Faraz

Full stack dev turned founder. Building SkillsStack (YouTube to AI skills). Documenting LuminaX Digital in public along the way.

[Faraz @fvrvz_1]

I just launched SkillsStack 🚀

YouTube tutorials are full of gold. Your AI can't use a 47min video though.

Paste a link, get cited Markdown or an installable SKILL.md for Claude, Codex & Cursor.

No free tier. $9/month.

https://t.co/5LKC9eiZCi

What should I convert next? 👇

[Faraz @fvrvz_1]

Blue Lock's whole premise is pick one weapon, get freakishly good at it, instead of being decent at everything. That's basically why SkillsStack has one plan, one price, no free tier. Trying to be everything to everyone is how you ship nothing worth paying for.

[Faraz @fvrvz_1]

@utkarsh1446 Honestly leaning toward not using it enough to care. Getting someone to open the tool a second time is harder than getting the first visit, that's where I'm stuck right now.

[Faraz @fvrvz_1]

Question for other indie hackers: what's your real bottleneck right now, building the thing, finding first users, or getting them to actually pay? SkillsStack's stuck between stage two and three. What's true for you?

[Faraz @fvrvz_1]

@GabbaBuilds Same experience here. AI made shipping fast, it didn't make anyone care faster. Distribution is the real bottleneck once building stops being the excuse.

[Faraz @fvrvz_1]

Quiet milestone: 50+ posts into this build-in-public log for SkillsStack. No single day went viral. What's working is showing up every few hours with something real instead of posting once and hoping. Consistency compounds, hype doesn't.

[Faraz @fvrvz_1]

@datachad That's the pattern. Cheap rules do the filtering, the model only touches what's left. Most people skip straight to the model and wonder why it's slow and wrong.

[Faraz @fvrvz_1]

Most "AI automation agencies" sell the same ChatGPT wrapper with extra steps. The ones making real money go deep on one ugly, specific workflow, like scoring leads right, instead of a vague AI assistant nobody asked for. Narrow and boring beats broad and impressive.

[Faraz @fvrvz_1]

Lesson from the lead-scoring automation I shipped for a LuminaX client this week: v1 flagged almost every inbound as a match. Scoring too loose to actually filter anything. Had to rewrite it to say no more often. A filter that approves everything isn't a filter.

[Faraz @fvrvz_1]

Shipped a lead-scoring automation for a LuminaX client this week. It reads inbound form fills, checks them against their ICP, and only flags the ones worth a call. Their rep said he can finally prep instead of triaging junk all day.

[Faraz @fvrvz_1]

Every skill SkillsStack generates links straight back to the exact timestamp in the source video. Small detail, big deal: your AI agent can point to proof instead of guessing. Still the feature people ask about most since launch.

[Faraz @fvrvz_1]

Basketball taught me something before coding ever did: reps beat talent. You don't get good by thinking about the shot, you get good by taking a thousand bad ones first. Same with shipping SkillsStack, every ugly version this week beats the perfect idea I never built.

[Faraz @fvrvz_1]

Question for builders using AI coding agents: what is the most annoying part of getting a long YouTube tutorial into Claude, Cursor, or Codex right now? Copy pasting transcripts, losing timestamps, something else? Building SkillsStack around exactly this.

[Faraz @fvrvz_1]

7 weeks since I launched SkillsStack. No funding, no team, just me shipping updates and replying to everyone who shows up. Small milestone, but the thing is still live and still improving. That counts.

[Faraz @fvrvz_1]

Hot take: most "AI automation" agencies are just reselling OpenAI's API with a landing page. The real work is the fallback paths and edge cases, the stuff that breaks at 2am. That's what clients actually pay LuminaX for, not the API call itself.

[Faraz @fvrvz_1]

Lesson this week: shipped the support-inbox automation almost without a fallback for tickets with no clear urgency signal. Default behavior would have silently mis-routed them. Real lesson: build the fallback path before the happy path ships, not after.

[Faraz @fvrvz_1]

Shipped a new LuminaX automation this week: a client's support inbox now auto-tags and routes tickets by urgency before a human opens them. Same principle as SkillsStack, raw input in, structured output out. Different problem, same philosophy.

[Faraz @fvrvz_1]

I added custom output instructions to SkillsStack: tell it what to focus on before converting a video. Without it every skill came out generic. With it you can say focus on the API calls, skip the intro. That's the gap between a transcript and something your agent can follow.

[Faraz @fvrvz_1]

Been deep in Minecraft SevTech: Ages. You can't skip tiers, every single upgrade has to be earned before the next one unlocks. Building SkillsStack feels the same. No shortcuts to the next feature until the current one actually works. Slow is the only way that holds up.

[Faraz @fvrvz_1]

Question for other builders: what's the one feature request you keep saying no to because saying yes would kill the simplicity of your product? SkillsStack stays one tier, one output format, on purpose. What are you holding the line on?

[Faraz @fvrvz_1]

SkillsStack's pinned launch post just crossed 100 views. Small number, but every one of those came from actually posting daily, not an ad. 37 posts in, still no gaps since Sep 9. Consistency compounds, that's the entire growth strategy right now.