Pulse

Project Overview

Pulse redefines how people consume news by combining AI-driven editorial analysis with real-time fact verification. The platform ingests content from multiple RSS sources, groups semantically similar stories, and generates editorial rewrites in five distinct styles — from deep analytical breakdowns to quick TL;DR summaries. Six specialized AI experts provide unique perspectives across politics, economics, science, sports, culture, and health.

The Problem

Traditional news platforms deliver single-voice reporting that often lacks depth and verification. Users are overwhelmed by information pollution, with no easy way to cross-reference claims, understand multiple perspectives, or verify facts in real-time.

The Solution

Pulse’s editorial render pipeline uses Gemini 2.5 Flash to extract structured fact tables from articles, generate rewrites in 5 editorial styles, and deterministically verify numbers and dates to prevent hallucinations. Combined with Polymarket prediction integration and smart bookmarking, users get a complete intelligence toolkit.

Key Features

Code Highlight

// Editorial render pipeline with fact verification
async function renderEditorial(article: Article) {
  const factTable = await gemini.extract({
    model: "gemini-2.5-flash",
    prompt: buildFactExtractionPrompt(article),
    schema: FactTableSchema,
  });

  const rewrites = await Promise.all(
    EDITORIAL_STYLES.map((style) =>
      gemini.generate({
        prompt: buildRewritePrompt(article, style),
        facts: factTable, // deterministic verification
      })
    )
  );

  return { factTable, rewrites };
}