# Marcus Chen, Staff Engineer at Rowspan Labs — read of learns-yo, May 20 2026

> 8 years in TypeScript, currently the guy who owns the PR queue at a 22-person B2B SaaS shop and watches half our engineering time disappear between Linear ticket and first commit.

## How I got here

Searched "typescript code generation from spec" because I was annoyed after watching a junior engineer spend four hours translating a 40-line Notion doc into a service file. One of the top results was a Reddit thread where someone linked this. I opened it between standup and lunch. I gave it about six minutes.

## What I clicked first

The hero line stopped me: "Pseudocode in. TypeScript that passes review on the first try." That is the exact sentence I would write if I were describing my problem to someone. I did not bounce. That is rare.

The code example below it is genuinely the best thing on the page. It shows a real-looking output: aliased imports, a real error type, a logger call, proper async. It is not Hello World. It looks like something a senior engineer at a real shop would have written. I stayed for another three minutes because of that single block.

## Where I paused

The spec table. "Context: Up to 200k tokens of repo context per translation." That is a real number. That is someone who actually thought about how large a codebase gets. And then right below it: "Privacy: BYO model key, self host the indexer, no training on your code." That combination, if it works, is actually the answer to the question my CTO would ask in the first thirty seconds.

## What I distrusted

Two things broke my trust hard.

First, the "Who this is for" copy reads: "Tutors with $500K+ revenue, training-program operators, certification course creators with a list." That is a different product's ICP copied onto this page. Someone forgot to change the template. That is not a minor typo. That tells me this is a factory process that is churning out idea pages and sometimes does not fully update them before publishing. It made me reread everything above with more suspicion.

Second, the pricing reveals the actual product: you are not buying software. You are buying a business plan. "Adopt the build: $99-$199. Dossier plus the working code starter, brand assets, copy library, and outreach pack." I came here thinking I was looking at a tool I could use next week. I am actually looking at a startup idea for sale. The page does not make that clear until you are halfway through it. The "Get API key" button at the top is not a free trial button. That is a mismatch between what the CTA implies and what the product actually is.

Also, "$-20,990 Year-1 take-home" displayed right on the product page is a strange editorial choice if you are trying to get someone to adopt this.

## What would convince me

I would want to see a screen recording of someone pasting an actual messy Notion spec, not a clean markdown file, and watching the output come back matched to a real repo with a real pile of legacy decisions. Not a hand-picked example. A live session with a real engineer who did not prep anything.

And I would want one design partner quote that says something specific, like "we stopped doing manual spec translation for a category of ticket and measured the delta." Not a testimonial that says "game changer." Numbers, with context.

## What I'd ask in an email reply

1. The code example in the hero matches a very clean project structure. What does the output look like when the codebase has mixed conventions, because we inherited half our imports from a contractor who did not follow our standards?

2. How does the indexer handle monorepos where the conventions vary by package?

3. Is there actually a working API I can hit right now, or is this still pre-build, and if I pay the $99 adopt fee am I getting a starter template or a deployed service?

## Verdict: on-the-fence

The core idea is real and the code example is honest. But the mismatched ICP copy and the late reveal that this is an idea-for-sale rather than a tool-for-use are both problems that I would need answers to before forwarding this to my CTO. If someone emailed me and addressed those two things directly, I would probably take a thirty-minute call.

---
*Memo by skeptic persona, generated 2026-05-20. Studio breaks own self-grading loop.*