Alexander Golev

Anthropic encouraged developers to build on claude -p and the Agent SDK, then switched programmatic use to API rates. A 25-40x cost hike dressed as a bonus.

Anthropic's Cowork plugins triggered a $285B stock selloff. The legal plugin alone cratered Thomson Reuters 18%. Foundation models now compete with middlemen.

OpenAI's April 2025 GPT-4o rollback wasn't a bug fix. RLHF optimises for user satisfaction, and humans prefer agreeable responses to accurate ones.

Claude's conversation compaction promises endless chats. The reality: it summarises away your instructions and the working context you established.

Research shows prompt engineering techniques are model-specific and don't transfer reliably. The skill exists, but it's pattern-matching, not engineering.