If AI Ruins the Value of Software…

 

I just watched DeepSeek create a video game from a few prompts. And it is becoming clear that thanks to AI, anything we can conceive of that can be created by software is inevitably going to become ubiquitous, and thus valueless. Isn’t it only a matter of time before a Microsoft Word or Excel clone can be generated and delivered without having to pay Microsoft? If so, is this not true for any software at all (at least those not subject to the strict engineering QA requirements for safety-critical systems like commercial aircraft)?

If AI really can make all software cheaper, with extremely rapid development times, where does this lead us? Where do people add unique value?

I think the value of people moves away from software (beyond running the inputs and outputs to AI itself). And it moves back into traditional hardware – everywhere AI and automation cannot penetrate. Blue collar jobs, construction. In high tech, the “cookbook” sciences like steel making and metallurgies… everything that only a craftsman can do.

Because not everything in the world is software.

Is this right?

Published in General
This post was promoted to the Main Feed at the recommendation of Ricochet members. Like this post? Want to comment? Join Ricochet’s community of conservatives and be part of the conversation. Join Ricochet for Free.

There are 54 comments.

Become a member to join the conversation. Or sign in if you're already a member.
  1. iWe Coolidge
    iWe
    @iWe

    To follow… if AI can produce any software, then what unique value does Word or X or Facebook even offer, besides an inertial customer base?

     

    • #1
  2. David Foster Member
    David Foster
    @DavidFoster

    The inertial customer base (aka the ‘network effect’) is hugely important.  Suppose it were possible to create an equivalent of Ricochet with a few prompts. You’d still have a long road ahead to build a community interested in communicating with & interacting with one another.

    OTOH, suppose it were easy to use AI prompts to improve search for Ricochet posts relevant to a particular topic…that would add value to the customer base that already exists.

    • #2
  3. iWe Coolidge
    iWe
    @iWe

    David Foster (View Comment):

    OTOH, suppose it were easy to use AI prompts to improve search for Ricochet posts relevant to a particular topic…that would add value to the customer base that already exists.

    It sure would.

    • #3
  4. David Foster Member
    David Foster
    @DavidFoster

    I used a LLM to create a post, with several follow-on prompts to improve it:  Retrotech: An Automated Flour Mill in 1785.

    And I used a different LLM to bring a historical individual, a noted diarist, into the present where she continued her diary: LLM Time Travel: Fanny Kemble in 2024.

    • #4
  5. Dr. Bastiat Member
    Dr. Bastiat
    @drbastiat

    This will be very interesting…

    • #5
  6. Henry Racette Member
    Henry Racette
    @HenryRacette

    I have a nephew who recently graduated from college with a degree in something-related-to-movies who is gradually breaking into the voice actor business. He has an agent now and paid work, and he might be able to establish himself, but AI will soon make entry in to that business extraordinarily difficult, I expect, as it replaces “stock” voices on lots of small projects.

    Software will be similar: it’s going to decrease the need for the least skilled people. It’s hard for me to imagine AI producing finished products of any significant complexity any time soon, but it’s easy to imagine AI tripling or quadrupling the productivity of a good programmer. I’m sure that’s coming, and the result, I suspect, will be that weaker programmers will find it harder to get work.

    My own field is, I think, pretty AI-proof, at least for now. I write software for  complex machines that don’t yet exist. That means that there are no extant examples on which AI can be trained and, until AI can sit in the room with the engineers and collaborate on the design with them, flesh-and-blood programmers like me will be required.

    At the rate things are moving, I’ve got another few years. Probably.

    • #6
  7. DonG (¡Afuera!) Coolidge
    DonG (¡Afuera!)
    @DonG

    It used to be that software engineers made money by plugging 1’s and 0’s into the right place.   Nobody does that now.  Today’s coders are now completely isolated from the actual hardware running the code. Perhaps driving AI to get software is just the next evolution in software engineering.

    • #7
  8. She Member
    She
    @She

    iWe:

    If AI really can make all software cheaper, with extremely rapid development times, where does this lead us? Where do people add unique value?

    I think the value of people moves away from software (beyond running the inputs and outputs to AI itself). And it moves back into traditional hardware – everywhere AI and automation cannot penetrate. Blue collar jobs, construction. In high tech,  the “cookbook” sciences like steel making and metallurgies … everything that only a craftsman can do.

    Because not everything in the world is software.

    So, my takeaway from this is:

    “Don’t Learn to Code.”

    Is this right?

     

    • #8
  9. iWe Coolidge
    iWe
    @iWe

    She (View Comment):

    So, my takeaway from this is:

    “Don’t Learn to Code.”

    Is this right?

    Seems like it.

    • #9
  10. kedavis Coolidge
    kedavis
    @kedavis

    DonG (¡Afuera!) (View Comment):

    It used to be that software engineers made money by plugging 1’s and 0’s into the right place. Nobody does that now. Today’s coders are now completely isolated from the actual hardware running the code. Perhaps driving AI to get software is just the next evolution in software engineering.

    Yes, it does seem as if first-step software such as operating systems, compilers, etc. will continue to be human-driven since at those stages there is nothing to “train” an AI to start with.  And for systems with narrower distribution, there may never be.

    But to follow on from what Henry notes, beginner/entry-level/not-very-good programmers were never working on that stuff anyway.

    • #10
  11. Randy Weivoda Moderator
    Randy Weivoda
    @RandyWeivoda

    Something tells me it’s not as easy as that.  You probably can tell an AI to write a spreadsheet or word processing program and maybe it can come up with something useful.  It’s got existing programs to copy from.  How good will the programs be?  Will there be any innovations over what the existing programs deliver?  Can an AI do beta testing to make sure it works right?   What about creating software to do something new, where it cannot copy from existing examples? 

    I could picture it being a useful tool for programmers. It would allow them to do a bunch of routine stuff that builds the basic structure of a program, and then they could do a bunch of fine-tuning to make it work well. 

    • #11
  12. kedavis Coolidge
    kedavis
    @kedavis

    Randy Weivoda (View Comment):

    Something tells me it’s not as easy as that. You probably can tell an AI to write a spreadsheet or word processing program and maybe it can come up with something useful. It’s got existing programs to copy from. How good will the programs be? Will there be any innovations over what the existing programs deliver? Can an AI do beta testing to make sure it works right? What about creating software to do something new, where it cannot copy from existing examples?

    I could picture it being a useful tool for programmers. It would allow them to do a bunch of routine stuff that builds the basic structure of a program, and then they could do a bunch of fine-tuning to make it work well.

    That’s certainly possible. And as with other things, it could actually turn out to be somewhat counter-productive if it takes as long or longer to verify what the AI produced, as to do it from scratch by people.  Lawyers who submit AI-produced documents to courts are finding that out, as AI will apparently just cite made-up cases that don’t exist, as well as mis-stating/mis-representing cases that DO exist…

    • #12
  13. Randy Weivoda Moderator
    Randy Weivoda
    @RandyWeivoda

    iWe: Isn’t it only a matter of time before a Microsoft Word or Excel clone can be generated and delivered without having to pay Microsoft?

    If someone wants programs with the same basic functionality of Word and Excel without paying Microsoft, that has been available for years.  A bunch of us on Ricochet are happy users of LibreOffice, which you can legally download and use for free.  If you like it, you can kick in a few bucks to support the project.  I think I sent them $15 or $20.  I sometimes use Excel at work, but very much prefer Calc, the LibreOffice spreadsheet program.

    I just don’t see every Tom, Dick, and Harry creating their own software through AI, rather than getting something made by professionals. 

    In my former job (in my family’s business), we had a custom database program made for us because there wasn’t anything commercially available that suited our needs.  This project started back in the late 1980s.  There are likely fine software packages available off the shelf for that purpose today, but not then.  My father hired a programmer, didn’t get very far, fired him, and restarted with someone else again and again.  Eventually, I took over developing this and we got a set of programs that worked very well.  The programmer (Mike) had several different clients past and present and said that he wished that he had someone like myself at each of those businesses, because most people have a vague notion of what they want their software to do and are terrible at explaining it.  And then when Mike gave them what he thought they wanted, they didn’t test it and it would be months before they got back to him and said that wasn’t what they wanted at all.  I have heard from other programmers who have had similar experiences with clients.  This suggests to me that the average person is not going to be able to describe to an AI what they want and how it should work, since the average person cannot adequately communicate these things to another human being.

    • #13
  14. kedavis Coolidge
    kedavis
    @kedavis

    Randy Weivoda (View Comment):

    This suggests to me that the average person is not going to be able to describe to an AI what they want and how it should work, since the average person cannot adequately communicate these things to another human being.

    Quite true, also the people who do certain tasks such as entering payroll data may themselves have no real idea what the requirements are that the software must adhere to by federal or state laws, etc.  Basically, the software cannot be designed by the people who actually use it; and the users then have to conform to the software, not the other way around.

    I benefited by my work either being used by the business I worked at, when I was in-house, or by other local customers as well as distant ones when I worked at “software houses,” so I could see how people were actually using it and could make changes so it was easier to use while still meeting the requirements that the direct users may have been unaware of.

    For example, a direct user – aka., a data entry person – may think it was pointless to enter “days worked” in addition to “hours worked” as part of payroll.  And if they were designing/describing what THEY THOUGHT a payroll system “should be,” they would likely leave out the “days worked.”

    But “days worked” was required for various reasons such as whether some hours were “overtime” or not, whether an employee qualified as “full time” during a given period for insurance purposes…

    So while “days worked” couldn’t be eliminated, I did make allowances for making it easier for the data-entry people to use.  Such as, the usual default for “days worked” on a weekly payroll entry schedule was 5.  But if the people who, say,  operated a particular machine only worked 3 days, or even better if it was a 3-day week for EVERYONE because of a holiday or something, it was annoying for the operator to change 5 to 3 EVERY. SINGLE. TIME.

    But on the other hand, if just ONE PERSON only worked 3 days in a week, changing the default to 3 just for that one person and then having to change back again to 5 for everyone else, could lead to mistakes as well.

    So I made it so that if the “days worked” was changed to a different number, and the SAME different number, for, like, 3 people in a row, then that change became the default.

    They loved it.

    Those systems also had “intelligent terminals” where the screens had pre-determined fields for text, numbers, etc.  And the original simple design of the programs I “inherited,” “positioned” the entry cursor to each data field as they went through.

    However, the “smart terminals” were also mostly moving the cursor to the next field already, and if the operator was already typing the next field, the cursor would be re-positioned to the start and the terminal would beep, rather loudly in fact.

    The payroll-entry people were pretty fast, and so payroll-entry day would be accompanied by a lot of beeping and low muttering from that office.

    I modified that program so that as long as there wasn’t an error in the previous field, the repositioning was skipped and the terminal just moved through the fields without interference.

    They loved that, too.

     

    Somehow, I don’t think AI would have “thought” of either of those.

    • #14
  15. Percival Thatcher
    Percival
    @Percival

    Randy Weivoda (View Comment):

    iWe: Isn’t it only a matter of time before a Microsoft Word or Excel clone can be generated and delivered without having to pay Microsoft?

    If someone wants programs with the same basic functionality of Word and Excel without paying Microsoft, that has been available for years. A bunch of us on Ricochet are happy users of LibreOffice, which you can legally download and use for free. If you like it, you can kick in a few bucks to support the project. I think I sent them $15 or $20. I sometimes use Excel at work, but very much prefer Calc, the LibreOffice spreadsheet program.

    I just don’t see every Tom, Dick, and Harry creating their own software through AI, rather than getting something made by professionals.

    In my former job (in my family’s business), we had a custom database program made for us because there wasn’t anything commercially available that suited our needs. This project started back in the late 1980s. There are likely fine software packages available off the shelf for that purpose today, but not then. My father hired a programmer, didn’t get very far, fired him, and restarted with someone else again and again. Eventually, I took over developing this and we got a set of programs that worked very well. The programmer (Mike) had several different clients past and present and said that he wished that he had someone like myself at each of those businesses, because most people have a vague notion of what they want their software to do and are terrible at explaining it. And then when Mike gave them what he thought they wanted, they didn’t test it and it would be months before they got back to him and said that wasn’t what they wanted at all. I have heard from other programmers who have had similar experiences with clients. This suggests to me that the average person is not going to be able to describe to an AI what they want and how it should work, since the average person cannot adequately communicate these things to another human being.

    That is a very common issue. What the customers explicitly asks for is somewhat nebulous. What they don’t want will become apparent once they see it. The technique of “rapid prototyping” can ease the user interface issues, and with clever test cases can uncover others.

    • #15
  16. Randy Weivoda Moderator
    Randy Weivoda
    @RandyWeivoda

    kedavis (View Comment):

    Randy Weivoda (View Comment):

    This suggests to me that the average person is not going to be able to describe to an AI what they want and how it should work, since the average person cannot adequately communicate these things to another human being.

    Quite true, also the people who do certain tasks such as entering payroll data may themselves have no real idea what the requirements are that the software must adhere to by federal or state laws, etc. Basically, the software cannot be designed by the people who actually use it; and the users then have to conform to the software, not the other way around.

    I benefited by my work either being used by the business I worked at, when I was in-house, or by other local customers as well as distant ones when I worked at “software houses,” so I could see how people were actually using it and could make changes so it was easier to use while still meeting the requirements that the direct users may have been unaware of.

    For example, a direct user – aka., a data entry person – may think it was pointless to enter “days worked” in addition to “hours worked” as part of payroll. And if they were designing/describing what THEY THOUGHT a payroll system “should be,” they would likely leave out the “days worked.”

    But “days worked” was required for various reasons such as whether some hours were “overtime” or not, whether an employee qualified as “full time” during a given period for insurance purposes…

    So while “days worked” couldn’t be eliminated, I did make allowances for making it easier for the data-entry people to use. Such as, the usual default for “days worked” on a weekly payroll entry schedule was 5. But if the people who, say, operated a particular machine only worked 3 days, or even better if it was a 3-day week for EVERYONE because of a holiday or something, it was annoying for the operator to change 5 to 3 EVERY. SINGLE. TIME.

    But on the other hand, if just ONE PERSON only worked 3 days in a week, changing the default to 3 just for that one person and then having to change back again to 5 for everyone else, could lead to mistakes as well.

    So I made it so that if the “days worked” was changed to a different number, and the SAME different number, for, like, 3 people in a row, then that change became the default.

    They loved it.

    Those systems also had “intelligent terminals” where the screens had pre-determined fields for text, numbers, etc. And the original simple design of the programs I “inherited,” “positioned” the entry cursor to each data field as they went through.

    However, the “smart terminals” were also mostly moving the cursor to the next field already, and if the operator was already typing the next field, the cursor would be re-positioned to the start and the terminal would beep, rather loudly in fact.

    The payroll-entry people were pretty fast, and so payroll-entry day would be accompanied by a lot of beeping and low muttering from that office.

    I modified that program so that as long as there wasn’t an error in the previous field, the repositioning was skipped and the terminal just moved through the fields without interference.

    They loved that, too.

     

    Somehow, I don’t think AI would have “thought” of either of those.

    I always liked little tricks that could save keystrokes.  I asked the programmer if we couldn’t simplify entering dates, by allowing the user to enter T for Today and Y for Yesterday.  Then I expanded it to all these other shortcuts in date fields.

    T or +0 or -0 = Today’s Date
    Y or -1 = Yesterday’s Date
    +nnn = num of days into future
    -nnn = num of days into past
    Mdd = This Month, dd=Day
    Ndd = Next Month, dd=Day
    Ldd = Last Month, dd=Day
    Emm = Last Day, mm=Month
    C = Calendar

    Calendar would pop up a graphic calendar where you could arrow to the date you wanted.  I don’t know what percentage of our employees ever took advantage of these and various other shortcuts, though.  It seems like people beyond a certain age were satisfied to know one way of doing something and were damned if they were going to learn another way, even if it saved time and effort.  I wonder when luxury cars first started coming out with power door locks, did old people still unlock each door manually because they didn’t want to learn to push a button one way or the other?

    • #16
  17. kedavis Coolidge
    kedavis
    @kedavis

    Randy Weivoda (View Comment):

    I always liked little tricks that could save keystrokes.  I asked the programmer if we couldn’t simplify entering dates, by allowing the user to enter T for Today and Y for Yesterday.  Then I expanded it to all these other shortcuts in date fields.

    T or +0 or -0 = Today’s Date
    Y or -1 = Yesterday’s Date
    +nnn = num of days into future
    -nnn = num of days into past
    Mdd = This Month, dd=Day
    Ndd = Next Month, dd=Day
    Ldd = Last Month, dd=Day
    Emm = Last Day, mm=Month
    C = Calendar

    Calendar would pop up a graphic calendar where you could arrow to the date you wanted.  I don’t know what percentage of our employees ever took advantage of these and various other shortcuts, though.  It seems like people beyond a certain age were satisfied to know one way of doing something and were damned if they were going to learn another way, even if it saved time and effort.  I wonder when luxury cars first started coming out with power door locks, did old people still unlock each door manually because they didn’t want to learn to push a button one way or the other?

    I can see those kinds of shortcuts being useful in some situations, like for example retail point-of-sale.  But the systems I developed were being used for things like entering Accounts Payable invoices and Accounts Receivable invoices…  which arrive pretty randomly over the course of a month and could refer to a wide range of dates.  Trying to figure out “how many days in the past?” or “how many days in the future?” would actually slow things down.

    Also, those systems didn’t really do pop-ups.  The limit was pretty much looking up customers or vendors by name rather than number.

    • #17
  18. Randy Weivoda Moderator
    Randy Weivoda
    @RandyWeivoda

    kedavis (View Comment):
    I can see those kinds of shortcuts being useful in some situations, like for example retail point-of-sale.  But the systems I developed were being used for things like entering Accounts Payable invoices and Accounts Receivable invoices…  which arrive pretty randomly over the course of a month and could refer to a wide range of dates.  Trying to figure out “how many days in the past?” or “how many days in the future?” would actually slow things down.

    That’s fine.  I wasn’t suggesting that years ago you should have included such code in your programs.  I just thought I’d share a detail about my experience.

    • #18
  19. TBA, sometimes known as 'Teebs'. Coolidge
    TBA, sometimes known as 'Teebs'.
    @RobtGilsdorf

    ~busily seeds a Basti-bot with which to take over the internet~ 

    • #19
  20. genferei Member
    genferei
    @genferei

    AI won’t be reducing the value of software, although it might reduce its price.

    One area where I see it contributing to the massive increase in the aggregate value of all software is in one-shot or one-person code. There is a huge overhead today in creating a piece of software to solve a problem. The moment you have a user who isn’t the programmer, all sorts of things outside of solving the core problem become necessary (maintainability, security updates, feature creep…). Therefore developers are required. To pay their salaries sales must be made. Which means sales people. To pay their salaries you need a certain price point. Which means that any problem must be faced by a sufficient number of people who can convince sufficient other people (budget holders, typically not those who have the actual problem; and IT, who are going to have to maintain the human and technical infrastructure to support the solution) to say yes to the purchase and training costs. And now it’s 12 months later and a whole lot of hassle. 

    Or describe the problem to your agentic AI and get the 80% solution in the next five minutes. 

    By ‘problem’ here I don’t mean a platform-like thing like a word processor or a spreadsheet, but some task that might or might not be performed using a platform. ‘Find all the (pdf) invoices that come from suppliers with a Texas address’ is a problem, but a one-shot solution isn’t a product, so the ordinary software business isn’t going to help. An AI generated solution – smuggling general purpose programming back into the hands of users – won’t be properly documented, version controlled or unit tested. Or perfect. But it might save a human a day’s or a week’s work. 

    Happy days, and not a single (professional) programmer loses their job. 

    • #20
  21. Percival Thatcher
    Percival
    @Percival

    The AI will  require access – intimate access – to your current systems and data while designing and implementing solutions for you.

    Question: Is the AI going to keep the information it gleans while analyzing your systems and data to itself, or will it apply any of that knowledge to future products it produces for others, including your competitors? If someone asks for your customer list, will it hand the list over?

    • #21
  22. Henry Racette Member
    Henry Racette
    @HenryRacette

    genferei (View Comment):
    describe the problem to your agentic A

    I’ve been thinking of AI as a programmer’s assistant, something an experienced programmer would use to flesh out bits of code, quickly come up to speed on unfamiliar code libraries, etc. My oldest boy uses it that way already and says it saves him a lot of time.

    Using AI as an agent, as genferei says, is likely to have a bigger impact, once major software developers begin incorporating AI interfaces into their products. We’re not talking individual desktop applications here, spreadsheets and such, though I suppose they will make some use of AI as well. But large systems in banking, manufacturing, sales, customer support, health care, etc., require relatively low-skill programmers (or specially trained users) to create custom reports and presentations of data, and AI could easily replace that function in the very near future — if not already.

    Junior programmers get tasked with producing a lot of custom reports, spreadsheets, charts, and other management-support material that requires an understanding of basic database query language and the various formatting tools provided by enterprise-level software products. That function can readily be replaced by AI.

    Web page generation is another thing that AI can take over pretty quickly. A lot of (perhaps most) “custom” web developers already use canned website generation software that makes it easy to lay out basic sites quickly (often locking the customer into their platform in the process). Those custom developers won’t entirely go away, but big web hosting companies will implement AI-based web design services that cost anyone can use to quickly create and maintain their new website.

    Just as order-taking kiosks in fast food are eliminating entry-level jobs, AI will eliminate the need for a lot of low-skill developers. This has the effect of breaking off the lowest rungs on the career ladder.

    So what comes after “learn to code,” when that’s no longer a viable path for a large number of inexperienced young or displaced people? I don’t know.

    • #22
  23. Brady | @HerrForce1 Coolidge
    Brady | @HerrForce1
    @HerrForce1

    iWe (View Comment):

    She (View Comment):

    So, my takeaway from this is:

    “Don’t Learn to Code.”

    Is this right?

    Seems like it.

    “I just want to say one word to you. Just one word. Plastics.”

    • #23
  24. Keith Lowery Coolidge
    Keith Lowery
    @keithlowery

    I use AI to generate code pretty consistently. In the last quarter of last year I was able to build some software in 3 months that would have taken me 9-12 months prior to large language models. 

    So it has been a real boon to my personal productivity. But there is no way I could have gotten AI to do all the work for me for the simple reason that the complexity of what I was doing exceeded my ability to express it to the language model. I could actually write the code before I could express, in sufficient detail, what I wanted the AI to write. There seems to be an upper bound to how we take advantage of AI. The limit is associated, I think, with the total complexity of what we’re trying to do. I can tell AI to do some discreet functionality (e.g. “Write a function that plots the contents of a CSV file as a line graph using matplotlib”) and it’s really helpful. But it is another thing entirely to get it to do things larger and vastly more complex.

    David Gelernter made this observation about programming: 

    A good programmer can sit down at the keyboard and build a program – a working piece of software – with nearly the complexity of an aircraft carrier all by himself, to his own designs and no one else’s

    The barrier to using AI to write entire programs is that vast numbers of those programs have such complexity that expressing what you want the AI to do exceeds our abilities, at least where language models are concerned (e.g. Grok or ChatGPT). And if we could express it, it is very unclear that we could do so in less time and effort than simply writing the code in the first place. I have made the observation before that what seems relatively small in our heads will often expand to fill all available space when we try to express it in words. The ability of humans to hold ideas in our heads that feel simple, but are massive and complex to put into words, will place natural limits on AI for a very long time. IMHO

    • #24
  25. Randy Weivoda Moderator
    Randy Weivoda
    @RandyWeivoda

    Keith Lowery (View Comment):
    The barrier to using AI to write entire programs is that vast numbers of those programs have such complexity that expressing what you want the AI to do exceeds our abilities, at least where language models are concerned (e.g. Grok or ChatGPT). And if we could express it, it is very unclear that we could do so in less time and effort than simply writing the code in the first place. I have made the observation before that what seems relatively small in our heads will often expand to fill all available space when we try to express it in words. The ability of humans to hold ideas in our heads that feel simple, but are massive and complex to put into words, will place natural limits on AI for a very long time. IMHO

    Although I am not a programmer, this sounds 100% correct to me.

    • #25
  26. genferei Member
    genferei
    @genferei

    Percival (View Comment):

    Question: Is the AI going to keep the information it gleans while analyzing your systems and data to itself, or will it apply any of that knowledge to future products it produces for others, including your competitors? If someone asks for your customer list, will it hand the list over?

    If you’re paying, the AI (As A Service) vendors promise they won’t. It’s up to you what you think of those promises. If it’s Microsoft, and you’re already all-in on Azure and MS-hosted SharePoint and whatever fool name they give to the Office suite this week, what’s the difference? If it’s some CCP-funded Chinese service, then I would absolutely trust them to be snarfing up everything about you for the purpose of overthrowing western civilization.

    • #26
  27. Yarob Coolidge
    Yarob
    @Yarob

    Interesting question.

    As you saw from your DeepSeek video, AI-assisted coding is already a thing. See, for example, IBM’s watsonx Code Assistant (“Code smarter, not harder”). At the moment it’s positioned as a tool to “augment” developer skill sets, but we can’t be far off it replacing those skill sets completely.

    The major change, I believe, will not be code-related and technical, it will be allowing end-users to create applications directly without needing the assistance of IT specialists like systems analysts, database designers, and programmers. We’ve had code generators and automated software testing for many years, but the revolution will be in building systems without having to engage in the traditional lengthy and expensive processes that now have to take place before a line of code is written. Say goodbye to requirements definition and all that it entails—structure charts, functional decomposition of processes, data flow diagrams, data dictionaries, etc.—and replace them with accountants and hiring managers speaking into a microphone and clicking icons on a screen.

    The future? AI-generated software running on quantum computers. Last year, Google’s quantum Willow chip ran a benchmark in under five minutes that would have taken a contemporary supercomputer 10 septillion years (1 followed by 25 zeros) to run. Commercial quantum computing is not around the corner, but it is coming.

    And no, none of this will be free, because the value of software will increase, not diminish.

    • #27
  28. randallg Member
    randallg
    @randallg

    David Foster (View Comment):

    OTOH, suppose it were easy to use AI prompts to improve search for Ricochet posts relevant to a particular topic…that would add value to the customer base that already exists.

    Good idea. Ever notice that most sites’ search functions are completely useless.
    Google almost always works better.

    • #28
  29. kedavis Coolidge
    kedavis
    @kedavis

    Brady | @HerrForce1 (View Comment):

    iWe (View Comment):

    She (View Comment):

    So, my takeaway from this is:

    “Don’t Learn to Code.”

    Is this right?

    Seems like it.

    “I just want to say one word to you. Just one word. Plastics.”

    Sam Wainright to George Bailey, right?

    • #29
  30. Henry Racette Member
    Henry Racette
    @HenryRacette

    kedavis (View Comment):

    Brady | @ HerrForce1 (View Comment):

    iWe (View Comment):

    She (View Comment):

    So, my takeaway from this is:

    “Don’t Learn to Code.”

    Is this right?

    Seems like it.

    “I just want to say one word to you. Just one word. Plastics.”

    Sam Wainright to George Bailey, right?

    Not quite. Someone I don’t remember to a young Dustin Hoffmann.

    • #30
Become a member to join the conversation. Or sign in if you're already a member.