Reddit's "Draft Comment" feature now loads thread content correctly on Vercel. Previously, Reddit blocked requests from datacenter IPs, which meant the thread context the LLM needs simply wasn't there. The fix tries Reddit's authenticated API first, falls back to anonymous JSON, then uses ScrapeCreators for production where Reddit IP-blocks Vercel. The prompt itself is also substantially smaller — a 40-comment thread that previously sent ~91KB of Reddit metadata to the model now sends ~6KB, which is the part that actually matters.
Importing a brand into your own account now works. Two separate failures were hitting users: a race condition that returned "User account not found" immediately after signup, and a missing column that caused "Failed to import" for everyone else. Imports now seed a fresh agent from the brand's URL, name, and logo, then walk you through the standard onboarding flow — personas, topics, a live scan — rather than copying data that would have been overwritten anyway.
The AI Agent Activity dashboard now counts visits to legacy /ap/{username}/... URLs that crawlers still hit from old indexed links. Those URLs were real brand traffic that the path filter was silently dropping. A handful of other paths that were inflating the count with 404s and unverified hosts have been removed from the tally.
Marketing pages for industry, use-case, and team-specific URLs now run the public onboarding scan on submit rather than bypassing it for logged-in free users.