Loading page…
An AI cold-call and interview simulator that reps practise on every morning — $2,500+ MRR in under 40 days, built in two days after their first product got flattened by a Claude release
roughly break-even · a couple of hundred users · 20-30% of users log in daily · 30-40% of trials convert to paid · pricing moved from $49/mo flat to $50/mo (6 hours) and $99/mo (10-12 hours) · ~11 cents per minute of AI cost · top 10% of users were running ~$67 in the red
They spent five months building an AI calendar co-pilot, watched a single Claude release make it unsellable, then shipped a dumber product — a cold-call simulator — in two or three days and had paying users immediately.
Fewer bars = easier, cheaper, or faster for an AI-assisted solo builder. Editorial judgments based on the case details.
Eric and Chris ran a sales training and education business called Higher Levels for years before building software — a decade each in tech sales, roughly 40,000 YouTube subscribers, and an email list of about 15,000 people. They had entertained this product idea more than a year earlier and left it alone because the timing was wrong for their business. The first thing they actually built was not this. It was an integrated calendar and inbox co-pilot for top-performing reps, connected to calendar, email and call recordings, designed to eat administrative work. They spent five months on it, priced it at $79 a month, and had people who genuinely loved it. Then Claude's agentic co-working release landed. Eric adopted it for basically everything within days, and the arithmetic changed: "we had people that loved and used our product and were willing to pay at that time $79 a month for it, no problem... but with one update, even though it's not there yet, within 24 months, there's no way we're going to outlast." On top of that, every enterprise-account customer needed corporate approval to connect a Gmail account, so onboarding was a fight even when the buyer was enthusiastic. So they looked at where the signal actually was, and it was somewhere else entirely — a cold-call simulator they had built almost incidentally. They shipped it in two or three days and pivoted the company inside a week. The result at the time of the recording: over $2,500 in monthly recurring revenue in under 40 days, roughly break-even, a couple of hundred users, and three live conversations with companies doing more than a billion dollars in annual revenue.
Every paying user so far came from content made in 2026, and the founders front-run the obvious objection: the 40,000 YouTube subscribers and 15,000-person email list exist, but "we have not sent one email to our email list," and they did not post on LinkedIn at all until 2026. No ads, no cold outreach.
The quick-reference card — acquisition channels, replication playbook, risks & traps, and our verdict — is members-only.
Join the Unicorn Club — $5/moCancel anytime, in one clickPaste into Claude Code / Codex and get a working version of this product end-to-end.
Two production-grade prompts per idea — a full build spec and an SEO growth plan — plus 10 tailored build specs a week. Members only.
Join the Unicorn Club — $5/moCancel anytime, in one clickEvery claim traces back to a listed source — open any of them and check the number yourself.
Data credibility: Founder-reported by both cofounders on their own YouTube channel, with no dashboard, Stripe screenshot or third-party data shown at any point. Self-reported figures: over $2,500 MRR in under 40 days; roughly break-even; a couple of hundred users; 20-30% daily login rate; 30-40% trial-to-paid conversion; original pricing $49/mo flat, previous product $79/mo, current tiers about $50/mo for six hours and $99/mo for ten to twelve; approximately 11 cents per minute of AI cost; the top 10% of users running about $67 in the red on the old unlimited plan; SOC 2/ISO/GDPR estimated at $70,000-$100,000 all in; a 14-page security audit that was rejected after three weeks because the product was in beta; three live conversations with companies doing over $1B in annual revenue; and named individual users from Replit, HubSpot, Keller Williams, AlphaSense and Expensify. Their existing business, Higher Levels, is credited with roughly 40,000 YouTube subscribers and a 15,000-person email list, and they state repeatedly that they have never emailed that list. They are unusually explicit that they do not have answers to their own open questions, which is worth weighting positively, but the company is under two months old and nothing is verified. Region is the United States on Eric's own statement, 'I live in Austin, Texas,' where he attends a monthly accelerator meetup downtown; the CTO's location is not stated. The exact spelling of the product is given verbally as 'rhythm.ai' and spelled out on air as 'RIT thim.ai', so treat the domain as approximate.
The pivot product shipped in two or three days, went in front of people, and got paid users immediately at a flat $49 a month with unlimited usage — a price picked to acquire rather than to optimise, and one that did produce renewals.
The number that told them this one had traction was daily login: 20-30% of users come back every day, with no activation sequence and no onboarding tooltips in the product, because a rep rips one simulated call each morning before dialling real people.
The dead stretch was the five months before this product existed.
Rhythm.ai simulates live conversations a salesperson has to survive: cold calls, discovery calls and job interviews. A rep loads a scenario and gets a realistic voice counterpart on the other end — objections, brush-offs, hostility included. Eric's own description of the best-performing marketing asset is telling: "a YouTube video that I cut up and posted on LinkedIn of the world's hardest cold call simulator. So, I got absolutely destroyed by our own tool." Chris is candid that the winning product is technically the lesser one: "it's actually, in my opinion, kind of a dumber product. Like no disrespect to our CTO or anything like that, but it's more straightforward." And it requires zero integrations, which is exactly why it onboards in a minute where the calendar co-pilot took three weeks of security review. The layer they charge for on top of the raw simulation is where the domain expertise lives: custom scorecards, so a rep or a team can grade calls against their own sales methodology, plus the scenario programming that makes the counterpart behave like a real prospect. The feedback they keep getting is "this is the most realistic call simulator I've ever had and it's not even close." For teams, they added a second surface in one week once the B2B conversations got serious: team analytics, visibility, training paths and assignment tracking. Our read: that split is the important structure. The consumer buys realism; the manager buys observability. The realism is where the AI cost sits; the analytics are near-free to serve.
Subscriptions, and the pricing is the most instructive part of the whole breakdown because they got it wrong first and said so. They launched at $49 a month with unlimited usage, which acquired customers and did produce renewals. Then they measured consumption and found their power users were unprofitable: the top 10% were roughly $67 in the red each, against an underlying cost of about 11 cents per minute of simulated conversation. "AI is pure consumption now — it is a real challenge." So they moved to tiers deliberately modelled on the tools their buyers already use: about $50 a month for six hours of usage, about $99 a month for ten to twelve. The reasoning is behavioural rather than financial: "we're very early in testing of that two-tier model if nothing else to make it exactly similar to Claude because I hit the same exact thing with Claude and ChatGPT where it's like I'm using it so much... 'Hey, you've used out all your usage. You need to upgrade to $100 a month.' I'm like, dang, I don't care." Matching the mental model of ChatGPT and Claude pricing means the upgrade prompt feels normal rather than like a bait-and-switch. The conversion numbers they report are strong: 30-40% of people who start a trial become paying users, and 20-30% of users log in daily. They are roughly break-even at $2,500 MRR and are deliberately not accelerating until they know their margin. Chris's framing: "we might get to like 15,000 MRR, but what if we're losing $7,000 a month?" The consumer tier is also explicitly a lead magnet. Individual users at Replit, HubSpot, Keller Williams, AlphaSense and Expensify signed up out of pocket, which is the opener for the team deal: "even if we break even on those people and that allows us to then break into B2B motions and say, hey, three of your reps are already using it."
They pre-empt the obvious objection: yes, they have an audience, and no, that is not where the customers came from. "We did not post on LinkedIn until 2026... I don't want anyone to see this and think, wait a minute, I'm 2 minutes into this video, oh, they have 40,000 followers." They have never emailed their 15,000-person list. No ads. No cold outreach. Every paying customer so far came from content made in 2026 that had to perform on its own. The format is specific and worth copying exactly: not build-in-public, but free value on the topic the product serves, with the product used as the demonstration. "I'll make some of the content that we know works well and just give free value about interview advice, about cold call advice... And oh, by the way, if you want to reinforce this concept in the video, I'm going to use Rhythm AI to demonstrate the theory that I just talked about in practice." Eric is blunt about the alternative: "the videos that are getting customers are not like Chris and I are building in public, follow along as we build our AI SaaS. No... Everyone's going to watch one video and then realize you don't know what you're doing." The single best lead source was a YouTube clip recut for LinkedIn — the world's hardest cold-call simulator, with the founder losing to his own product. On an account he says he has never grown, just a decade of tech-sales connections, it got 200+ likes, well over 100 connection requests, and a wave of signups. The same LinkedIn posting also opened the enterprise pipeline with no outbound at all: three conversations with billion-dollar-revenue companies, sourced purely from posting and demoing publicly. And the early-user handling is deliberate: earliest users got a Slack channel and the founders' cell phone numbers. "we nurtured the [expletive] out of those people."
They build on Claude, and their unit economics are therefore consumption economics. About 11 cents per minute of simulated conversation is their working figure. On the old $49 unlimited plan the worst decile of users cost them roughly $67 more than they paid. Chris frames the shift precisely: "in the past you kind of more or less have fixed costs — you build it, it runs automatically... Now anytime someone uses it... you've got to count that in and have really great insights on your users." Their conclusion is that margin discovery has to happen before scaling, not after: "figure out your margins, figure out what you're actually scaling before you start amplifying it, throwing a bunch of dollars behind it." The enterprise side changes the shape favourably. Consumer usage is the variable cost; team analytics and visibility features are "kind of static features you build once and they cost pennies. But there's immense value to the organization." So the strategy is to sell the cheap-to-serve layer to the buyer with the budget. The biggest hidden cost is not compute. It is enterprise compliance. They have run the numbers on SOC 2, ISO and GDPR and put it at roughly $70,000-$100,000 all in, plus significant time to stand up the controls — and Chris used to sell SOC 2 for a living, so he knows. His warning is direct: "don't start building a tool for B2B enterprise if you don't have money to spend on just getting the preliminary things in place to even be evaluated." Build speed is the other cost story, and it is now trivially low. The cold-call simulator took two days. The full suite of team features and analytics took one week. Three people, no employees — they say they could not hire anyone today without going into debt.
Copy the moat definition, because it is the only honest one available for a small AI product in 2026: your moat is the domain expertise encoded around the model, not the model. Anyone can ask Claude for a cold-call simulator. Very few people can specify what a realistic gatekeeper objection sounds like, or what belongs on a scorecard for a specific sales methodology. Their read on the market share is worth internalising: roughly 20% of ChatGPT and Claude users will bother to build the thing themselves, which leaves 80% who will never DIY it. Copy the content format. Free, genuinely useful advice on the problem, with the product appearing as the demonstration of the point you just made. Not build-in-public. And copy the specific asset type that worked: put your own product in the hardest possible mode and lose to it on camera. Copy the pricing mimicry. Mirror the tier structure of ChatGPT and Claude so usage limits feel familiar rather than punitive, and pick the tier boundary from measured consumption rather than from a spreadsheet guess. Copy the consumer-first sequence when your product can support it. Individuals harden the product, and their employers' logos are the opener for the team deal. Their estimate for going B2B cold is sobering: minimum six months before an enterprise trusts a new vendor, and the first one will be small and probably a friend. What to change: they had a decade of sales credibility and 40,000 subscribers to draw on even if they didn't monetise the list. If you don't, budget more time for the content engine to bite. And do not repeat their first product. A tool whose value overlaps directly with what a frontier assistant will ship next quarter is a tool with a 24-month clock on it — which is exactly why they are staying bootstrapped rather than raising against that clock.
The founders name the biggest risk themselves, and they have already been hit by it once. A single Claude release made their first product unsellable after five months of work. The same thing can happen to this one. Eric's framing of the enterprise objection is the sharpest line in the transcript: "if you don't have a good answer to like why are you going to be better than Claude in five years, that company's looking at — hey, we already committed to spend a hundred million with Claude, this looks cool, but my approach is going to be our team needs to try to build that internally first." Second: consumption margin. At $2,500 MRR they are roughly break-even. Their own worry is signing a big contract that looks great and loses money — a hundred reps at 11 cents a minute is a very different cost curve from a couple of hundred casual users, and they have not yet proven the tiers hold at team scale. Third: the consumer segment self-destructs on success. Interview practice is the clearest case — "as soon as they get that job, they're done." And they estimate a natural ceiling of only 10-20% of salespeople who care enough about their job to put a personal card down for tools. Fourth: the enterprise gate. They already lost a trial after filling in a 14-page security audit, three weeks later rejected because the product was in beta. SOC 2 at $70-100K is not optional for the deals they want, and it is a large bill against $2,500 MRR. Fifth: competition. Their own space already contains funded players — "they've raised 50 to $100 million" — and the founders are candid that they don't see how some of those survive either. On evidence: this is the founders talking about their own company on their own channel, with no dashboard shown. They are unusually forthright about what they don't know and explicitly say so repeatedly, which is a point in their favour, but every number here is self-reported and the business is under two months old.
The four founder playbooks — first customers, first dollar, the flywheel, the silent stretch — are in the Unicorn Club, on this idea and on all 1,403 others. The story, the numbers, the sources and every later section stay open to everyone.
Join the Unicorn Club — $5/moCancel anytime, in one click