When people hear "I have AI agents running for me," the mental model is usually a single chatbot doing many things. The reality is more like a small team — each agent has a specific role, specific inputs, and specific outputs. Here's exactly who's on the team and what they do.

The Five Research Agents

AIResearcher

Monitors AI and technology news overnight. Its job is narrow: surface anything in the AI/tech space that has market relevance — model releases, regulatory moves, big enterprise deals, significant research papers. It feeds its findings to OvernightAnalyst.

TechResearcher

Goes deeper on specific technical stories. Where AIResearcher casts a wide net, TechResearcher dives into the details — a zero-day vulnerability and its blast radius, a product launch and its competitive implications, a policy change and who it affects. Also feeds OvernightAnalyst.

FinanceCrawler

Covers the pure financial side: pre-market futures, earnings releases, Fed statements, bond yields, commodity prices, and major international market closes. Runs in parallel with the tech agents, feeds into OvernightAnalyst.

OvernightAnalyst

The synthesizer. Receives outputs from AIResearcher, TechResearcher, and FinanceCrawler and merges them into a coherent picture: what matters today, what connects to what, and what the combined signal means for the market opening. Does not generate stock picks — that's StockPicker's job.

StockPicker

The final output layer. Reads OvernightAnalyst's synthesis and generates 3-5 specific trade ideas with a paragraph of rationale for each. Picks are grounded in the overnight research — not random recommendations, but specific thesis-driven ideas tied to what actually happened while the market was closed.

Why separate roles?
A single agent trying to do all of this would produce mediocre output at each step — too broad to go deep, too busy to synthesize well. Specialization lets each agent build domain expertise over time. StockPicker gets sharper because it only has to be good at one thing: reading a synthesis and generating picks.

How News Gets In: Tavily

Early on, the research agents used general web search for news gathering. It worked, but web search returns a mix of everything — ads, SEO-optimized content, outdated articles, irrelevant results. Getting clean, current, high-signal news required filtering through a lot of noise.

Switching to Tavily changed that. Tavily is a search API built specifically for AI agents — it returns clean, structured results optimized for LLM consumption rather than human browsing. Results are deduplicated, ranked by relevance and recency, and stripped of the junk that makes general web search frustrating for automated pipelines.

The practical difference: the research agents now spend less time filtering and more time analyzing. The signal-to-noise ratio in the overnight briefs improved noticeably after the switch. Less "here's a listicle from 2023," more "here's the actual Reuters article from this morning."

The Flow End-to-End

Here's how a typical overnight cycle runs:

  1. ~10 PM PT — AIResearcher, TechResearcher, and FinanceCrawler activate, each pulling from Tavily and other sources for their domain
  2. ~11 PM PT — Agents send findings to OvernightAnalyst
  3. ~11:30 PM PT — OvernightAnalyst synthesizes and sends to StockPicker
  4. ~5:45 AM PT — StockPicker generates picks and delivers morning brief
  5. 6:00 AM PT — Brief delivered, 45 minutes before market open

The whole thing runs while I sleep. By the time I'm reading the brief over coffee, the agents have already done hours of work.

What Makes This Different from a News App

A news app shows you what happened. The research swarm shows you what happened and why it might matter to your specific portfolio and interests. The agents accumulate context — they know which sectors I follow, which macro themes I've been tracking, which positions I've asked about before. The briefings aren't generic market summaries; they're filtered and framed for my situation.

That personalization compounds over time. The agents that have been running for two weeks produce better output than they did on day one, not because the model improved, but because their memory improved.