On the future of (software) apps

coming off the news of jonny ive and his ai design firm being acquired/bought by OpenAI, it brings to the forefront that the physical computing devices we use are going to change very quickly and look very different. yes we knew this, but now it feels more real with plans to release their first device next year.

although sama rudely gave us nothing on product specifics in the official launch video after making the claim that it was “single coolest piece of technology the world will have ever seen”, in another conversation he stated it will have no screen and the team hopes to “wean people off of screens”. amazing. this is what was so great about Deep Research when it first came out — less fumbling with internet search meant less time on screens and more focused action.

now no screen altogether? no more apps (as we know an “app” to be today)?

think about the transit app i just made. instead of having to look at and interact with a screen to get the arrival times of upcoming trains, you could speak? touch? move? point? with a device that’ll magically? give you the answer. see those question marks are the main thing preventing us normal folk from understanding how this new wave of technology will look like in our daily lives. something some people are calling NLX (natural language interface) because of how the generative ai models revolve around natural language as both inputs and outputs. a NLX requires new UI elements (prompts, plans, chains of thought become new elements just like how the touch screen birthed the “pinch-to-zoom” or completely gesture-only interfaces requiring no buttons in place of swipes) but also new thinking as to how to deal with both parts of natural language — the part we see (text) and the part we hear/speak (voice) (note: yes, the blind/deaf also require touch and that’s another problem that needs to be considered because technology should be accessible to all). this new interface that centers around language has different structures and environments than what we are used to with a GUI (and seemingly more variable). GUIs are structured, fixed, and predictable (your phone screen is your phone screen) in many ways more-so than what the structure would be for delivering the same information when i’m walking down the sidewalk needing to know when the next train arrival time is. the elements in that environment are invisible. so then it becomes screen optional? or it’s a new form of screen (AR) like meta raybands or the resurfaced google glass. or maybe it’s a voice agent, but few are willing to forego privacy to talk to themselves on a busy train. this makes the case for NLX to be an add-on to the existing GUI we are familiar with today. most likely, NLX will start that way despite humans insane ability to adapt to technology quickly (remember doing internet-things before chatGPT?). it’s useful to consider the extremes (a no-screen future), but they won’t be eliminated entirely because some experiences undoubtedly benefit from a screen — like information-dense (maps) long-form content (videos). it’s interesting that I just presented the idea that NLX would be used in situations that require short-from content with short-term memory needs because that’s exactly where the majority of our current screen time is being spent — on the fleeting tik-toks and tweets. yet, one of the main points with a GUI is that it requires a lot of unnecessary effort to find a single piece of fleeting information you need (take out your phone?, open an app?, navigate the app?, search for it?). another angle to consider is how use cases continue to remain for old tech (way of doing things) when new tech comes out (e.g. complete control of on-prem vs the cloud).

one thing is for sure: software “apps” (applications that help you do things) will look very different very soon

this isn’t even considering the new way that knowledge work will be conducted when compute and intelligence is in seemingly infinite supply and the limiting factor becomes your ability to orchestrate the intelligent resources that you have. when to course correct? this is a new kind of leadership skill.

in this environment, your judgement is your differentiator. when everyone has access to the same information you do, what makes your decision different? access to the same information already exists to a degree today (just like how it was with libraries), but AI tools make it exponentially more true. judgement = the unique sense-making behind (1) what information you used and (2) how you used that information to get to your answer/decision. reasoning and intuition are unpredictable, independent processes that make humans human and inform our judgements. news flash: AI’s can reason so therefore they can also form their own judgements. if you train something to excel at and learn how to reason, it will therefore become an independent thinker itself (congrats, you now have to compete against a digital workforce, too! oh, and AI becoming self-aware? yeah, that’s inevitable because self-awareness is beneficial to any reasoning-optimizing machine). not all judgement is the same — there’s more but remember, an AI won’t have human judgement in the same way that you won’t have my judgement. all this to say that your uniqueness matters more now than ever.

quick side note: this is also connected to the main thesis of my presentation on the future of Product Management as a craft. i was bullish on PM becoming more important in the AI-world because of their responsibility in determining what is built and why. combine that with the person responsible for best-understanding the most important stakeholder to the business (the customer), you get a very important person.

the uniqueness behind the product judgements you make will, again, become even more of a differentiator when the “building things” part gets easier. of the three build stages (knowing what to build, how to build it, and building it), the “building it” part is going to be solved fairly soon. most of software by the end of the year (“most of” because although writing the code may be solved, there’s still all the work outside of the IDE to connect different services, platforms, and tools that need to be done). robotics shortly after.

what am i going to do about this?

1. spend less time fighting with app ‘building’ when it’s a technical code issue
2. spend more time understanding the what and why behind products and our own judgements
3. continue to learn about and use AI, consider natural language as an interface, put a premium on human experiences, and take my health seriously (AI will never be able to run a sub-3 hour marathon for me!)

figuring out what needs to be built, why, and how will forever remain a mystery.

edit: one area i didn’t discuss above, but is likely to have a big impact on future of software apps, is the browser/web. as cloud and other technologies make the browser feel native, fast, and reliable like they do in native apps, more of the work we do (although so much of it already is) is likely to become browser-based or internet-native. it’s a matter of collaboration, accessibility, and transparency — values customers increasingly care about when options are plentiful like they are today. in addition, experiences on the web are becoming on-par with their native app counterparts (one slight lag: mobile — mostly due to apple’s grip on safari web handling) AND AI-first, generative apps work better on the web because because the web gives you more flexibility to dynamically load models, stream content, and plug into back-end APIs without rigid app store constraints or client-side processing limitations.