Jul 8, 2025 · 25m · top-founders

My Agent Army Brings Me $500,000 Per Month (Exploding SaaS)

Romain Torres · 15m spoken Nathan Latka · 6m spoken
0:00 / 0:00

gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions

Host Nathan Latka interviews Arcads founder Romain Torres to uncover how his bootstrapped AI video advertising startup scaled to $6 million ARR with just six employees and a fleet of over 100 autonomous AI agents.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Nathan holds 28.3% of the talking time here. How this is scored →

Nathan as informed peer 5.6 Guest teaching 4.9 Guest disagreement 1.3 Nathan pushing back 2.4
05100:0010:0020:000:52–4:16 · Nathan as informed peer 5/10 Scaling to $6 Million ARR with a Tiny Team Nathan digs into Romain's ARR metrics and go-to-market channels. Romain explains that contrary to Nathan's assumption about physical UGC brands, most customers sell digital products and mobile apps.4:17–7:39 · Nathan as informed peer 5/10 Building Automated AI Outbound and Sourcing Agents Nathan inquires about outbound tooling, prompting Romain to explain how they built custom LLM agents to replace manual sourcing and outreach. Nathan presses on exact launch revenue figures.7:39–10:22 · Nathan as informed peer 5/10 The Origin Story: Building the WeFast Mobile App Nathan asks how Romain entered the women's fasting app market with WeFast. Romain walks through market research, community dynamics, and paid acquisition mastery that led to Arkads.10:22–14:23 · Nathan as informed peer 6/10 Bootstrapping and Maintaining High Profit Margins Nathan repeatedly attempts to pin Romain down on exact profit margins and ad spend. Romain resists giving concrete percentages while describing experimental multi-account organic TikTok automation.14:23–20:01 · Nathan as informed peer 5/10 Live Demonstration: Building AI Growth Agents in Gumloop Romain shares his screen to demonstrate how he builds autonomous growth agents in Gumloop, educating Nathan and the audience on node architectures and few-shot tweet generation.20:02–23:42 · Nathan as informed peer 7/10 Advanced Autonomous Agents: Competitor Scraping and Hooks Nathan demonstrates marketing expertise by proposing a modular copywriting framework (hook, body, CTA). Romain responds with his alternative perspective on how LLMs perform best on full-sample duplication.23:43–25:16 · Nathan as informed peer 6/10 Final Growth Insights and Interview Conclusion Nathan summarizes the core metrics, business evolution, and agent infrastructure of Arkads in a rapid-fire outro recap while Romain shares final thoughts.0:52–4:16 · Guest teaching 4/10 Scaling to $6 Million ARR with a Tiny Team Nathan digs into Romain's ARR metrics and go-to-market channels. Romain explains that contrary to Nathan's assumption about physical UGC brands, most customers sell digital products and mobile apps.4:17–7:39 · Guest teaching 5/10 Building Automated AI Outbound and Sourcing Agents Nathan inquires about outbound tooling, prompting Romain to explain how they built custom LLM agents to replace manual sourcing and outreach. Nathan presses on exact launch revenue figures.7:39–10:22 · Guest teaching 5/10 The Origin Story: Building the WeFast Mobile App Nathan asks how Romain entered the women's fasting app market with WeFast. Romain walks through market research, community dynamics, and paid acquisition mastery that led to Arkads.10:22–14:23 · Guest teaching 5/10 Bootstrapping and Maintaining High Profit Margins Nathan repeatedly attempts to pin Romain down on exact profit margins and ad spend. Romain resists giving concrete percentages while describing experimental multi-account organic TikTok automation.14:23–20:01 · Guest teaching 7/10 Live Demonstration: Building AI Growth Agents in Gumloop Romain shares his screen to demonstrate how he builds autonomous growth agents in Gumloop, educating Nathan and the audience on node architectures and few-shot tweet generation.20:02–23:42 · Guest teaching 6/10 Advanced Autonomous Agents: Competitor Scraping and Hooks Nathan demonstrates marketing expertise by proposing a modular copywriting framework (hook, body, CTA). Romain responds with his alternative perspective on how LLMs perform best on full-sample duplication.23:43–25:16 · Guest teaching 2/10 Final Growth Insights and Interview Conclusion Nathan summarizes the core metrics, business evolution, and agent infrastructure of Arkads in a rapid-fire outro recap while Romain shares final thoughts.0:52–4:16 · Guest disagreement 1/10 Scaling to $6 Million ARR with a Tiny Team Nathan digs into Romain's ARR metrics and go-to-market channels. Romain explains that contrary to Nathan's assumption about physical UGC brands, most customers sell digital products and mobile apps.4:17–7:39 · Guest disagreement 1/10 Building Automated AI Outbound and Sourcing Agents Nathan inquires about outbound tooling, prompting Romain to explain how they built custom LLM agents to replace manual sourcing and outreach. Nathan presses on exact launch revenue figures.7:39–10:22 · Guest disagreement 1/10 The Origin Story: Building the WeFast Mobile App Nathan asks how Romain entered the women's fasting app market with WeFast. Romain walks through market research, community dynamics, and paid acquisition mastery that led to Arkads.10:22–14:23 · Guest disagreement 3/10 Bootstrapping and Maintaining High Profit Margins Nathan repeatedly attempts to pin Romain down on exact profit margins and ad spend. Romain resists giving concrete percentages while describing experimental multi-account organic TikTok automation.14:23–20:01 · Guest disagreement 1/10 Live Demonstration: Building AI Growth Agents in Gumloop Romain shares his screen to demonstrate how he builds autonomous growth agents in Gumloop, educating Nathan and the audience on node architectures and few-shot tweet generation.20:02–23:42 · Guest disagreement 2/10 Advanced Autonomous Agents: Competitor Scraping and Hooks Nathan demonstrates marketing expertise by proposing a modular copywriting framework (hook, body, CTA). Romain responds with his alternative perspective on how LLMs perform best on full-sample duplication.23:43–25:16 · Guest disagreement 0/10 Final Growth Insights and Interview Conclusion Nathan summarizes the core metrics, business evolution, and agent infrastructure of Arkads in a rapid-fire outro recap while Romain shares final thoughts.0:52–4:16 · Nathan pushing back 2/10 Scaling to $6 Million ARR with a Tiny Team Nathan digs into Romain's ARR metrics and go-to-market channels. Romain explains that contrary to Nathan's assumption about physical UGC brands, most customers sell digital products and mobile apps.4:17–7:39 · Nathan pushing back 2/10 Building Automated AI Outbound and Sourcing Agents Nathan inquires about outbound tooling, prompting Romain to explain how they built custom LLM agents to replace manual sourcing and outreach. Nathan presses on exact launch revenue figures.7:39–10:22 · Nathan pushing back 2/10 The Origin Story: Building the WeFast Mobile App Nathan asks how Romain entered the women's fasting app market with WeFast. Romain walks through market research, community dynamics, and paid acquisition mastery that led to Arkads.10:22–14:23 · Nathan pushing back 6/10 Bootstrapping and Maintaining High Profit Margins Nathan repeatedly attempts to pin Romain down on exact profit margins and ad spend. Romain resists giving concrete percentages while describing experimental multi-account organic TikTok automation.14:23–20:01 · Nathan pushing back 1/10 Live Demonstration: Building AI Growth Agents in Gumloop Romain shares his screen to demonstrate how he builds autonomous growth agents in Gumloop, educating Nathan and the audience on node architectures and few-shot tweet generation.20:02–23:42 · Nathan pushing back 4/10 Advanced Autonomous Agents: Competitor Scraping and Hooks Nathan demonstrates marketing expertise by proposing a modular copywriting framework (hook, body, CTA). Romain responds with his alternative perspective on how LLMs perform best on full-sample duplication.23:43–25:16 · Nathan pushing back 0/10 Final Growth Insights and Interview Conclusion Nathan summarizes the core metrics, business evolution, and agent infrastructure of Arkads in a rapid-fire outro recap while Romain shares final thoughts.

speaking balance: gold is Nathan, purple is the guest (3 minute bins)

0:00 · Nathan 57.9% · guest 42.1%0:00 · Nathan 57.9% · guest 42.1%3:00 · Nathan 18.5% · guest 81.5%3:00 · Nathan 18.5% · guest 81.5%6:00 · Nathan 15.4% · guest 84.6%6:00 · Nathan 15.4% · guest 84.6%9:00 · Nathan 20% · guest 80%9:00 · Nathan 20% · guest 80%12:00 · Nathan 25.4% · guest 74.6%12:00 · Nathan 25.4% · guest 74.6%15:00 · Nathan 42.3% · guest 57.7%15:00 · Nathan 42.3% · guest 57.7%18:00 · Nathan 10% · guest 90%18:00 · Nathan 10% · guest 90%21:00 · Nathan 30.3% · guest 69.7%21:00 · Nathan 30.3% · guest 69.7%24:00 · Nathan 47.2% · guest 52.8%24:00 · Nathan 47.2% · guest 52.8%
Sharpest disagreement ▶ 10:59 Romain blocks profit margin probing

Romain playfully refuses Nathan's repeated attempts to lock him into an exact margin percentage, maintaining strict privacy on internal spend numbers.

Hardest push from Nathan ▶ 10:51 Nathan pushes for specific margin percentages

Nathan refuses to let Romain off with vague answers about being profitable, directly asking whether margins are 10 percent or 60 percent.

Biggest teaching moment ▶ 17:49 Romain demonstrates AI agent workflow construction

Romain conducts an in-depth technical walkthrough in Gumloop, showing how custom inputs and inspiration lists can automate complex copywriting tasks.

Nathan holds their own ▶ 21:31 Nathan challenges prompt architecture with copywriting theory

Nathan leverages direct marketing knowledge to challenge Romain on whether breaking prompts into hooks, curiosity, and call-to-action components yields superior agent outputs.

the scores for every segment, with the reasoning behind each
ChapterTopicNathan as informed peerGuest teachingGuest disagreementNathan pushing backWhy
Scaling to $6 Million ARR with a Tiny Team 5412 Nathan digs into Romain's ARR metrics and go-to-market channels. Romain explains that contrary to Nathan's assumption about physical UGC brands, most customers sell digital products and mobile apps.
Building Automated AI Outbound and Sourcing Agents 5512 Nathan inquires about outbound tooling, prompting Romain to explain how they built custom LLM agents to replace manual sourcing and outreach. Nathan presses on exact launch revenue figures.
The Origin Story: Building the WeFast Mobile App 5512 Nathan asks how Romain entered the women's fasting app market with WeFast. Romain walks through market research, community dynamics, and paid acquisition mastery that led to Arkads.
Bootstrapping and Maintaining High Profit Margins 6536 Nathan repeatedly attempts to pin Romain down on exact profit margins and ad spend. Romain resists giving concrete percentages while describing experimental multi-account organic TikTok automation.
Live Demonstration: Building AI Growth Agents in Gumloop 5711 Romain shares his screen to demonstrate how he builds autonomous growth agents in Gumloop, educating Nathan and the audience on node architectures and few-shot tweet generation.
Advanced Autonomous Agents: Competitor Scraping and Hooks 7624 Nathan demonstrates marketing expertise by proposing a modular copywriting framework (hook, body, CTA). Romain responds with his alternative perspective on how LLMs perform best on full-sample duplication.
Final Growth Insights and Interview Conclusion 6200 Nathan summarizes the core metrics, business evolution, and agent infrastructure of Arkads in a rapid-fire outro recap while Romain shares final thoughts.

Statements from this episode (13)

Assertion Not checkable as stated
Arcads reached $6M ARR, adding $1M in a single month
“So I think one month ago we were at five million AR. Now we are at six.”
Romain Torres Jul 8, 2025 ▶ 1:08
Assertion Not checkable as stated
Arcads runs its $6M ARR business with a five-person engineering team
“This, the team is still five people engineers, very talented engineers.”
Romain Torres Jul 8, 2025 ▶ 1:14
Disclosure
Torres: Arcads relies most on Meta ads made with its own tool
“The first one and the most important one is like actually paid ads. We use our own tool to promote our tool. So we do some paid ads on Meta to like basically show what the tool can do and We just prove it works by using it, right?”
Romain Torres Jul 8, 2025 ▶ 3:04
Disclosure
Arcads Built Internal AI Agents for Lead Sourcing and Outreach
“So we ended up like building a lot of internal AI agents to do exactly this.”
Romain Torres Jul 8, 2025 ▶ 5:30
Assertion Not checkable as stated
Torres: Arcads hit $5k MRR in its first week
“First week was like five K of monthly recurring revenue after like literally one week.”
Romain Torres Jul 8, 2025 ▶ 7:07
Assertion Not checkable as stated
Arcads hit $1M ARR six months after onboarding its first customer
“I think it was, yeah, the first time was in June, I think.”
Romain Torres Jul 8, 2025 ▶ 7:32
Assertion Not checkable as stated
Arcads is highly profitable, fully bootstrapped, and avoids VC funding entirely
“What I can tell you is like, archives is very profitable. We are like bootstraps. We are like, not like relying on VC money to fund the paid ads and the growth.”
Romain Torres Jul 8, 2025 ▶ 10:34
Insight
Torres: TikTok suppresses organic reach on accounts posting 10 times daily
“Actually, I very quickly realized that if you post 10 times a day on the same accounts, TikTok will not push your content too much because they don't really want people to be posting that much.”
Romain Torres Jul 8, 2025 ▶ 12:39
Assertion Not checkable as stated
Teens make six figures monthly on TikTok Shop using Arcads AI avatars
“Teens create TikTok shop accounts and they just like create a ton of content on TikTok using Arkad's avatars to promote those supplements brands and say sometimes like do really, really good revenue, like sometimes like six figures a month without any paid ads…”
Romain Torres Jul 8, 2025 ▶ 13:52
Assertion Not checkable as stated
Torres: Arcads has close to 4,000 paying subscribers
“No, it's a little bit more. It's like closer to 4000 now.”
Romain Torres Jul 8, 2025 ▶ 14:44
Insight
Torres: AI excels at duplicating high-performing content, not creating
“AI, what you have to understand is AI is very good at duplicating stuff. It's I think it's gonna be good at creating very soon, but right now, like, it's, like, really world class at taking something that is good and doing something similar.”
Romain Torres Jul 8, 2025 ▶ 17:50
Disclosure
Arcads uses AI agents to scrape, rewrite, and replicate competitor ads
“I have one that maybe I can show you, but it might be a bit too complex, but that basically will send me messages in Slack every single day when my competitors run new ads that are worth replicating. It will scrape the ads libraries and yeah, it will download …”
Romain Torres Jul 8, 2025 ▶ 20:24
Insight
Torres: ChatGPT excels when replicating examples rather than following rules
“My understanding of like where chat GPT really excels is not when you give them like tips on like, this is how you write a good hook. And this is how you write a good body, but it's more like do this, but in this different way. And that's why like, I like to g…”
Romain Torres Jul 8, 2025 ▶ 21:59
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 2,600 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.