Added a new pre-seed YC company, @usemoss to Plushcap: plushcap.com/moss. H/t to @diabhey for telling me about them. They’re working on pushing down the latency threshold on retrieval which is a blocker for conversational agents, especially Voice AI agents 🧵

As with any good early stage DevTool, most of their content is in their docs docs.moss.dev/docs, with a few recent technical blog posts they wrote on their approach to solving issues in this problem space

For example, here’s their recent post "The Retrieval Latency Tax: Why Your AI Agent Feels Slow": moss.dev/blog/retrieval… Plushcap blog post analysis page here (only 3 posts so far): plushcap.com/companies/moss…

Moss competes in the real-time data infrastructure space (a heterogenous grouping, some more focused on scaling protocols versus latency requirements): plushcap.com/competitive/sp… As well as Voice Agents: plushcap.com/competitive/sp…

Would be interesting to see content from these guys on what problems are truly solvable with their edge-based approach, versus what is still intractable, for example 3rd party data services like Clearbit, Crunchbase, ZoomInfo, etc can’t be served this way without pre-caching.

Lots of room for improvement with voice agents, most of them still suck (pressing “0” on repeat while shouting “talk to operator” is still the default). Check Moss’ homepage out here: moss.dev

@fullstackpython Smart take edge excels at low-latency but struggles with fresh 3rd-party data (Clearbit, ZoomInfo) without heavy pre-caching. Biggest unsolvable edge limit today?