Guilherme Anjos @GuiOSlab

60 seconds, zero humans: a lead submits a form, Claude scores it 10/10 hot, writes the reply in the owner's voice, books the call, logs everything. This is n8n + one well-designed prompt. No vector DB, no fine-tuning — the model returns strict JSON and the workflow branches on it. Watch it run ↓ Open-source workflow + ready-to-import version: …ad-agent.guilhermecbanjos.workers.dev

Most small businesses lose leads to one thing: slow follow-up. Whoever replies first usually wins. So I built an AI agent in n8n + Claude that answers, qualifies, and books every inbound lead in under 2 minutes — 24/7. • Scores each lead 1–10 (hot/warm/cold) with Claude • Auto-replies to the good ones in your voice, with your booking link • Logs the rest for manual review — never spams Runs on your own accounts, pennies per lead. Ready-to-import template + a 15-min setup guide: …ad-agent.guilhermecbanjos.workers.dev #n8n #automation #AI #nocode

what's your MRR and how many months did it take to get there? i'll go first: $0, month 11. normalize the slow part.

founders: go ask ChatGPT "what are the best tools for [your category]" right now. are you in the answer? reply with yes/no. curious how bad this actually is.

This was my score :) Im super super happy, from 0 to 43 in about a month is a huge deal! And this is the live proof that Vaulto does work!

i built vaulto to see if AI actually recommends your product. then i made myself guinea pig zero. first scan of my own site: 0/100. invisible. i posted the screenshot anyway. been running my own fix kit since. today: 43/100. AI partially knows me now. slow, compounding, real. the fixes work.

if you're building something, scan your own site free, no signup: vaulto.sh/scan

i built a tool that scores how visible your site is to AI. first thing i did was scan my own site. 0 out of 100. i sat there for a minute. then i screenshotted it and blogged about it. if my own scanner roasts me, at least i know it's honest.

@Suzaan2302 both, honestly. the next jump: claude can write your description, but will it recommend your product when a buyer asks it "best X for Y" cold, without you naming the brand? that's the part most stores haven't checked yet. have you tried asking it that way?

it's less bias than citation density. AI recommends whatever shows up most in the english dev content it trained on. resend/brevo are all over indie-hacker blogs; zoho's transactional arm just has fewer of those mentions despite being solid. are you seeing this across categories, or mostly dev tooling?

45 days ago i started building a marketplace for buying and selling startups. deal rooms, escrow, all of it. 3 weeks in, i deleted the entire thing. 658 commits later it's an AI visibility scanner instead. the hardest part of building solo isn't the code. it's admitting the thing you spent a month on was the wrong thing.

@kiranvoleti transactable sits on top of cited, not instead of it. the agent still has to pick you before it can act, so citation is upstream. the new bar is machine-actionable: clear pricing, no human-only checkout. seeing agents skip cited-but-not-actionable sites yet?

@ImNotAV1rus narrow + connected is how you win early citations, and it makes you trackable: pick the 5 prompts those traders actually type, then watch weekly if AI names you or a competitor. that's a feature in vaulto.sh, free scan to see where you stand now?

@TerekhinIvan the "chatgpt recommends new domains day one" bit is the one i'd replay. no backlinks needed because it isn't pagerank, it surfaces the clearest semantic match. did those day-one wins hold, or decay once competitors showed up?

@AjayGB4 the split is real, but #2 is a different beast. crawling gets you in the index; citation needs the engine to pick you as the best source. submission alone can't force that. does IndexerNow stop at pushing the crawl, or does it surface why a URL still isn't getting cited?

@DariaStrategy @decentwealthapp the "listen to users" answer is the real AEO moat. chatgpt recommends what people genuinely say about you on reddit and reviews, so trust-first work IS the optimization, not a softer alternative. how do you separate an earned mention from a gamed one when you track it?

@nitish_fixaeo this kills the "one fix ranks you everywhere" pitch. the engines pull from different places: perplexity leans live web + reddit, chatgpt leans corpus + bing. per-engine work, not universal. do you tell clients to pick one engine and go deep, or spread across all four?

@LocalFalcon the hard part is non-determinism. ask chatgpt the same question twice and you can get two different brands, so a single check is noise. we sample each prompt 5-10x and track frequency-of-mention instead of a rank. are you doing repeated sampling, or one-shot checks?

@adamkillam the agent split is the part most people flatten. when the reader is an agent, not a human, marketing copy stops mattering. agents parse docs, schemas, machine-readable claims. do you optimize the same pages for both, or build a separate machine-readable surface for agents?

@altivai @ahrefs signal-vs-noise is the whole game here. most AEO "facts" are one person's anecdote repackaged as law. tracing evidence is the right wedge. where do you draw the line to publish, a minimum number of independent sources, or do you re-test it yourselves?