Todd

Building AI, Optimizing Life.

Consistency is greater than intensity.
Questions are better than answers.
Living beats a resume.

[Todd @piechow]

is @vercel this bad at customer support?

I have dozens of projects in my account. Paying customer for over a year.

Account inaccessible. totally locked out due to not getting a 2FA.

I've opened 5 support tickets over the span of a month. Not one reply.

[Todd @piechow]

CBL surrendered three more retail properties this week. CMBS on each. Six surrenders in recent months from one operator.

Every buyer reading a retail OM right now has seen those headlines. They're categorizing your asset before they read page one.

[Todd @piechow]

Amazon is bragging that Rufus Auto Buy users save 20% per purchase.

That 20% comes from your sale prices.

Thread on what the Spring Sale (March 25) actually sets up for Q2 👇

[Todd @piechow]

Stanford and BetterUp studied AI slop in the workplace. They named it 'workslop.'

40% of desk workers got it from a colleague last month.

Cost: $186/employee/month. $9M/year at 10K employees.

That's just processing it. Here's the part that actually matters 🧵

[Todd @piechow]

Amazon's Big Spring Sale starts March 25.

So do Rufus ad charges.

Same morning. Most brands have only one of these on their calendar.

[Todd @piechow]

The industrial lease spread was $2.20/SF a year ago. Now it's $1.28.

Most industrial offering memorandums haven't caught up. Here's what that means for brokers right now:

[Todd @piechow]

Amazon's FBA commingling ends March 31. Most brands are focused on the FNSKU logistics.

The Amazon Rufus connection nobody's making: 🧵

[Todd @piechow]

Gmail now scores email content for quality before deciding where it lands.

Gemini reads your first 100–200 characters. The AI slop patterns that cost you reader trust now also cost you inbox placement.

Thread on what changed and what to check 👇

[Todd @piechow]

Sale-leaseback volume hit $14.4B in 2025. Q4 was up 56% quarter over quarter. PE firms are the driver — SLBs lag M&A activity by 6-12 months, and the 2025 deal wave is feeding into the market right now.

[Todd @piechow]

DoubleVerify found a $40K/month AI content farm and they left their LLM system prompt in the source code.

The instructions: front-load sensation. Inject emotion. Manufacture urgency.

Those are Sloppy Joe flags. Written as a to-do list.

Thread 🧵

[Todd @piechow]

71% of single-tenant net lease transactions last year were done by private buyers — 1031 exchangers, family offices, HNW individuals. By deal count, per Marcus & Millichap's 2025 STNL report.

Most net lease OMs are built for the other 29%.

[Todd @piechow]

The 50/70 rule: let AI write 50-70% of your content, edit the rest, and you've solved the AI slop problem.

You haven't. Thread on why.

[Todd @piechow]

$936B in CRE loans mature in 2026. 19% more than last year. Multifamily alone is up 56%.

That's a lot of deals hitting the market where the seller isn't there by choice.

Here's what changes in the offering memorandum when that's the context. 🧵

[Todd @piechow]

When shoppers ask Rufus "Is this durable?" it doesn't look at your listing.

It reads your reviews.

Amazon Rufus has 4 distinct retrieval modes. Most brands are only optimizing for one. 🧵

[Todd @piechow]

Investors run a stress test on every CRE deal they seriously consider — whether you give them one or not.

If your offering memorandum doesn't include a sensitivity table, they build it themselves. With their assumptions. Not yours.

Why most OMs skip it (and why that's backwards): https://t.co/SJwNtEbM7F

[Todd @piechow]

Amazon Rufus is flagging your negative review themes to shoppers — even when your listing says nothing about them.

[Todd @piechow]

OpenAI trained the em dash out of GPT-5.1 after it became known as "the ChatGPT dash."

The vocabulary moves. The structures don't. Here's why that matters for your writing:

[Todd @piechow]

Buyer-side AI is now screening offering memorandums before analysts open them. PropRise cuts screen-and-kill from 30 min → 10. Dealpath extracts financials in 60 seconds.

[Todd @piechow]

Amazon Rufus returns the same 2–3 products per keyword. Nearly every time.

Someone holds those slots in your category right now.

A 1,300-product study just revealed exactly what those products have in common. 🧵

[Todd @piechow]

GPTZero's detection rate dropped from 70% to 18%.

Not because AI got smarter — because someone built a humanizer. One extra step. Detector goes blind.

Thread on why statistical AI detection is breaking, and what actually holds up 👇