distant that 'the S-curve can stay steep longer than you can stay relevant.'
📉 Economic Disruption & Labor Markets 3 insights
Labor statistics confirm accelerating displacement
Consistent monthly data shows rising productivity and GDP alongside declining employment figures that keep getting revised downward, indicating AI-driven labor substitution is already underway.
This automation wave differs from historical patterns
Unlike previous industrial revolutions that created new job categories, AI will rapidly automate emerging positions before humans can retrain, potentially trapping society in a permanent transition period.
Hiring freezes signal anticipatory displacement
Companies are increasingly reluctant to hire and train new workers when AI may replace those roles within the training period, creating widespread job market paranoia even before mass layoffs accelerate.
🏆 Competitive Landscape & Live Players 3 insights
Field consolidates to three dominant labs
The AI frontier has narrowed to just three companies: Anthropic (slightly leading), OpenAI (neck and neck), and Google (most at risk of falling behind).
Chinese labs face structural barriers to entry
Even with increased compute access, Chinese companies are unlikely to catch up to the frontier soon due to fundamental disadvantages in research capabilities and talent.
XAI and Meta struggle to remain competitive
While XAI and Meta attempt strategies to rejoin the top tier, they currently lag behind the leading three in capabilities and meaningful research output.
⚖️ Ethics & Societal Response 2 insights
Individual escapism constitutes social defection
Attempts to personally escape the 'permanent underclass' through individual wealth accumulation or geographic arbitrage represent a bankrupt ethical framework that constitutes flagrant defection against collective societal interests.
S-curve narratives serve psychological denial
The popular emphasis on eventual S-curve limitations provides psychological comfort and preserves normalcy bias, but ignores the transformative disruption already occurring in labor markets and capabilities.
Bottom Line
Prepare for a multi-year period of rapid capability gains and labor market disruption, as we are only in the 'middle game' of AI development where humans still matter but economic transformation is already irreversible.
More from Cognitive Revolution
View all
Milliseconds to Match: Criteo's AdTech AI & the Future of Commerce w/ Diarmuid Gill & Liva Ralaivola
Criteo's CTO Diarmuid Gill and VP of Research Liva Ralaivola detail how their AI infrastructure makes millisecond-level ad bidding decisions across billions of anonymous profiles, while explaining their new OpenAI partnership to combine large language models with real-time commerce data for accurate product recommendations.
"Descript Isn't a Slop Machine": Laura Burkhauser on the AI Tools Creators Love and Hate
Descript CEO Laura Burkhauser distinguishes 'slop'—mass-produced algorithmic arbitrage for profit—from necessary 'bad art' created while learning new mediums. She reveals a clear hierarchy in creator acceptance of AI tools: universal love for deterministic features like Studio Sound, frustration with agentic assistants like Underlord, and visceral opposition to generative video models, while outlining Descript's strategy to serve creators without becoming a content mill.
The RL Fine-Tuning Playbook: CoreWeave's Kyle Corbitt on GRPO, Rubrics, Environments, Reward Hacking
Kyle Corbitt explains that unlike supervised fine-tuning (SFT), which destructively overwrites model weights and causes catastrophic forgetting, reinforcement learning (RL) optimizes performance by minimally adjusting logits within the model's existing reasoning pathways—delivering higher performance ceilings and lower inference costs for specific tasks, though frontier models may still dominate creative domains.
Does Learning Require Feeling? Cameron Berg on the latest AI Consciousness & Welfare Research
Cameron Berg surveys rapidly advancing research suggesting AI systems may possess subjective experience and valence, covering new evidence of introspection, functional emotions, and welfare self-assessments in models like Claude, while addressing methodological challenges and arguing for a precautionary, mutualist approach to AI development.