# Ryan Kowalski, Senior Engineer at Meridian Financial (220 people, Chicago) — read of Model Selector, June 15, 2026

> Nine years writing backend services, two kids under five, and a notes app with forty-three half-finished side project ideas I never launched.

## How I got here

A guy in a Hacker News comment thread about "profitable micro-SaaS in the AI space" linked to what he called "the most honest idea marketplace I've seen." I clicked. I've been looking for something to build on train rides for about eight months. This came up when I searched "llm cost routing tool" on Google later that same day, which probably means their SEO is working, or Google is just starved for content in this niche.

## What I clicked first

The headline landed. "Cut your LLM costs 40 to 60 percent with intelligent model routing." Clear, specific, believable range. I've personally complained about our team's OpenAI bill to our VP twice this quarter. For half a second I thought I was about to buy a tool that does this thing. Then I kept reading and realized I'm being sold the *idea* of building that tool. That's a meaningful gear-shift the page doesn't warn you about until well past the fold.

## Where I paused

The scoring block. Specifically this line: "$-26,890 Year-1 take-home (Fermi)." Negative. They are telling me, in the hero section of a product pitch, that I will likely lose twenty-six thousand dollars in year one. Then directly below it: "1 in 8 Meaningful-success odds." I stopped for a solid ninety seconds here. Either this is performance honesty designed to make me trust everything else on the page, or they genuinely believe those numbers and put them front and center anyway. Both interpretations are interesting. Neither makes me feel great about handing over ninety-nine dollars.

## What I distrusted

The axes feel made up. "Buyer clarity: 10/10" and "credibility: 9/10" are great. But those are the easiest things to score well on for any idea with a clear problem statement. "Financial upside: 1/10" is the number that matters for an indie builder and it's a one. I don't know what the ten axes are or how they weight against each other. The page links to "How scoring works" but I didn't click through during my first read. I probably should have. The fact that I didn't suggests the scoring block isn't earning trust as fast as it's spending it.

Also: "We shipped the strategy package; you ship the customer conversations." That's doing a lot of work as a disclaimer. The gap between "strategy package" and "validated idea" is exactly where most indie projects die. I've lived in that gap.

## What would convince me

One thing: a builder who bought the $99 package, shipped something in the actual LLM routing space, and has real churn or retention numbers from even five paying customers. Not a testimonial quote. A Loom video, or even a Twitter/X thread with the account still up and replies from actual users. The reason the $-26,890 Fermi number stings is that there's nothing on the page to argue against it. If someone with a similar background to mine (engineer, no audience, no distribution) had turned this into $800 MRR in six months, that would reframe everything.

## What I'd ask in an email reply

1. The LLM routing space already has LiteLLM and a few others. What's the differentiated wedge here, and does the $99 dossier address that directly or is it generic GTM?
2. That $-26,890 year-one number, where does most of the loss come from? Is it time cost (i.e., opportunity cost of my hours), or are there real cash expenses assumed in the model?
3. You mention "no code changes needed" as a feature of the product concept. Did you actually prototype the middleware layer, or is that a UX assumption baked into the dossier?

## Verdict: on-the-fence

The honesty is real and I appreciate it, but "buyer clarity 10/10" combined with "financial upside 1/10" is a rough combination to sell. I'd want to read what the dossier actually contains before I'd pay even five dollars, and the page doesn't give me enough of a sample to know if it's sharp or generic.

---
*Memo by skeptic persona, generated 2026-06-15. Studio breaks own self-grading loop.*