Are we cooked?

A developer's candid and viral reflection has struck a chord in the tech community, articulating a deep-seated anxiety about the nature of AI's impact. After subscribing to OpenAI's GPT Codex and Anthropic's Claude in December 2025, the programmer experienced a dramatic reduction in manual coding, not because the AI produces flawless code, but because it fundamentally assumes the intellectual burden of problem-solving and implementation. This shift led to an existential realization: contemporary large language models (LLMs) represent the automation of intelligence itself, a qualitatively different disruption than historical automation of physical labor.

The post argues that neural networks like GPT-4 and Claude 3 Opus are not merely advanced tools but systems capable, in principle, of automating any new cognitive task that arises. This has prompted the developer, who describes their previous optimism as 'copium' or coping-based optimism, to consider abandoning software development for a field like biotech research. However, they express a paralyzing fear that the relentless pace of AI advancement—evidenced by models gaining multimodality, longer context windows, and agentic capabilities—means even scientific discovery may not be a long-term sanctuary from automation.