# Priya Nair, Director of Product at Loopfield (62 employees, B2B SaaS) — read of Adaptive AI Companion, June 14 2026

> 9 years in product, most of it at companies where I was building the AI roadmap before anyone called it that. Currently evaluating whether I leave to build something of my own.

## How I got here

Lenny's newsletter had a link to "honest startup ideas" or something like that. The framing was unusual enough that I clicked. I was on the train, had 12 minutes before my stop. This is the kind of page I read with one eyebrow up.

## What I clicked first

The hero copy "An AI That Actually Knows You" is fine, not offensive. But the thing that actually pulled me in was the scoring table. "$-11,780 Year-1 take-home (Fermi)" and "1 in 6 Meaningful-success odds" sitting right on the page, not buried -- that is not how most idea marketplaces talk. I stopped scrolling. That's the hook.

## Where I paused

"After 10 conversations it knows you better. After 100, it anticipates what you need."

I sat with that for a bit. That's a real claim about a behavior that's very hard to verify from a product page. I want to believe it. The problem is ChatGPT already has memory, Claude has Projects, Gemini is working on it. All of those are free or bundled with something I already pay for. So the question I'm actually asking is: what does "better" mean here, and what is the switching cost of learning this versus just using what I have?

## What I distrusted

Two things. First: "Psychological Model Building." That phrase is doing a lot of work and I don't know what it means operationally. Is this a graph of topics I've mentioned? Is it actual vector embeddings? Is it a structured JSON object I could inspect? The page doesn't say. "Psychological" sounds clinical in a way that makes me think the copy is covering for something that's actually pretty simple underneath.

Second: there's a "Try it Live" section in the nav and what looks like a live result demo on the page, but from what I can read in the text dump, I have no idea what that actually shows. If there's a real live demo and it works, that would completely change how I read this page. If it's a canned response or a screenshot, that's the kind of thing I notice.

## What would convince me

A head-to-head comparison. Specifically: someone who uses ChatGPT memory AND this product for 30 days and writes down, in plain language, the three moments where this one did something the other didn't. Not a feature list. A diary entry. Two paragraphs.

Also: the "financial upside: 2/10" score -- I'd want to understand the assumptions behind that. If the $-11,780 Year-1 number is based on a specific pricing model and customer acquisition cost, I want to see the math. The page says "Fermi" which I respect, but a Fermi estimate is only as good as its inputs, and those are invisible here.

## What I'd ask in an email reply

1. ChatGPT just updated their memory to store way more context and it's included in the plan I already pay for. What does your product do in month 3 that ChatGPT memory doesn't?

2. The "Portable Profiles" feature says I can export my personality model and use it across AI systems. Has anyone actually done this with a non-Wishdeal AI? What does that handoff look like technically?

3. The $99 tier includes "working code starter." What stack, and how far does it actually run before I need to bring in real engineering?

## Verdict: on-the-fence

The transparency about negative returns and low success odds is genuinely unusual and I respect it. But I'm buying an idea package in a space where the infrastructure players are giving away the core behavior for free, and the pain score (4/10) lines up with my honest read of my own frustration with current AI memory -- it's real, but it's not the thing keeping me up at night.

---
*Memo by skeptic persona, generated 2026-06-14. Studio breaks own self-grading loop.*