---
title: "Youngjun Yoon (@cpcai) | Hashnode"
description: "Solo developer shipping AI and full-stack MVPs in 3-5 days | founder of CPC AI, Seoul. Read the latest articles by Youngjun Yoon on Hashnode."
url: "https://hashnode.com/@cpcai"
---

@cpcai

# Youngjun Yoon

@cpcai · Joined November 2026

Solo developer shipping AI and full-stack MVPs in 3-5 days | founder of CPC AI, Seoul

## Comments

[The template hash is the one I would take, and honestly, it is not in there today. The manifest records the render hash, which tells me the output changed but not why — a layout change and a font fallback look identical in that field. Putting the template's own hash beside it means the manifest can answer "which videos came from the version before the tweak" without diffing files. Adding it before the next layout change is the cheap moment to do it. On the voice model: it is not pinned, and I would not pretend the cache substitutes for pinning it. TTS is cached per text hash, so anything already shipped keeps its audio and a provider update cannot reach back and change it. What the cache cannot protect is the next render — if the model shifts between Monday and Friday, two videos from the same template can sound different and nothing in the manifest says so. Pinning the model id and recording it in the manifest is the same fix as the template hash, and it is the one I would add first. Your font point is the one that costs a whole night: a silent fallback is the worst failure mode in this pipeline. That is why the render step runs with the network blocked and dies on the first frame rather than producing a video in the wrong typeface.](https://hashnode.com/posts/from-headline-to-a-27-second-vertical-video-the-unattended-render-pipeline-behind-a-crypto-market-app/6af7c2d4ecb2109c71008c5e/comment/6af9414866f102f4b888c56a)

Comment · Article · 1d ago · [From headline to a 27-second vertical video: the unattended render pipeline behind a crypto market app](https://hashnode.com/posts/from-headline-to-a-27-second-vertical-video-the-unattended-render-pipeline-behind-a-crypto-market-app/6af7c2d4ecb2109c71008c5e)

[Thanks — that is the right question, and it is the part of this process I had to learn the hard way. I draw the line at what changes the definition of done, not at what surprises me. The quote covers the six checks; inside that boundary, anything I did not expect I absorb, and then add it to the checklist so the next quote does not miss it. Invoicing a client for my own incomplete checklist would be the wrong bill. What does trigger a re-scope is a surprise that moves the end state: the client does not own the Stripe or the DNS account, the auth model has to be rebuilt rather than finished, or the release itself depends on a third party — merchant verification, SMS sender registration, store review. Those are not hours I underestimated, they are a different job. I say so on day one with the evidence (the commit, the 401, the dashboard I cannot enter) so the client can judge the change instead of taking my word for it. Absorbing everything is tempting because repricing mid-project costs trust, not just time. That is why the five questions carry the weight: when I know who owns the accounts and what "done" means before I quote, the greps have far less room to surprise me. When one still does, I would rather re-scope early and honestly than quietly shrink the scope. Your framing of pricing the answer rather than the hours matches how I work. How do you handle the case where the definition of done is clear but the blocker sits with a third party — quote it as a range and hold the client to the same assumption, or stop at the boundary of what you control?](https://hashnode.com/posts/how-i-quote-a-rescue-job-in-30-minutes/6af8075b70ba34ed06c5962f/comment/6af93ed990de5c4b0bac390b)

Comment · Article · 1d ago · [How I quote a rescue job in 30 minutes without reading the whole codebase](https://hashnode.com/posts/how-i-quote-a-rescue-job-in-30-minutes/6af8075b70ba34ed06c5962f)
