Specialist agents, not one voice
Market, news and macro, fundamentals, and attention each have their own agent. They write separately before anyone synthesizes a thesis.
WebserviceX
Agentic AI equity researchfor the S&P 100.
Specialist agents take seats, disagree, and hand work forward. A 12-month investment view, a nearer overlay, and one five-rung published rating. Dated notes — not a chatbot wrap, and not a brokerage.
Press / to search a ticker
Coverage
Coverage updating · batch 2026-08-14
- Covered
- 0/101
- Full-desk
- 101
- QC pass
- 91%
- 5D tactical
- —
- Batch notes
- 101
Full desk
Eight agentic steps. One note.
01
The name
A company, on a date
02
Four agents
Market · news · fundamentals · attention
03
Disagreement
A case for, and a case against
04
Year-long
The owner’s view
05
Nearer
The next several weeks
06
Risk
A challenge before anything ships
07
Published
The rating on the note
08
Fit to show
Unfinished writing does not go out
Agentic AI
Specialist agents on an equity-research desk
Agentic here means more than one model completion. Distinct agents take seats, act on a shared as-of picture, hand work forward, and a published note is the end of the chain — thesis, catalysts, invalidation, and a five-rung rating.
They act, then they hand off
Each agent completes a job and passes evidence forward. A research manager, trader, risk seats, and a portfolio manager continue the chain — that is the agentic part.
Disagreement is required
Bull and bear agents argue the long and the underweight case from the same pack. A rating that skips that step is not presented as a full desk.
A published note, not a chat log
The output is a dated equity-research note: thesis, catalysts, invalidation, and one of five ratings. Unfinished agent chatter does not ship.
Dual horizon
Research Manager, Trader, then Portfolio Manager
After the specialist agents, three desk agents own the horizons. That hand-off is the agentic chain; the language is equity research.
12-month investment view
Research Manager
The structural owner’s view: what must be true over the next year. Not what the public record scores.
Tactical overlay (about 2–8 weeks)
Trader
The nearer stance: setup and unwind. This is the view the public record looks back at.
Published desk rating
Portfolio Manager
The rating on the note. Reconciles the two horizons and any risk challenge. Stated as one of the five rungs.
What a note contains
The anatomy of a published note
As-of date
The session the note is written for. Later knowledge is out of bounds.
Thesis
What must be true for the year-long view.
Evidence
Dated facts from price, news, the business, and public attention — not invented figures.
Catalysts
What could change the picture next, if the pack names it.
Invalidation
What would kill the thesis. A note without this is unfinished.
Published rating
One of BUY, OVERWEIGHT, HOLD, UNDERWEIGHT, or SELL — stated plainly.
Rating scale
Five rungs, used everywhere
The desk does not use Strong Buy or Strong Sell. Every published view is one of these five tokens.
| Rating | Meaning |
|---|---|
| BUY | High-conviction add. The desk expects material outperformance over a longer horizon. |
| OVERWEIGHT | Favorable risk/reward. Add versus a relevant equity benchmark. |
| HOLD | Fair risk/reward. Maintain; no new initiative. Unscored on the public record. |
| UNDERWEIGHT | Unfavorable. Trim versus the benchmark. |
| SELL | Exit or avoid. Material downside, or the thesis has broken. |
Built to be inspected
Not a chatbot wrap—an agentic research desk
Dated notes
Each note keeps the day it was written. Earlier dates stay readable beside the latest.
Fit to publish
Writing that still sounds like homework does not go out. A note must state a clear view.
A later look
We look back at the nearer view over a short stretch of sessions. Incomplete runs are left out. This is not a grade of the year-long thesis.
Educational purpose only. Not investment advice.