AI & SocietyOctober 5, 2026

The iPhone Killed the Calculator. Now the LLM Is Coming for Your Software Company.

When the iPhone arrived, it did not have to be a better calculator, camera or map to stop people from buying those things. Once they were inside the phone, it only had to be free and already in your pocket. The LLM is about to do the same to software. It will take out software companies in industries that never thought of it as a competitor, and it will not have to be better than their products either. It only has to be inside the tools their customers already use every day.

Sahra-Josephine Hjorth in a pink dress holding an early iPhone on Paseo de la Reforma in Mexico City, the Angel of Independence behind her

I bought the first iPhone in 2007, when I was in college and lived in Washington, D.C. It was heavy and slippery, and while it felt revolutionary, I also kept my BlackBerry. There are few moments in our history when something is released and you know it is going to be monumental, but you do not yet understand how. The iPhone was one of them. ChatGPT, released in November 2022, was another.

The people who understood the iPhone least were the companies it was about to take out. The company that made your car's navigation system did not think a phone was its competitor, and neither did the companies making cameras, calculators or alarm clocks. Yet the iPhone's home screen held a calculator, a clock with an alarm, maps and a camera, and every one of those icons had been somebody's entire business. Most software companies are in the same position today, and they do not see it either.

The first iPhone was worse at almost everything

That first phone cost $499 or $599 with a two-year AT&T contract. It could not install apps, because there was no App Store. It had no GPS, so the maps could only guess roughly where you were. The camera could not record video, and the phone could not even copy and paste. Microsoft's CEO, Steve Ballmer, laughed at it, calling it the most expensive phone in the world and saying business customers would not want a phone without a keyboard. I understood the doubt, and it is part of why I held on to my BlackBerry. I was skeptical of typing on a screen, and the phone was so slippery that it made me feel clumsy.

Then it started to improve, and much of the improvement arrived without anyone buying a new phone. In July 2008, Apple released a free software update that added the App Store to the phone people already owned. Suddenly anyone could build software for it, and the phone in your pocket could do things it could not do the week before. A year later, another free update added copy and paste. What software could not fix arrived in new models instead: the iPhone 3G added GPS in 2008, and the iPhone 3GS could record video in 2009. Every year the phone did more, and much of it came as a download rather than a purchase.

The map company never saw a phone as a competitor

In October 2009, Google announced free turn-by-turn navigation for phones, and that day TomTom's stock fell 21 percent and Garmin's more than 16. A satnav, the GPS box people stuck to their dashboard for directions, cost a few hundred dollars. Overnight, the same job became an app that cost nothing.

Two years later, TomTom's investor relations director, whose job was to reassure the company's shareholders, told a reporter that satnavs were "definitely not doomed," and offered an example to prove it: "Just because you have a camera on your phone doesn't stop you from buying a digital camera." He picked the worst possible example. Compact camera shipments had peaked the year before at more than 108 million. In 2025, the industry shipped 2.4 million.

Figure 1

The camera the phone made unnecessary

How to read it. Both bars are drawn on the same scale. The orange bar is the more than 108 million compact cameras shipped in 2010, the peak year. The pink sliver is the 2.4 million shipped in 2025, after a small recovery. The dashed outline is what disappeared. TomTom's investor relations director used cameras as his example of a category the phone would not replace one year after this peak.

Source: Digital Camera World, CIPA data, as cited in the essay.

The map companies were not careless. They watched their competitors closely, and their competitors were other map companies. When you map your market by category, the threat that ends the category never shows up on the map.

I have spent this year writing about the ways an industry can disappear. The capital can leave. The need can move into a new container. The customer can start building the product instead of buying it. The industry can compress until far fewer companies do the same work. The iPhone shows a fifth way, and it is the hardest to see from the inside: the whole category gets absorbed by something that was never in it.

A better product can become a worse purchase

The iPhone calculator was no better than a good calculator, and its camera was no better than a good camera. The iPhone was not competing against the best map or the best calculator. It was competing against free and already in your pocket. Nobody shops for a calculator while holding one, so the specialist is never rejected. It is simply never considered, and a product that is never considered leaves no trace. Nobody writes down why the customer did not buy it, nobody calls to ask why they left, and there is no warning before it simply stops being bought.

Figure 2

Never rejected, simply never considered

How to read it. A diagram of the argument, not data. Before the phone, the specialist competes: the customer compares satnavs and buys the best one. After the phone, the same need goes straight to the phone that is already in the pocket. The satnav never enters the comparison, so it loses without ever being rejected, and no lost-deal report records it.

That is why it is so hard to see from inside a product team. TomTom kept shipping better devices, and every demonstration was honest. The alternative in the customer's pocket was improving faster. A company can make real progress on its product while the argument for buying it gets weaker.

The LLM does not need the app

An LLM is the technology behind ChatGPT, Claude and Gemini: a system you talk to in ordinary language that writes, explains, translates, calculates and increasingly acts for you. More than a billion people a week now open ChatGPT, bringing whatever they need help with that day: a translation, a contract, a lesson, a spreadsheet, a gift for their mother.

Today's LLMs are the first iPhone. They still make things up, get numbers wrong and fall short of good specialist software at plenty of jobs, and software companies are comparing them feature by feature with their own products. That is Ballmer's mistake again. Many people do pay for these assistants, but they pay for a dozen other reasons, so for any single software product the assistant is already paid for. Every specialist is now competing for an additional purchase, another login and another place to explain what the customer wants.

And the LLM goes further than the iPhone did. The iPhone turned devices into apps, but you still had to find and open the right one. The LLM turns apps into a sentence: nobody needs a translation tool when they can just ask for the translation. It also knows who is asking. The phone never knew why you wanted the map, while an assistant you have used for months may already know the brief, the failed attempts and you. A specialist that starts from zero has to be better by enough to make the customer explain everything again.

Investors have started to picture what fits inside that sentence. When Anthropic released a set of business tools for Claude in January, software stocks lost more than $285 billion in four days. I made this argument about edtech in my first essay. It holds for every product whose job fits in a sentence.

Figure 3

The phone took the gadgets. The LLM takes the software.

9product categories folded into one phone. Five show a documented decline; for four there is no clean sales data.

Satnav

documented decline

Replaced by: Free turn-by-turn navigation on the phone

The day Google announced free navigation for phones (October 28, 2009), TomTom fell 21% and Garmin more than 16%.

Source: TechCrunch

documented declineabsorbed, no clean sales data

How to read it. Switch between 2007 and 2026, then tap any square or sentence to see what it replaced and the evidence. On the phone, each square is a product category that used to be a separate thing you bought. In the LLM, each category becomes the sentence a person now types instead of opening a product. On the phone side, filled squares have a documented decline and outlined squares do not, because nobody tracks those categories cleanly. The LLM side is my argument, counted in categories rather than companies, because it is still too early to say which companies fall. Wristwatches are left out on purpose: luxury watches survived the phone, so they do not belong on this list.

Some will survive by going where the phone cannot follow

A software company's strongest answer is that it holds what a chatbot does not: the customer's data, the connections to their other systems, the security approvals and the signed contracts. Holding those things buys time, but it is not what saved the map companies. Garmin survived by moving: it had record revenue in 2025 earned from fitness watches, boats and aircraft, the places a phone could not follow. The map company survived by becoming something else.

Others will survive inside the new device, as thousands of companies did on the iPhone. They built their businesses as apps, and Apple took up to 30 percent of everything they sold. They survived, but they paid rent to the company that owned the phone. Figma, Adobe, Salesforce and HubSpot already run inside ChatGPT. Some will keep the sale but lose the customer's decision. A booking site may still take your hotel reservation and get paid for it, but the assistant has already decided which hotel you want, so the booking site no longer shapes the choice.

Every software company is the map company now

In 2007, most of us held the thing that would take out the map companies in our hands and thought we were holding a phone. The map companies looked at the same object and saw nothing to worry about. Nobody was wrong about what it was that year. Everyone was wrong about what it would become.

I can feel the same shift in my own life right now. One by one, I am canceling subscriptions I used to pay for without thinking, because the job now happens inside ChatGPT or Claude. Translation went first. Transcription went next.

As a Dane living in Mexico, I basically lived in Google Translate. I am trying hard to learn Spanish, but most of my life here happens in Spanish, and I cannot always keep up. The hardest place is the parents' group chat for Club América's Nidos Águila academy, where my son plays. The messages come fast, and there are a lot of them. Before, I copied each message into Google Translate, read it in English, wrote my answer, translated it back into Spanish and pasted it into the chat. Google Translate saw one message at a time. It did not know who the coach was, what had been decided last week or what I was trying to say. Today I simply pass the messages to Claude. It translates them, remembers what was said before, and drafts my reply in Spanish, so the whole conversation happens in one place.

I also used to pay a monthly subscription to a transcription service, a tool that turns a recording into written text. I used it for my voice notes, interviews and work conversations: I uploaded the recording, waited for the text to come back, and then copied it into wherever I was actually working. Today I send the voice note straight to Claude. The text appears in the same window where I am already writing, and in the next sentence I can ask Claude to shorten it, summarize it or turn it into an email. Does Claude have a billion different categories where I can tag and store every recording? No. But it gets the job done at no extra cost, inside the system I already use all day, so I canceled the subscription. The transcription company did not lose me to a better transcription tool. It lost me to a sentence.

And it is not just me. For years, programmers who got stuck took their problem to Stack Overflow, a public website where strangers answer each other's coding questions. You wrote up your question, posted it and waited, sometimes for hours, for someone you did not know to reply, and the answer might not even fit your code. Today a programmer pastes the error and the code straight into their preferred LLM and gets an answer in seconds, written for their exact problem, with no stranger and no waiting. New questions on Stack Overflow have fallen from about 200,000 a month to almost none. None of these products lost to a better competitor in its own category. They lost to a tool people were already using for everything else.

I write this from both sides. I have spent twelve years selling B2B SaaS, software that businesses pay for as a monthly or yearly subscription, and today I own, advise and invest in a series of software companies. In 2007, I was the customer holding the phone. Today I am canceling other people's subscriptions while selling my own, and that puts me at risk of being the map company too.

Today, more than a billion people a week talk to an LLM and think they are using a chatbot. Most software companies look at it and see a tool that still makes mistakes, and they cannot yet fathom what it is about to take from them. That is exactly how the iPhone looked from the map company's desk.

The iPhone never had to be a better calculator, camera or map. It only had to be free and already in your pocket. The LLM does not have to be better than your software either. It only has to be the window your customers already have open. The map company found out in a single day in October 2009. Most software companies will find out the same way: not from a competitor they were watching, but from something their customers were already using every day.