Everyone I talk to is asking the same question right now. Is this market too hot?

Positioning decides how fragile a rally is. When everyone is already long, there is nobody left to buy.

But how would you actually measure that? Is the crowd all-in or still hedged? Is the bullishness broad, or stuffed into a few names?

You can feel a hot market. Proving it with data is the hard part.

So I had Claude Code build me a full market positioning study off an options flow API.

I built a market-positioning scanner. It reads options activity to measure how bullish or bearish investors are and identify which trades are becoming crowded. Think of it as a market thermometer.

I used Unusual Whales for my setup. Any options data API with flow and volatility endpoints works, like Polygon or Cboe DataShop.

Here’s the plan:

1. Pull the 2026 options tape for the whole market
2. Score what the crowd is buying
3. Score what protection costs
4. Rank today against every day this year

The Prompt

I started with a fresh Claude Code session and one “goal-based” prompt.

The prompt sets the goal, defines the data workflow, and tells the agent to check what the API can actually answer before designing the study.

Full prompt here: Market_Positioning_Study_Prompt.pdf​

If you want to customize it, swap in your own universe or your own question. The structure holds for any options API.

You need two things: Claude Code and an options data subscription with API access. The full study used about 4,500 API calls, inside a single day’s allowance on my plan.

The Result

Full report here: Market_Positioning_Consensus.pdf. It came back at 103 pages.

The short version: the market flipped from cautious to bullish in the last week of August, and the study called it overly bullish on August 26.

Here is what that means in practice:

• If you are adding risk here, you are joining a crowd that arrived last week. Traders spent most of August cautious, then chased upside for six straight sessions. The move you are buying into is already positioned for.
• Hedges are the cheapest they have been all year. Nobody is paying for downside protection right now, so if you have been waiting to hedge a big position, the market is practically giving it away. Cheap insurance while everyone piles in is the classic complacency setup.
• This exact setup marked the two tops of 2026. The study flagged “everyone bullish” twice before this year: late January, right before a 9% drawdown, and early August, right before a fade. Both times, chasing was the wrong move over the following weeks.

None of that says the market must fall. It says the easy part of this move already happened, and the crowd is leaning one way with no seatbelt on.

I also had the agent zoom into semiconductors, since that is where everyone’s book is concentrated.

Two practical reads there:

• Memory is the most crowded semiconductor cluster, while $TSM is the most crowded individual name. The memory names ($MU, $WDC, $SNDK, $STX) carry more bullish positioning than the AI logic names. If you own the memory supercycle, know that the options market already owns it with you.
• Interestingly, $NVDA is the least crowded of all 14 chip names in the study. Everyone assumes NVDA is the consensus trade. By positioning, the crowd has moved on to the derivatives of the AI story and left the original behind.

Here’s the bigger takeaway. A study like this used to require a quant team and a serious data budget.

The models are commoditized now. What is scarce is the data your agent can reach. My entire edge in this study was one API subscription.

Research capacity is becoming a function of which feeds you can query.

The funds building proprietary data pipes understand this, and the advantage compounds.

I want to note the limits. This is one data vendor, one year of history, and positioning describes where the crowd stands today. It is a weak forecaster on its own, and none of this is investment advice.

Run the prompt against your own coverage universe and see where your names rank. Let’s check back in a month on whether the August 26 signal meant anything.

Personal

I’m making these API-driven studies a regular rotation in the letter.

If there’s a dataset you want the agents pointed at next (13F filings, credit, prediction markets), hit the reply button. I read everything.