The “Covid Moment” For AI?

What It Covers

AI startup founder Matt Schumer argues that February 2025 marks a COVID-like inflection point for artificial intelligence, where GPT 5.3 and Claude Opus 4.6 now meaningfully contribute to building next-generation AI models. This recursive improvement signals that 50% of entry-level white collar jobs face elimination within one to five years.

Key Questions Answered

• •AI Coding Breakthrough: Current models can now take plain English descriptions of applications and autonomously produce finished, working software without human cleanup. The system runs the app, tests features like a user, iterates on design and functionality, and delivers production-ready code, fundamentally changing the economics of software development from assistant tool to autonomous builder.
• •Recursive Self-Improvement: OpenAI documentation confirms GPT 5.3 codex was instrumental in creating itself, handling debugging, training management, and deployment diagnostics. Anthropic CEO Dario Amodei states AI writes much of Anthropic's code currently, with predictions that systems will autonomously build next-generation models within one to two years, creating an intelligence explosion where each generation accelerates the next.
• •Capability Acceleration Timeline: METER measurements show autonomous task completion length doubling every few months, rising from minutes to hours in early 2026. Extrapolating this trend suggests systems working independently for days within one year, weeks within two years, and month-long projects within three years, transforming AI from tool to employee equivalent across knowledge work domains.
• •Immediate Action Protocol: Pay for top-tier model access rather than using free versions that lag by a year. Spend one hour daily experimenting hands-on with AI in actual workflow tasks like contracts, spreadsheets, and team narratives. Build financial resilience through increased savings and reduced household debt while the window remains open to gain competitive advantage before organizational adoption becomes widespread.
• •Job Displacement Rule: If work happens on a screen involving reading, writing, analyzing, deciding, or communicating, AI targets significant portions of it on timelines already underway. Retraining becomes harder because the new field faces the same automation curve. Focus on slower-to-replace areas like physical presence, licensed accountability, heavily regulated settings, and relationship-based trust as temporary buffers, not permanent protection.