Articles by tag: ai-degradation

Degradation of Claude Code: reasoning 3.5 times less

Analysis of 6852 Claude Code Sessions from AMD: drop in thinking depth, increase in errors and API costs by 80 times. Anthropic recommendations and switch to alternatives. Study the data for IT teams.

Degradation Claude Code: reasoning depth -73%

Analysis of 6852 Claude Code sessions: drop in thinking depth by 73%, cost increase ×122, blind edits 33%. Workaround and metrics for middle/senior dev. Check on your own tasks.

Self-training: why it works and how to avoid collapse

Breaking down self-training mechanisms on synthetics, model collapse risks, and strategies for middle/senior ML engineers. Filtering rules, verification, real anchor. Improve your training pipeline without degradation.