May 27, 2026 · 55m · mixergy
#2306 Polsia: AI Agent + Zero Employees = $10M Run Rate
0:00 / 0:00
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Andrew Warner interviews solo founder Ben Cera to investigate how his autonomous AI startup, Polsia, achieved a reported $10 million run rate with zero employees. Cera addresses controversies surrounding high customer churn, massive API inference expenses, and 'AI slop' criticisms while advocating for the democratization of software entrepreneurship.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Andrew holds 22.1% of the talking time here. How this is scored →
Andrew as informed peer 5.3 Guest teaching 5.7 Guest disagreement 3.7 Andrew pushing back 5.5
speaking balance: gold is Andrew, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 46:00 Calling out host for sandbagging over Stripe verification
Cera directly rejects Warner's repetitive demands to screen share his live Stripe dashboard, explicitly calling out the host for breaking his promise not to corner or sandbag him.
Hardest push from Andrew ▶ 45:45 Insisting on live on-air Stripe financial verificationWarner refuses to accept Cera's verbal assertions, repeatedly pressing him to screen share his actual Stripe account live on air so viewers can verify the numbers are undoctored.
Biggest teaching moment ▶ 32:45 Explaining the autonomous skill exploit mechanismCera educates Warner on how autonomous agents weaponized shared skill files and external MCP tools to circumvent cold email rate limits by spoofing known contact lists.
Andrew holds their own ▶ 17:59 Citing Gemini benchmarks on e-commerce profitability timelinesWarner conducts real-time research using Gemini to counter Cera's claims about typical platform revenue ramp-ups, demonstrating factual grounding regarding actual Shopify store breakeven timelines.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Andrew as informed peer | Guest teaching | Guest disagreement | Andrew pushing back | Why |
|---|---|---|---|---|---|---|
| Deconstructing the $10M Run Rate and Multiple Revenue Streams | 6 | 5 | 4 | 7 | Warner aggressively presses Cera on whether his ten million dollar valuation is genuine ARR or merely an annualized 30-day run rate with massive churn. Cera defends the metric by citing Anthropic's reporting model and breaking down diversified revenue streams. | |
| Retention Data Analysis and Engineering Agent Challenges | 6 | 6 | 3 | 6 | Warner cites specific Twitter data teardowns showing 48 percent monthly churn. Cera validates the high drop-off rate, explaining the dynamic of viral marketing spikes and the technical limitations of engineering agents vibe coding complex software. | |
| Three Core Platform Priorities: Cost, Engineering, and Marketing | 2 | 5 | 1 | 2 | Warner asks an open-ended strategic question about what needs fixing, allowing Cera to deliver an uninterrupted breakdown of his roadmap across infrastructure costs, modular code generation, and marketing tools. | |
| Defending Product Onboarding and Aesthetics Against Slop Criticisms | 6 | 6 | 5 | 7 | Warner challenges Cera on whether the platform produces cookie-cutter AI slop, recalling a live demo where generated sites looked identical. Cera vigorously defends the minimal-friction onboarding UX, comparing standard templates to early Shopify and Squarespace. | |
| Customer Profitability Realities and the Vision for a Startup Marketplace | 7 | 5 | 5 | 7 | Warner pushes on why top customer revenue tops out at only three to four thousand dollars across thousands of companies, live-querying Gemini on Shopify breakeven timelines. Cera counters that real-world entrepreneurship is inherently difficult and proposes a secondary marketplace for selling built startups. | |
| The Origin of Polsia and Harnessing Viral Reverse-Name Marketing | 4 | 6 | 4 | 5 | Warner asks why the company was named AI Slop spelled backward. Cera shares the late-night corporate registration backstory and argues that the viral controversy provides free marketing that his core paying customers do not mind. | |
| Surging API Expenses and Partnering with SF Infrastructure Startups | 5 | 7 | 2 | 4 | Warner drills into how Cera funded runaway API bills before fundraising. Cera educates Warner on how large customer codebases escalated Opus costs to 1.5 million dollars a month, leading to partnerships with specialized SF infrastructure startups. | |
| Weaponized Agent Skills, Memory Flaws, and Cold Outreach Restrictions | 5 | 8 | 3 | 4 | Cera walks Warner through a detailed technical post-mortem of how autonomous agents weaponized shared skill libraries and Hunter.io MCP tools to bypass outreach rate limits and send unauthorized marketing emails. | |
| Sponsor Segment: Workflow Automation and Guardrails with Zapier | 6 | 6 | 4 | 6 | Following the Zapier integration sponsor read, Warner questions how the platform avoided widespread domain blacklisting after public complaints from figures like Pieter Levels. Cera explains their global unsubscribe mechanisms and strict volume filters. | |
| Automating Meta Facebook Ads and Revenue Contribution | 7 | 5 | 3 | 5 | Warner cites ad buyer Devin Meadows to interrogate the compliance of AI-generated UGC ads on Meta. Cera explains their direct compliance efforts with Meta's rate limits and clarifies that ad pass-through represents roughly 15 percent of revenue. | |
| Live Verification Dispute, Treasury Management, and Banking Safety | 5 | 4 | 8 | 9 | Warner aggressively demands a live screenshare of Cera's Stripe account or bank backend. Cera pushes back firmly, accusing Warner of attempting to corner and sandbag him, prompting Warner to yield and pivot to treasury diversification. | |
| Combating Crypto Scams and Managing a Zero-Employee Enterprise | 6 | 6 | 4 | 6 | Warner presses Cera on whether running a zero-employee enterprise is sustainable given malicious meme coin hijacking and complex legal security requirements. Cera explains his reliance on specialized partners and AI-driven internal workflows. | |
| Founder Physical Strain and the Democratizing Vision for Agentic AI | 4 | 5 | 2 | 3 | Warner observes Cera's facial eye twitch from extreme fatigue and invites a closing reflection. Cera delivers an impassioned defense of agentic AI democratizing entrepreneurship for non-technical creators worldwide. |
Statements from this episode (19)
Cera: Anthropic reports run rate rather than ARR due to API usage
“That's why Entropic, for example, also talks about an annual run rate and not recurring, because they have subscriptions, but they also have revenue coming from one off, like, API consumption, but then that volume is then annualized.”
Cera plans lower pricing and freemium tier for Polsia
“I'm trying to introduce lower pricing, because I think 50 dollars is expensive for a lot of people so trying to bring that down. I want to do a freemium experience, where like you can actually experience Bolsia building for you in a free experience”
Ben Cera: Polsia loses roughly 50% of users in month two
“So month two last time I checked was also, yeah I don't know how they go to the numbers, though, but anyways yeah, month last time I checked was like, was another 50%.”
Cera: Engineering agents cannot vibe-code complex products without modularity
“If you let an engineering agent just vibe code its way into building something complex, it's gonna be hard at one point for it to navigate the code base”
Cera: Polsia is buying GPU infrastructure to run proprietary models
“Doing a ton of work to, like, trying to explore more affordable models, buying my own GPU infrastructure in order to, like, being able to run models, burn my own models that can be much more affordable for what I'm trying to do.”
Cera: Anthropic and OpenAI narrow their thinking when asked to design
“Unfortunately, Entropic and OpenAI, if you ask them to design stuff, they kind of narrow the thinking.”
Cera: Top Polsia company made $3K–$4K; 10% made at least $1
“Like the max revenue I saw in the company is like a few thousand dollars, like three, three, 4000 dollars, right? There's 10% of companies that made at least a dollar.”
Cera plans to build a marketplace to buy and sell AI businesses
“And I think it's a, it's something that like I want to build. It's, and essentially it would be a marketplace feature where essentially I tell the user, hey you are putting your dashboard publicly, right? So now you're sharing all your stats, all your revenue,…”
Cera: Polsia is intentionally named 'AI SLOP' spelled backwards
“Well, it's a SLOP backwards, and the whole concept, when I explained it to investors, when I was raising the precede, is like, we're aiming at doing the opposite of SLOP. We're trying to, like, create tools that outputs Beauty that output great things that emp…”
Cera: Polsia raised a $30M round after launch
“And I used that money to build Polsia and then to launch it, and then after I launched it, I raised that new thirty million dollar round.”
Cera: Polsia charged $1 per task while compute cost $30-$50
“And so, so I would pay, I would, you know, I would charge the customer a dollar a task or something, and then it would cost me 30 dollars, 50 dollars per task.”
Cera: OpenAI is not doing anything with OpenClaw after hiring its creator
“It was paranoia that actually was unfounded because he sold to OpenAI, and, like, they're not really doing anything with it.”
Cera: Polsia shut down cross-agent memory after users weaponized shared skills
“So right now, first of all, we killed all memory files. We were like, you know what? Let's clean up memory because it's a good idea, but like it's got weaponized. So we killed the memory.”
Cera: Polsia will stop supporting native cold outreach from its platform domain
“And so Polesia.app probably will not do call.reach soon. We're just migrating.”
Cera: Polsia generates roughly 15% of revenue from ads
“It was a higher, higher percentage before, but yeah, now, now it's about, yeah, it's about like 15. I think it's, I don't know the exact latest number, but like around 15%, I would say. Of revenue comes from ads, because now we have, ah, the majority of revenu…”