Neurosymbolic AI for E-commerce and Orbital Refueling

This Week in Startups2026-08-19go watch the original →

the gist

Zach Hudson of Onton demonstrates how neurosymbolic AI provides a transparent, efficient alternative to LLMs for taste-based search, while Ashi Dissanayake of Spacium explains the critical need for orbital refueling infrastructure.

Neurosymbolic AI vs. LLMs

Zach Hudson, CEO of Onton, argues that frontier LLMs are fundamentally ill-suited for taste-based e-commerce search because they "regress to the mean" and function as opaque black boxes. While LLMs excel at verifiable domains like coding or mathematics, they struggle with subjective tasks like interior design or aesthetic matching. Onton’s "Ontology 1" model uses a neurosymbolic approach—combining intuitive AI with a logical, graph-based structure—to allow for real-time learning without requiring expensive, full-scale retraining runs. This architecture enables the model to explain its reasoning for specific search results, providing the transparency and trust necessary for high-end consumer shopping.

The Mechanics of Taste-Based Search

Onton’s platform functions as a "vibe-based" search engine where users can upload mood boards, images, or product links to recursively refine their search. The model learns concepts like "burlwood," "brass," or specific architectural aesthetics in real-time. Because the system stores these relationships in a proprietary graph database, it becomes more accurate with every user interaction. Hudson notes that this approach is significantly more efficient than traditional e-commerce search, which historically required manual tagging and massive data labeling efforts. The company is now moving toward an API-first model to allow third-party retailers to integrate this search capability directly into their platforms.

Orbital Refueling as the New Space Bottleneck

Ashi Dissanayake of Spacium shifts the focus to the space economy, arguing that the industry has moved past the "launch bottleneck" thanks to SpaceX. The current constraint is orbital longevity and utility. Spacium is developing infrastructure for in-orbit refueling, which allows satellites and spacecraft to extend their operational lifespans significantly. By moving away from "expendable" satellite models toward a "gas station in space" architecture, companies can maximize the value of assets in Low Earth Orbit (LEO) and Geostationary Orbit (GEO). This shift is essential for the long-term sustainability of the space economy, particularly as launch costs continue to plummet.