Spectra

The Problem

I co-founded Spectra to fix a specific failure in how AI summarizes news: multiple agents scraped the same story from different sources to write a more neutral article, but early AI hallucination rates meant a confident, well-written summary could still be quietly wrong. Modern news consumption is fragmented and overwhelming on top of that, but trust was the problem that actually shaped the design.

• AI-generated summaries can be confidently wrong
• Information is scattered across sources with no shared context
• Feeds prioritize noise over relevance

The Solution

I designed Spectra so every sentence in a generated article carried its own interaction, letting a reader trace it back to the exact source it came from. Personalization and discovery mattered too, but nothing shipped until the trust layer did.

A more contextual way to consume news

AI-Assisted Story Exploration

With source-level trust handled, I could design the exploration layer on top of it: surfacing perspectives and related context without asking users to just take the AI’s word for it.

• Surface relevant perspectives and related topics, each traceable
• Summarize evolving stories without losing the source trail
• Reduce information overload without hiding where information came from

Personalized News Experience

The platform adapts to user interests and reading behavior to create a more focused and relevant feed.

• Custom curation from trusted sources
• Personalized topic discovery
• Reduced noise and repetitive content

Designed for engagement and clarity

Interactive Reading Experience

Instead of passively scrolling through headlines, users can actively explore topics, follow related discussions, and better understand the context behind stories.

• Conversational interactions powered by AI
• Social and editorial signal integration
• Designed for exploration, not just consumption

Outcome

Spectra received a national innovation grant for business validation. The sentence-level source attribution became the product’s core differentiator, and the same trust-through-evidence approach has shaped every AI product I’ve designed since.