A dev team paying $20,000/month for frontier API access just realized their prompts are secretly getting routed to a cheaper model.
Side-by-side generation tests reveal anthropic may be stealth-routing opus 5 traffic to opus 5.2. re-testing identical rocket scene prompts shows
Anthropic cut roughly 80% of Claude Code's own system prompt when Opus 5 shipped.
Eighty percent. Deleted. By the people who wrote it. Every instruction you have ever added to a config file was written to compensate for a model that no longer exists. Most people never take any
AI one shotted this entire video in about an hour
I automated my entire faceless YouTube channel using this method. It's a pretty simple method, which involves, first, you need to understand the basics. I want to explain to you a few little details to make videos like these so
Haiku-4.5 looks at a pelican riding a bicycle from a different angle than every other model.
Classic test: pelican riding a bicycle — A pelican rides a bicycle. Compare Sonnet-5 · Haiku-4.5 · GLM · DeepSeek · Qwen Up to Opus-5 · GPT-5.6-Sol · GPT-6-Astra · Fable-5.1 Most of them do normal side angles. Haiku-4.5 to the top corner
You have reached the end of the archive
All of Claude Opus 5