Meet CHOMPS: Our Personal AI Research Assistant

Keeping up with the pace of AI news feels like drinking from a firehose with no off switch. New models, product launches, research papers—thousands of them flood the internet every day. It’s not just hard to keep up; it’s impossible to make sense of what actually matters.

When OpenAI released their deep research agent, we saw a spark: what if we could adapt that same iterative search-and-synthesis agentic architecture to help us cut through the noise and zero in on what’s actually relevant to our work—right now?

We needed a way to find genuinely new ideas that weren’t just interesting, but timely and influential—things gaining momentum in real-time and gaining traction among the right communities. Traditional deep research agents are great at exploring idea spaces, but they miss two critical dimensions: timing and influence. They lack the intuition to know whether an insight is gaining heat this week or whether respected voices in your space are paying attention. Without these signals, you end up with a lot of data, but very little you can act on.

That's why we built CHOMPS, our internal AI-driven agent designed to help us surface and explore insights that matter most to our work.

The Challenge

• Too Much Information: Endless streams of AI news, new models, company launches, and tech shifts.
• Missing Timeliness: Traditional research tools struggled with highlighting what was important now, not just historically relevant.
• Social Proof Gap: Information lacked context around popularity or trending topics.

Our Approach

We built CHOMPS on top of Exa.ai—a neural search engine that delivers real-time, semantic search across the live web. This gave us the scaffolding we needed to move fast, stay current, and query with nuance.

But CHOMPS isn’t just a search wrapper. It’s designed to think. First, we prime it with a perspective: companies, thinkers, themes we’re tracking. Then it scans the past week’s web to see what’s shifted—what’s bubbling up that’s novel or counter-trend. From there, it clumps sources, identifies second-order effects, and proposes hypotheses—not just headlines.

CHOMPS then conducts weekly scans for recent events tied closely to these topics.

Here’s how the process works:

1. Initial Search: Every week, CHOMPS searches broadly to identify significant recent developments.
2. Insight Grouping: It clusters these developments, looking for patterns or themes that suggest novel insights.
3. Focused Deep-Dive: With these insights as a guide, CHOMPS performs a more targeted deep dive, exploring related but less obvious sources to enrich our understanding.

The Impact

CHOMPS doesn’t just summarize; it synthesizes with intent. The outputs are structured to be immediately usable—each insight comes with a crisp headline, clear context, and a sharp point-of-view. We’re not left sorting through a wall of links; we’re presented with themes worth acting on.

• Dramatically reduced time spent tracking news, freeing our engineers and strategists to focus on their core work..
• Regularly surfaces genuinely novel, actionable insights tailored specifically to our strategic focus.
• Continuously adapts and learns, becoming smarter with every cycle of research.

What’s Next

• Launching a private beta for external teams who want concise, relevant, weekly insights.
• Building out a crowdsourced "Scout Network" for niche insights and trusted publishing.

Key Takeaway

CHOMPS isn’t just another information aggregator. It’s our personal research assistant—smart enough to sift through the noise and deliver actionable insights exactly when we need them.

Interested in trying it yourself?

Join our private beta and start receiving actionable, clutter-free AI insights directly to your inbox every week.