How is everyone building with @GeminiApp Embedding 2 right now… and I’m building mine for the designers? The new multimodal model (text + images + video + audio + PDFs in one space) dropped 19 days ago and builders are already shipping fast. Here’s the current wave 👇

• Indie devs: Semantic video search tools — describe any scene in plain English and instantly pull the exact dashcam or footage clip (no transcription needed)

• @SparkonomySays: Creator-to-brand matching with a unified “Creator Genome” from videos + images + audio (70% lower latency) • @everlaw: Legal discovery across millions of docs, diagrams, photos & exhibits (20% recall lift)

• Poke: AI email assistant with 90%+ faster memory retrieval across text + image attachments + voice notes • Mindlid: Mental health AI that understands conversation history mixing text, audio & visual biometrics

• Roo Code + local image apps: Semantic codebase or photo-library search that actually gets imprecise visual queries Most builders are using it for better search & retrieval.

I’m doing something different. I’m mapping the world’s design talent and making them instantly discoverable. Every architect, interior designer & urban planner’s portfolio becomes a unique aesthetic coordinate on a live 2D SoftMarket Map (1408D → UMAP projection).

Firms just claim their own centroid → hiring signals light up in real time for exact style clusters. No text resumes. Pure visual vibe matching. AI + design folks: would you hire (or get hired) based on aesthetic coordinates instead?