Anthropic’s new opt-in policy shows the next UX edge: trust by default.
Smart builders will turn this into a product.
I’ll help you ship the MVP in 4 weeks.
Lovable didn’t just build an AI agent, they debugged it like pros.
By baking in PostHog from day 1 (feature flags, LLM traces, experiments), they caught GPT‑5 weirdness before it broke prod.
MVP tip:
→ Don’t bolt on analytics later
→ Track LLM behavior now
→ Use flags to ship...