The AI trade has been selling off hard recently. Alongside it, rumors have been circulating about funds unwinding positions, including VARA.

It raises a useful question if you manage a concentrated portfolio:

How much do I own relative to how much I could reasonably sell each day?

You probably know your biggest positions. You know your sector exposures. But how long would it take to reduce those positions without accounting for a huge share of daily trading?

This illiquidity factor can augment volatility which is critical to understand in times like now.

So this week I'll show you a useful trick in Claude Code to manage this risk.

I used AI to build a dashboard that works through that question using VARA’s disclosed equity holdings.

Check out the dashboard here to follow along: https://fasinisterra.github.io/vara-liquidity-risk/​

What stood out? I'll walk you through it below:

Was this email forwarded?
Subscribe to get more AI investing breakdowns just like this!

A smaller position can take longer to sell

I started with Value Aligned Research Advisors’ 13F holdings as of June 30, 2026 (yes this is bit stale but I recommend you doing this on your live portfolio)

For each position, I asked: how long would a full exit take if I traded at 5% of its typical daily dollar volume?

Three holdings make the point for VARA's portfolio:

• Micron: about $644 million, less than one trading day.
• Bloom Energy: about $1.24 billion, roughly seven trading days.
• Comfort Systems: about $351 million, roughly fourteen trading days.

Comfort Systems is the smallest position here. Under the same assumptions, it takes about twice as long to sell as the much larger Bloom Energy position.

That changes how I’d review the portfolio. Sorting holdings by dollars alone would put the biggest names at the top of my list. Comparing those dollars with trading volume brings a different set of positions into focus.

Under times of margin call or distress, this illiquidity can result in greater fire sales.

Where I’d look more closely

Nine positions take more than ten trading days in this scenario. Together, they represent about 16% of the value of the reconstructed equity portfolio.

Most of the portfolio sits below that threshold. The useful finding is the group of positions where an exit could take longer.

If I were running this on my own holdings, I’d want to know whether those timelines fit my cash needs, financing and willingness to keep holding through a drawdown.

I’d also test a partial reduction. Selling a quarter of a position is a different problem from exiting the whole thing.

What happens when the assumptions change?

Next, I halved the assumed daily dollar volume while keeping participation at 5%.

The share of the portfolio in positions taking more than ten days rises from about 16% to 48%.

How I’d replicate this if I were you

Start with a holdings file that includes share counts. Have AI match each security to its price and volume history, then calculate position size against daily dollar volume.

From there, vary the assumptions:

• How much of each position do I want to sell?
• What share of daily trading am I comfortable taking?
• What happens if less volume is available?

AI makes it easier to assemble the data and run these comparisons.

For me, the useful question is: which positions would take longer to reduce than I’m comfortable with?

You can explore the assumptions in the dashboard below.

Explore the liquidity dashboard: https://fasinisterra.github.io/vara-liquidity-risk/​

-Felipe

Personal Updates

I'm in Australia right now, running AI training sessions for our corporate strategy clients.

We've made a lot of progress in how we use AI for investor relations, particularly in checking whether the story a company tells lines up with its numbers.

Let me know if you're around Sydney!