AI has failed to replace a single software application or feature
23 comments
AI has completely killed Google Translate. It is significantly superior in every way.
GT often uses the wrong word or changes the tone of a message. AI always gets the intent right and always seems to use the most appropriate words given the original intent/meaning (or at least something way better than what GT does). And, whenever there is doubt, I can argue with it, so AI is happy to explain the nuances and differences between the possibilities.
Edit: I recently had to send a semi-formal email requesting something from a government employee in a different country (using a language I'm a beginner at), and AI was immensely helpful in getting the right tone (neither informal or too formal) and everything else right. The Google Translate version of what I had originally written was miles and miles and miles worse than what AI helped me craft.
GT often uses the wrong word or changes the tone of a message. AI always gets the intent right and always seems to use the most appropriate words given the original intent/meaning (or at least something way better than what GT does). And, whenever there is doubt, I can argue with it, so AI is happy to explain the nuances and differences between the possibilities.
Edit: I recently had to send a semi-formal email requesting something from a government employee in a different country (using a language I'm a beginner at), and AI was immensely helpful in getting the right tone (neither informal or too formal) and everything else right. The Google Translate version of what I had originally written was miles and miles and miles worse than what AI helped me craft.
> AI always gets the intent right and always seems to use the most appropriate words given the original intent/meaning
How do you know? Why are you using something to translate if you’re a native speaker of both languages?
How do you know? Why are you using something to translate if you’re a native speaker of both languages?
I said I am a beginner in one of the languages.
How do I know their answer is best? I can verify their answers through other means, and I understand both languages enough to realize one answer is more appropriate than the other, even if I can't come up with those words directly using only my brain.
How do I know their answer is best? I can verify their answers through other means, and I understand both languages enough to realize one answer is more appropriate than the other, even if I can't come up with those words directly using only my brain.
Google Translate uses AI...
https://blog.google/products-and-platforms/products/search/g...
https://blog.google/products-and-platforms/products/search/g...
True, but Google Translate was already "AI". They previously used LSTMs. And before LSTMs, it was ML-like statistical translation.
Conventional web search: I use Google's AI Mode almost exclusively when I use Google.
Refactoring tools in my IDE: In some cases where I could use the refactoring tools in my IDE I will ask the assistant to do something for me instead, of course it will also make changes that the refactoring tools won't do such as tear apart a complicated if-then-else ladder.
Photo retouching: there are plenty of photo retouching jobs that can be done easily with AI or with the tools built into Photoshop, which one is better depends on the situation. I really wish I had an AI tool to make masks ("cut out the person") that I could then use with the other tools or with AI generation.
Refactoring tools in my IDE: In some cases where I could use the refactoring tools in my IDE I will ask the assistant to do something for me instead, of course it will also make changes that the refactoring tools won't do such as tear apart a complicated if-then-else ladder.
Photo retouching: there are plenty of photo retouching jobs that can be done easily with AI or with the tools built into Photoshop, which one is better depends on the situation. I really wish I had an AI tool to make masks ("cut out the person") that I could then use with the other tools or with AI generation.
I think you are asking the wrong question. AI has not replaced Excel, but it has replaced the need for me to hire developers to build features.
I am a non-technical founder (background in growth/marketing). Two years ago, building a production web application required either (1) learning to code for months/years, (2) hiring engineers at $150-200K/year, or (3) outsourcing to contractors and praying.
None of those options were viable for me to validate an idea quickly.
With Cursor, v0, and similar tools, I built our entire frontend in production. Not a prototype, not an MVP in the old sense, but actual production code serving real customers. Features that would have cost $5-10K to spec and outsource, I can now build in an afternoon.
The tools did not "replace Excel" or any specific application. What they replaced was my dependency on other people to execute technical work. That is a much more profound shift than feature replacement.
The comparison to traditional applications is a category error. AI is not competing with Excel. It is competing with the labor market.
I am a non-technical founder (background in growth/marketing). Two years ago, building a production web application required either (1) learning to code for months/years, (2) hiring engineers at $150-200K/year, or (3) outsourcing to contractors and praying.
None of those options were viable for me to validate an idea quickly.
With Cursor, v0, and similar tools, I built our entire frontend in production. Not a prototype, not an MVP in the old sense, but actual production code serving real customers. Features that would have cost $5-10K to spec and outsource, I can now build in an afternoon.
The tools did not "replace Excel" or any specific application. What they replaced was my dependency on other people to execute technical work. That is a much more profound shift than feature replacement.
The comparison to traditional applications is a category error. AI is not competing with Excel. It is competing with the labor market.
The big one for me is shell scripting and working with the terminal in general. For everything except for the simplest commands, I find myself preferring to ask in plain English, and while imperfect, it has both saved me time and decreased the number of times I've accidentally deleted/broken things, compared to me doing it manually.
We've already reached the point where effectively using AI means learning how to resist it.
Transitions take time. 0 -> 1 replacement of existing interfaces is (1) not easy and (2) not a practical approach to market adoption.
"slapping on a chatbot" is a v0 attempt at re-imagining what software looks like. It's not very inventive and sometimes it sucks, but it's easy to understand and implement and we're very early in this era.
The distribution of change also isn't uniform. Excel might not have changed dramatically, but software engineering apps are evolving rapidly. Clawdbot/moltbot hint at new forms of personal computing. Look for the future where the optimism is.
"slapping on a chatbot" is a v0 attempt at re-imagining what software looks like. It's not very inventive and sometimes it sucks, but it's easy to understand and implement and we're very early in this era.
The distribution of change also isn't uniform. Excel might not have changed dramatically, but software engineering apps are evolving rapidly. Clawdbot/moltbot hint at new forms of personal computing. Look for the future where the optimism is.
It’s killed my need for junior or mid level “human LLMs” on cloud + app dev projects I’m responsible for leading. I don’t need junior and mid level ticket takers anymore just to do grunt work
I always had to do the requirements gathering, detailed work items, detailed architecture when assigning the work to someone else. Now I have the bandwidth to do it myself and not have to coordinate with three other people.
I always had to do the requirements gathering, detailed work items, detailed architecture when assigning the work to someone else. Now I have the bandwidth to do it myself and not have to coordinate with three other people.
To focus on Excel / Google Sheets itself, AI makes it a lot more usable in that you can describe what you want and it can generate formulas for you. I have a running spreadsheet that I had to manually manipulate because I didn't know how to automate some parts and a few minutes with an LLM fixed it for me.
I agree that it hasn't killed software, but it has made a lot of tasks that used to take longer a lot easier. For instance, it's much easier to add features or build an MVP than it used to be, but only if you know what you're doing.
Visual Studio Code for one, as well as Jetbrains stuff.
I just opened up an old PC to install Moltbot. I deleted WebStorm while waiting on the installation as that's one piece of software I don't expect to use again.
I just opened up an old PC to install Moltbot. I deleted WebStorm while waiting on the installation as that's one piece of software I don't expect to use again.
But it has filled everything else. Current software is solving all the problems we know how to solve. There s really no need to replace it.
But AI is not a software application, it does not aim to replace software applications.
It's like saying that the discovery of steel did not replace any existing weapon or tool.
It's like saying that the discovery of steel did not replace any existing weapon or tool.
Try "Claude in Excel", its pretty good.
This is certainly a take.
Thanks G
You've written and deleted several of these "AI doomer" posts this week. I think you're projecting a deep-rooted personal desire and not really addressing the pragmatic preexisting demand that makes AI important. War, misinformation, automation, fraud, pornography, none of these applications are going away. In many ways, they're more accessible and rewarding than ever before. If you want to hedge a bet on humanity making terrible choices, then AI is a perfectly antifragile investment.
It's not pleasant to imagine the full spectrum of AI applications. The same could be generalized for edtech, defense, surveillance, security and privately-owned prison economics. Alas, they're still with us and immensely lucrative.
It's not pleasant to imagine the full spectrum of AI applications. The same could be generalized for edtech, defense, surveillance, security and privately-owned prison economics. Alas, they're still with us and immensely lucrative.
Nobody uses Excel anymore, now everyone uses Microsoft 365 Copilot!
complete nonsense, like saying cars have failed to replace a single horse application or feature.
I'm so sorry to hear that you are falling behind. With this attitude you are going to keep falling behind. Good luck.
Now you might ask: what about AI native applications? Well, as it turns out, most of them are essentially clones of existing software but with a chatbot slapped on top. Because of the error prone nature of AI, any application that leverages it also has to surface all of the controls required to override all of its decisions. So you end up with a traditional software application plus AI.
AI promised to transform and even replace software applications, but all it did instead was augment them with an unreliable chatbot. All of the old fields and buttons are still there, but now there’s an additional text field that you can type into and hope for the best.