AI Engineer host Annika Sokhal pitched Moshi as a 24/7 quant analyst that watches 2.4 billion data points and pushes insight to traders instead of replacing them.
The argument starts with scarcity. Sokhal said the world made about 175 zettabytes of data in 2025, attention stayed flat, and the gap keeps widening for investors paid to synthesize.
Her answer turns on efficiency. The brain runs on about 20W with 86 billion neurons, while the systems trying to copy a sliver of it need a power plant, Sokhal said, so she asked, "why are we spending everything trying to push it out of the loop?"
Edge, in her definition, is three things: seeing news first, knowing blind spots and second-order effects, and having conviction to size a trade. If everyone has the same models, MCP connectors and prompts, that edge cannot be downloaded, she argued. It lives in a trader's head, built over decades on a desk, and walks out when they retire.
Moshi claims to feed that brain, not replace it. Sokhal described six deterministic agents watching markets 24/7, ingesting PDFs, emails and subscriptions, learning 300 new things in four seconds, and measuring every claim for hallucination and traceability. She cited up to 35% more efficiency for users who offload the grind and keep the judgment, but the talk showed no live audit, no named client, and no public benchmark for that hallucination scoring. Sokhal spent nearly a decade trading energy derivatives at JP Morgan before coming to the US on a Fulbright to study data science at Rice, and now runs Houston-based Moshi as a VC-backed real-time quant platform, per DEV Community.
