# Marcus Okafor, Staff Engineer (Platform Infra) at Meridian Payments — read of codebase-indexing-saas, June 24 2026

> 14 years in software, last 3 in LLM tooling. We have 90 engineers and a RAG pipeline we built ourselves in 2024 that works about 60% of the time and everyone hates it.

## How I got here

Searched "semantic code search API reduce context window tokens" Friday night. I do this on Friday nights because Saturday mornings I'm coaching my kid's U10 basketball practice and I have zero headspace. One of the results pointed here. I assumed I was about to see a product I could sign up for.

## What I clicked first

The hero pulled me in immediately. "Query any codebase like a database. 99% fewer tokens." That is exactly the sentence I want to read. Our homegrown pipeline sends 80k tokens to Claude for what should be a 3k-token answer and we pay for it every day. So I kept reading.

"Cross-File Dependency Mapping. Understand how changes ripple through your system." Also real. I've wanted this for 18 months. "Ray Tracing for Agents. When your AI agent makes a coding decision, trace through all dependencies in milliseconds to predict side effects." Okay, now I'm leaning forward.

## Where I paused

Right here: "Honest disclosure: we don't have live customers on this idea yet. We shipped the strategy package; you ship the customer conversations."

I read that three times. Then I scrolled back to the top. Then I looked at the nav: "For FAQ Pricing Honest Hire team to build." Then I looked at the pricing: "Browse Free. Unlock the dossier $5. Adopt the build $99-$199."

This is not a product. This is a startup idea for sale. They are selling me the business plan for a product I came here to buy. The thing I want does not exist yet. The "Get API key" button in the nav goes nowhere useful I'd assume, because there is no API. There's a dossier.

## What I distrusted

The self-scoring is deeply strange on a product page. "61/100 Adoptability. financial upside: 1/10." They are telling me, in the hero area where a normal product would put a testimonial, that their own idea has terrible financial upside. I respect the transparency in theory but it's disorienting when I came here thinking I was buying a subscription.

"99% fewer tokens" appears as a bullet point with zero methodology. Is that compared to sending the whole repo? Compared to their own baseline? Compared to what we're doing? That number means nothing without one sentence of context, and they didn't include it.

The whole "Wishdeal Factory scores every idea against 10 Adoptability axes" framing reads like a meta-layer that made sense to whoever built this site but is genuinely confusing if you arrived thinking you were evaluating a product. I had to reorient my entire mental model halfway down the page.

## What would convince me

If this were actually a product: one real customer quote with a company name and a token reduction before/after. Not a percentage. A number. "We were sending 140k tokens per agent call. Now we send 9k." That's it. That's the whole testimonial I need.

If it's an idea-sale (which is what it is): show me one person who bought the dossier and built something with it. A tweet, a URL, a GitHub repo, anything. Right now I'm being asked to trust a Fermi estimate with a negative projected income from a studio that is openly grading its own ideas.

## What I'd ask in an email reply

1. Is there any version of this that I can actually call today, even in beta or rough form, or is the product literally the strategy doc?
2. The "Ray Tracing for Agents" feature -- is that a novel indexing approach or is that a description of something that already exists (like tree-sitter plus a graph traversal)?
3. Who is the "$99 adopt" tier for -- someone who wants to build this business, or someone who wants to use the software? Because those are completely different people and I am only one of them.

## Verdict: on-the-fence

The underlying idea is one of the most clearly defined problems I've seen pitched in this space, and someone should build this. But I came here as a buyer of software and found a buyer of founders. If I were looking to start a company, I'd be curious. I'm not, so I'll probably close this tab and be mildly annoyed for about 20 minutes.

---
*Memo by skeptic persona, generated 2026-06-24. Studio breaks own self-grading loop.*