# Derek Okonkwo, Senior SWE at Carta (side-project builder) — read of Error Log Condenser, June 9 2026

> 9 years backend (Python, Node, k8s at scale), two failed micro-SaaS attempts, currently trying again. 6-year-old daughter, so my build window is 9:30-11pm every night after she's down.

## How I got here

Saw someone in the Indie Hackers Slack drop a link with the comment "Wishdeal Factory is building ideas as packages now, interesting model." Clicked out of curiosity because I've been sniffing around dev-tools niches for months. I was not searching for a log compressor. I already want to build something. This felt like it might be a shortcut.

## What I clicked first

The problem framing hit immediately. "The AI spends 35% of the response tokens regurgitating your own logs back at you." That's accurate. I've watched Claude literally paste back my entire traceback before it says anything useful. But then I read further and got confused about what I'm actually looking at. Is this the product, or am I being sold the idea to build the product? It took me longer than it should have to realize the answer is: both, sort of, neither.

## Where I paused

The math block. "$600/month of pure noise" on 50 debug cycles per week. I actually stopped and did the multiplication myself. 50 cycles, 20K tokens saved per cycle, $0.03/1K. That's... $30/month, not $600. The $600 figure requires 1,000 debug cycles per week, which is 20 per business day, which is not one team, it's a department. The math is labeled "why it matters" but it quietly doubled somewhere. I didn't trust anything quantitative after that.

## What I distrusted

The page opens as if I'm a developer who needs this tool right now, then halfway through reveals I'm being sold a dossier. "We don't have live customers on this idea yet" buried below a convincing product pitch is a weird move. The hero copy is written to a developer debugging a production incident. The pricing section is written to a founder who wants to build a business. These are two different people with two different problems and the page tries to be both simultaneously. The "Wishdeal Factory scores every idea against 10 Adoptability axes" section lands like a gear shift with no clutch.

The "open source" mention also goes nowhere. No GitHub link, no repo name, nothing. That's the single most important trust signal for a developer-tools product and it's just... a bullet point.

## What would convince me

The $99 "adopt the build" tier says I get "working code starter." I need to see what that actually is before I'd pay anything. A GitHub link to the actual repo, even a half-baked one, would tell me more in 30 seconds than the entire page. Specifically: what language, how many lines, is the rule engine actually extensible or is "customizable rules" just a config JSON. Show me one rule definition. That's the product. That's what I'm buying.

For the business idea angle: one real number from a real person. Not a Fermi estimate, a Slack message from a developer who tried it and said "I saved X on my OpenAI bill last month." Even anecdotal. The "honest disclosure" section is genuinely refreshing but it stops right at the edge of useful.

## What I'd ask in an email reply

1. The page says "open source" but there's no link. Where's the repo, and what's the actual license? Is the rule engine shipped code or is that the part I'd be building?
2. The $99 tier gives me "working code starter." What does done look like? Is this a CLI tool, a browser extension, a web app? What stack?
3. Your strongest axis score is "buyer clarity: 10/10" but I, the buyer, am not clear on what I'm buying. Who is the actual customer you're imagining: the developer who wants the tool, or the founder who builds and sells it?

## Verdict: on-the-fence

The underlying problem is real and I've felt it personally. But the page is trying to do two jobs at once and it does neither cleanly enough to get my $99 tonight. I'd reply if there was a GitHub link I could inspect first.

---
*Memo by skeptic persona, generated 2026-06-09. Studio breaks own self-grading loop.*