Most people need months to find out if an idea has legs. Jure Sotosek found out his did in a weekend, and then spent two years turning that weekend into a business doing more than $1M a month.
Sotosek is a Slovenia-based software engineer who used to build real-time audio tools for Microsoft Teams. That background mattered more than it might seem, because the product he eventually built, ParakeetAI, lives or dies on how fast it can listen, think, and respond.
ParakeetAI is a real-time AI call assistant. It listens in on a conversation, a job interview, a sales call, a recruiting screen, and quietly feeds the user contextual answers while the conversation is still happening. Today it has more than 1.5 million users and generates roughly a billion organic views a month. It started as a weekend build with no funding behind it.
A Weekend, Not a Roadmap
The first version of ParakeetAI came together over a single weekend, built on Next.js, Speechmatics for transcription, and OpenAI's models underneath. There was no extended validation phase and no slow crawl toward a minimum viable product.
Sotosek had already tried and failed at more than ten products before this one. That history did not slow him down. It sharpened his instinct for spotting a weak signal early, so he could move fast on this idea instead of over-building it.
He Charged Money Before He Validated Anything Else
Most early-stage advice says to validate interest first and worry about pricing later. Sotosek did the opposite. He put a price on ParakeetAI from day one.
Someone entering a credit card is a much stronger signal than positive feedback.
A sign-up tells you someone is curious. A charge tells you someone has a problem worth paying to solve. He treated that difference as the whole point of an early launch.
The Real Engine Was Distribution, Not the Product
Sotosek had already run a hardware product, WrumerSound, that generated 200 million views across TikTok, Instagram Reels, YouTube Shorts, and Facebook Reels. He brought that same organic content machine to ParakeetAI.
Scale through creators, eventually working with more than 100 creators producing hundreds of videos a day
Treat content like performance marketing, systematically testing hooks, formats, scripts, and angles rather than relying on a handful of viral hits
Push volume relentlessly, generating one to two billion organic views across three-month stretches
Test secondary channels, Reddit, Google Ads, and SEO all showed promise but never matched organic short-form video
Solving the Latency Problem
A real-time call assistant has almost no room for lag. ParakeetAI has to run speech-to-text, pass that through an LLM, and return an answer, all while the conversation is still moving. Getting that pipeline fast enough to feel invisible was one of the harder engineering problems Sotosek had to solve.
Outgrowing the Interview Use Case
ParakeetAI launched as an interview assistant, but that use case has a built-in ceiling: once someone lands the job, they stop paying. Sotosek expanded the product into recruiting conversations, sales calls, and other continuous, recurring B2B situations where the need for a real-time assistant does not disappear after one good outcome.
What He Would Tell Someone Starting Today
Optimize for speed of learning, not for how impressive the product looks. A rough version that teaches you something fast beats a polished one that teaches you nothing.
Think about distribution before writing too much code. Have at least one plausible answer for where your first thousand customers come from.
Be persistent about becoming an entrepreneur, but flexible about the specific product. The ten failed ideas before ParakeetAI were not wasted time, they were the filter.
What Comes Next
Sotosek says he would rather run a small, highly efficient team of exceptional people than grow headcount simply because the company is growing. With AI still early in what it can enable, he is treating ParakeetAI less like a finished product and more like the first use case of a much longer run.
You can follow Jure Sotosek's work directly at parakeet-ai.com, on X, or on LinkedIn.
If this was useful, pass it on to someone sitting on an idea they are afraid to ship this weekend. Sometimes the fastest way to find out if something works is to charge for it and see who pays.