Superwhisper founder Neil Chudleigh, second from left, poses with his team in their Toronto office. The AI-powered app converts spoken words into text in real time, helping users draft messages, e-mails and documents.EDUARDO LIMA/The Globe and Mail
In the great span of human history, the ubiquity of typing is a relatively recent development. The first practical typewriter was patented in 1867, while the QWERTY keyboard layout didn’t emerge for another decade or so. But we’ve had the capacity for spoken language for 100,000 years, if not more.
That makes Neil Chudleigh more of a traditionalist, even if his method of working is, well, untraditional. As often as he can, he clips a small microphone to his shirt pocket or the rim of his hoodie and murmurs to his devices instead of typing. “This is a bit of a new thing,” he said of the wireless mic tucked into his pocket. “A lot of people use a podcast mic.”
Dictation is having a moment, and it has everything to do with artificial intelligence. LinkedIn co-founder Reid Hoffman, for one, wrote last year that he has been “voicepilled,” which he defined as the moment you realize you can “amplify your ability” by talking to your computer.
This is particularly true among software engineers. Many already use tools such as Anthropic’s Claude Code to develop applications based on plain-language instructions. But pressing keys individually is an impediment to productivity, which is why some developers are adopting AI-powered speech-to-text apps to verbally issue instructions to Claude. It’s either the platonic ideal of translating thought into action, or a sign of our productivity-obsessed working world reshaping our behaviour.
After a launching Superwhisper, Mr. Chudleigh, 32, quit a full-time job at a company he co-founded to focus on the app.EDUARDO LIMA/The Globe and Mail
Mr. Chudleigh is both a participant and driving force of the trend. In July, 2023, he made a dictation app called Superwhisper that uses AI models to transcribe spoken words and process the copy into different styles, such as a formal e-mail or a casual text. Within a few months, he was earning more money through the app than he was at his full-time job – at a company he co-founded, no less – and he quit.
He runs his new company, which is registered as SuperUltra, from Toronto. The app has hundreds of thousands of weekly active users from companies such as Meta Platforms Inc. META-Q, OpenAI Global LLC, Coinbase Global Inc. COIN-Q and Dropbox Inc. DBX-Q, while non-techies use it to take notes and dictate e-mails, texts and Slack messages. Employees at U.S. financial institutions are among its customers, too. Revenue is modest, in the seven-figures, but growing quickly. It offers a lifetime subscription option for about US$250, in addition to monthly plans.
Business Brief Newsletter
A daily look at the most important business stories that are making news and moving markets, written by Chris Wilson-Smith
Already have account?
Mr. Chudleigh, 32, only hired his first employee in August, 2025, and has taken no outside funding. The company now employs six people and five contractors and in April moved out of a bland office on the outskirts of Toronto and into a trendy neighbourhood in the west end. In an office of Superwhisper users, the team has developed a sixth sense for whether someone is making conversation or talking to their computers.
But Superwhisper faces a lot of competition in its quixotic goal of supplanting keyboards. Its biggest rival is likely Wispr Flow. Based in the AI mecca of San Francisco, Wispr boasts some 80 employees and US$81-million in financing. There are many other apps, too, sometimes with variations of “whisper” in the name. And the risk of a tech-savvy customer base is that they can code their own version, possibly even while using Superwhisper.
Mr. Chudleigh demonstrates how to use his Superwhisper voice-to-text app. One user who types at 60 words a minute says he can dictate at more than twice the speed using the app.EDUARDO LIMA/The Globe and Mail
Mr. Chudleigh has the advantage of being early to market. He launched his app about a year before Wispr, and it spread via word-of-mouth. Some of it can be traced to Andrej Karpathy, a founding employee of OpenAI and a former University of Toronto student. Mr. Karpathy is a bit of an AI guru these days and is credited with coining the term “vibe-coding” in a post on X last year to refer to building software with AI, as opposed to writing every line by hand. In the same missive, he cited Superwhisper as part of his workflow. “I barely even touch the keyboard,” he wrote.
David Brustein, a systems administrator in New York, purchased a lifetime subscription to Superwhisper and uses it for dictating e-mails, texting his mom and as part of his coding setup. The point is all about saving time. He can type some 60 words a minute but dictates at 162, according to stats provided by Superwhisper. He was hard-pressed to say what exactly he’s doing with these time-savings, though.
There are pitfalls with dictation. Whereas typing forces us to focus our thoughts, speaking aloud lends itself to stream-of-consciousness ramblings that would be unfair to foist upon our colleagues in a message. “Sometimes I will get a little verbose,” Mr. Brustein acknowledged. Because Superwhisper allows for customization, he’s gotten around this problem by introducing a step for an AI model to summarize his messages before firing them off, while keeping his tone of voice.
As AI upends shopping, retailers tailor their pitches for chatbots
AI superusers are pulling ahead and widening the productivity gap
Others find value in verbosity. Bob Bradley, vice-president of data science and AI engineering at Geotab Inc., uses a rival app called OpenWhispr, including for building web applications. “When an idea goes from my brain to my fingers to my keyboard, I’m filtering that information,” he said. “I find stream-of-consciousness much more informative and results in a much better output.” Nothing gets lost travelling from brain to computer, in other words.
Voice-to-text technology is not new and can be traced to the 1950s when Bell Labs built a machine to recognize the numbers zero to 9 when spoken by its inventor. Capabilities have progressed in fits and starts since then, notably with digital assistants such as Siri and Alexa.
Co-workers of Neil Chudleigh, Austen Shutherland, top, Nico DiPlacido, left, and Brendan James.EDUARDO LIMA/The Globe and Mail
Mr. Chudleigh has long been interested in voice-to-text and alternative ways of working with computers, such as eye-tracking software. He has a custom-built retainer that allows him to control his computer with his tongue and by tilting his head, though it’s an accessibility device that isn’t practical for him to use. After studying mechanical engineering and computer science at Western University, he co-founded an affiliate marketing company in 2015 and relocated to San Francisco to go through the storied Y Combinator accelerator program before returning to Toronto. (The company, PartnerStack, was acquired by AppDirect in April.)
While always a voice-to-text adherent, the technology never worked well enough for his liking. Advances with AI models have changed that. OpenAI released a speech-recognition model in 2022, while Google has its own offerings. Toronto’s Cohere Inc., meanwhile, released an open-source model in March that boasts the lowest English-language error rate, according to one ranking.
Mr. Chudleigh built Superwhisper to take advantage of these improvements, and the app allows users to choose from a variety of AI models capable of speech recognition. “It was a great time to solve my own need,” he said. While he’s determined to keep the operation lean and avoid outstanding financing, he’s made a few hires. Nico DiPlacido, an early employee at PartnerStack, jumped ship in March to work on Superwhisper, which he was already using. “I actually had to stop using it and start typing again so that I could maintain my skills,” he said.
Some Superwhisper users, including the team's Nico DiPlacido, right, continue to type to maintain that skill or complement voice-to-text in their work.EDUARDO LIMA/The Globe and Mail
Even fans of the app still have need of a keyboard. Robert Oberg, a content creator in Mexico, will speak aloud to Superwhisper for 10 or 20 minutes about ideas for a newsletter post, and then write the traditional way. “I have never been very interested in replacing typing, because it is something physical that exercises my brain,” he said. Still, he has noticed some dictation-creep, and he’s started to speak to Superwhisper to respond to e-mails and messages.
Russell Dunphy, head of engineering at a company in Britain, has tried both Superwhisper and Wispr Flow, but recently prompted Claude Code to make his own version. It took about 45 minutes. “I’ve made an objectively worse version. But it’s okay in the ways that are what I need,” he said. His employer is evaluating whether to switch to his app and save about £400 (roughly $740) a month.
At Superwhisper, Mr. Chudleigh didn’t seem worried about the myriad dictation apps, nor that vibe-coded versions were going to erode his customer base. “A lot of people frame startups as zero-sum games. I don’t think that’s the case,” he said. “If you’re talking about replacing every keyboard on the planet, it’s just such a huge market.”