Risky strategy How start-ups fake technical progress

Faking it is part of the business. Start-ups have to convince investors that their product already works or is about to. Sometimes this works well, but it can also land those responsible in prison.

An acquaintance once told me how he started an IT job for which he was not qualified at all. His first task was to get a program up and running that wasn't working properly. The only thing he knew was that in many programming languages there has to be a semicolon at the end of every statement, so he looked for a missing semicolon and actually found it. The program then ran.

In the weeks and months that followed, the friend acquired the knowledge he claimed to already have through self-study and on the job. In the process, he developed into a capable software developer who worked for this company for fifteen years in increasingly specialized tasks.

"Fake it til you make it" is the name of this principle, in other words: "Fake it until you make it." It seems to be particularly popular with inventors and start-ups. Investors are easier to convince if the product already works. This has always worked well in the past. But sometimes it doesn't. There is a fine line between realistic promises and fraud.

Fireflies.ai is an AI bot that creates session logs. In the age of Gemini and ChatGPT, the product no longer seems revolutionary. In 2017, when the two founders launched their online service, it certainly was. And as they recently announced, they didn't have a functioning AI application back then either. Instead, they dialed into the meetings themselves and recorded what they heard.

In this way, they gained enough customers and interest from investors to develop their AI. Today, Fireflies.ai is worth one billion dollars. A few days ago, one of the founders, Sam Udotong, revealed the origin of the product in a LinkedIn post. The opinions in the comment columns range from "absolutely great" to "obvious fraud" and "espionage", as a person was listening in on potentially confidential meetings instead of a bot.

Strictly speaking, the Fireflies founders did not defraud their customers, as they provided them with the service they paid for. However, they were lying when they claimed that it was already their AI that was doing the work. It is unclear whether they also told investors that their protocol bot was already working or whether they only had a prototype that was 100 percent controlled by humans.

In the early phase of the Industrial Revolution, machines that performed human activities were in vogue - especially as demonstration objects to amaze the public.

The "Schachtürke" was a chess machine built by the Hungarian civil servant Wolfgang von Kempelen. It had its premiere before the eyes of Queen Maria Theresa of Austria-Hungary. A wooden figure in oriental garb pushed the pieces across the board. The machine impressed not only because it obviously knew the rules, but also because it played quite well. The Shaft Turk went into operation in 1770 and fascinated people for more than 30 years. An experienced chess player actually sat inside and moved the wooden piece so that it moved the pieces to the correct square.

Now, von Kempelen never intended to actually build an autonomous chess-playing machine. In his case, it was actually cheating. For decades, he and, after his death, his successor toured Europe and even the USA with the Chess Turk. It was not until more than 60 years after the premiere that the actual functioning of the machine became public knowledge.

In English-speaking countries, the chess machine is known as the "Mechanical Turk". Under the same name, Amazon launched a platform in 2005 on which people can perform simple, repetitive tasks online for a small fee, such as data entry, transcription, but also photo and video editing.

Elon Musk's company Tesla not only builds more or less self-driving cars, but also humanoid robots. A whole guard of these human-like creatures served drinks, danced and played rock-paper-scissors with the guests at an event for investors in October 2024. Even there, doubts arose because the robots spoke with very different voices, for example.

A little later, Tesla also admitted that the Optimus humanoids had still been controlled by humans that evening. In the meantime, the Tesla robots have learned a lot. They have mastered various kung fu moves, household chores - including cooking - and common factory tasks, including quality control.

The fact that Tesla's party robots were still dependent on human help in 2024 only briefly caused ridicule. Development is progressing, but there is no launch date for the Optimus. Tesla has announced that it wants to use the robots in its own company first.

"Fake it til you make it" becomes a real problem when, despite all efforts, it is not possible to get out of the fake phase. This was the case with Theranos, a company that had launched a revolutionary device for blood analysis. This was designed to carry out 240 different analyses with a much smaller amount of blood than any other known device. Part of the success story of the company, which was valued at 10 billion dollars at its peak, was that it was founded in 2003 by the then 19-year-old Elizabeth Holmes. As CEO, she was considered the world's first self-made billionaire, and in 2015 the Times magazine listed her as one of the 100 most influential people in the world.

Unfortunately, "Edison", Theranos' blood analysis device, never even came close to delivering what the company promised. But it was already in use. In fact, the company used laboratory equipment from other manufacturers.

Because Holmes kept up the deception for so long - including inviting then Vice President Joe Biden to a similarly fictitious laboratory in 2015 - she was charged with fraud and sentenced to 11 years in prison in 2022, which she began in 2023. The Theranos company was dissolved in 2018.

Nick Swinmurn turned deception into a real multi-billion-dollar business with the online shoe store Zappos. What he discovered as a business idea in 1999 was considered highly risky at the time: would enough people actually order shoes online without trying them on in store?

Because he didn't have the capital to set up a huge shoe warehouse at the beginning, he made customers believe that they were ordering from him. In reality, he or one of his employees would go in search of a store that had the desired pair in the desired size and color in the display. In the early years, Swinmurn even took some of his product photos directly in shoe stores in his area. Today, dropshippers apply the same principle, albeit by ordering from other online stores for their customers.

Zappos had its breakthrough with the offer to take shoes back free of charge if they didn't fit, and not to charge postage for delivery or returns. This increased demand and interest from investors. Zappos had proven that many people were actually willing to buy shoes online. In 2002, Zappos entered into a partnership with UPS and built its own warehouse near the logistics company's global air freight hub. In 2009, in the midst of the financial crisis, Amazon bought Zappos for 1.2 billion dollars.

The pretense of technical progress made by vehicle manufacturer Nikola Motor Company turned out less well. A video published in 2018 purported to show a truck powered by a fuel cell at full speed. In fact, it had no functioning drive and instead rolled down a sloping road, which was not visible in the advertising clip.

This and other accusations came from a short seller who had bet on Nikola's share price falling. He published an entire dossier full of accusations about how Nikola deceived investors with empty promises. Three months earlier, the vehicle manufacturer had gone public and two days before the allegations of fraud came to light, General Motors had acquired a stake in Nikola.

The CEO subsequently resigned and was sentenced to four years in prison for fraud in 2022. In March 2025, President Trump pardoned the fallen manager. The company went bankrupt in summer 2025.

Software solutions are harder to understand than truck engines, especially when they are only demonstrated in an online video. In 2007, Drew Houston demonstrated in a video what his product called Dropbox could do, namely synchronize files between different computers and make them accessible in an online directory. Cloud storage was still new at the time and Houston did not know how popular his Dropbox would actually be when he had developed it to market maturity.

The video went viral and he reportedly had tens of thousands of requests from users wanting his Dropbox within a day. He realized that the demand was indeed there and found his business partner along the way. Although Houston also pretended to have a functioning product, he subsequently actually realized it.

In the start-up world, this is also known as a minimum viable product, i.e. the minimum - or most cost-effective - functioning product that demonstrates, in the sense of a prototype, what benefits the product will one day bring. The question is then how much of the promise the people behind the new technology are already keeping and how much of it they will only deliver in a few years' time - if the development is successful.

Dropbox has been. The company claims to have 700 users. In 2024, it generated a turnover of 2.5 billion dollars.