Do we need to "develop" apps anymore? Can consumers develop their own apps yet?
The recent years of post-pandemic employment slump has been capturing headlines as tech leaders across the board have cut their staffing dramatically to achieve more efficiency in their businesses. Seeing hundreds of thousands of layoffs is a grim reminder of the boom/bust cycle that hangs over high-beta companies that leverage investment to scale up for unforeseeable future market demand.
An interesting twist in the headlines over the last year is more of these companies seeking to “flatten their organization structures.” Often these stories have focused on machine learning automation of routine tasks and in some cases delegating creative engineering or product design work to automated tools.
Legislators, who earn their salaries from taxing humans, wouldn’t want more mechanized labor in the work force because typically machines don’t pay taxes. But seeking to legislate staff retention is a blunt instrument because businesses exist to render services or produce tools and utilities, not to instrument human job/salary payouts. Requiring companies to maintain inefficient labor levels for community’s sake is outside the structure of how businesses operate for the benefit of their customers.
On the inverse side, technically inclined creators and entrepreneurs are enormously excited about the emergence of technical tools that allow them to massively scale up the impact of their own talents and productivity to bring new services and products to market, broadening the base of who in our community can create whatever they envision. More people being able to launch companies and produce more beneficial services, content and products of their own is an enormously beneficial market development. The reduction of funding waste for a new single-founder company development is a tremendous boon to our society and the democratization of access to the market. If a person can launch their company without going considerably into debt, it reduces the cost of innovation and invention leading to a more dynamic and diverse economy.
Being an industry veteran since 1995, I can see booms and busts of investment t cycles as the dialectic process of integrating innovation into society advances. I see both sides of the dialectic process as a kind of random walk with a feedback mechanism of the market (us) conferring the growth potential to the preferred end product or service. So on the macro scale, while I appreciate the frustrations and delights of the mechanisms of innovation personally, I focus on the systems and process broadly to determine the end outcome’s value.
My colleague Ben Morrow recently shared an article from Critini Research pondering the future impact of AI systems on software vendors. In it, the authors describe legacy SaaS system vendors as “friction with a friendly face”. Middle layers of the software ecosystem have profited by making complex systems and processes more fluid with vertically specific solutions that are now open to replacement by lower cost DIY tools that obviate those vendors.
The phrase du jour in the tech industry for this is “vibe-coding” whereby a person leverages open software or framework tools to create from scratch what they would previously have to outsource to a competent vendor. As a web developer who has built dozens of sites from the early days using WYSIWYG editing tools like Adobe DreamWeaver/GoLive to web editors like SquareSpace and Wix today, I am familiar with the hassles and delights of creating unique web experiences for my audiences of readers, listeners and watchers of my content.
I tried the new vibe-coding web design platforms such as Replit, Lovable and Caffeine with my colleagues. I am delighted with the way that these tools democratize access to web design to anyone with a bit of time, patience and passion. Their cost is 3-4 times more than conventional web hosting. But the ease of access means that more websites and creative work can find their way to audiences worldwide. A broader distribution of web authorship is an enormously positive outcome in my mind.
The biggest take away for me from this year’s Google I/O developer conference was a product idea that isn’t yet public. It’s a riff on the vibe-coding trend, but this one focuses on Android app creation, not just web pages. This July Google will unveil their toolkit for vibe-coding Android APK apps to list directly on the Play store. Register here if you’re curious: ai.studio
Vibe-coding apps from open source GitHub frameworks (to build native applications on Android or iOS devices) is still somewhat complex for techies like me. But Google plans to launch an app on mobile devices that authors other apps for the same device. Vibe approach to this is to type or say “I need my new app to do these few functions” and let the AI Studio draw the necessary component parts together for you without you having to to pull code together yourself. You can go from use case to functional app, then list the app on the Play app store after review and sharing to a few peers.
In reaction to this development, my colleague Carlos Freitas said, "Why do you even need to have an app? You can just have your AI/agent, which knows where to find the information and will have the negotiation capability as well!"
Though Google's AI Studio is not yet in the hands of consumers and developers, it provokes a lot of consideration for people like me who have studied phones and app frameworks a good deal. On one level, it reminds me of the WYSIWYG editors of the 1990s which made the personal webpage a broad market phenomenon. On another level it appears to invoke the elimination of all future “native app” development. To break this concept down I should share some personal history that makes me think this way.
My company was a not-for-profit open source developer community. They wanted to provide an open source framework that would allow other companies to build web-phones. So they launched one called FirefoxOS which other companies could freely replicate under their own brands. In the process of web-enabling every component of a modern phone, we created a JavaScript operating system and enabled ways for websites served on the phone to dialog with the user and query the device for location/camera/battery/proximity-to-server and general things Android and iOS offer with hidden code. Part of this process was the enablement of sockets called “intents” expressed by the user or site to create the full experience through dialog over the web session. There were share-intents, save-intents and consent dialogs that the developer or site host could use. This came to be known as “progressive web apps” that allowed webpages to save content offline in low connectivity contexts. But generally the upshot of this phase of innovation was that we were able to get away from always pushing software onto user’s phones until they demanded the requisite service or capability from a web host. This is what makes me see around the corner of what is next for agentic phone capabilities.
In the I/O demo of AI Studio, a user speaks or texts a prompt to the native app and the frameworks to achieve the intent construct the scope of the app from capabilities resident on the device or accessible in the cloud. In theory the new app assembles on the fly to suit the desired scope of functions. Flipped around, subtracting the “developer” any user could assemble the fulfillment of their intents without having to bounce off to the Play store to buy an app from another developer. They develop for themselves without the need for an intermediate facilitator or gatekeeper.
If Android or iOS devices can service a user need directly without the need for a cumbersome purchase funnel through an app marketplace, shouldn’t they? Thousands of app developers may groan at this concept. “The OS is going to compete with me on the device I am building for? Unfair!” (This is a lot of the thinking behind the “Sherlocking” concept in iOS developer circles. How long can you participate in an app market before the OS maker integrates your feature directly, obviating your work?) Seen in this context, app and extension marketplaces are like Petri dishes where we web developers can build and test until our services are crucial enough that they demand being surfaced to everybody without hasslesome intermediary steps.
Discussing this concept with my developer friend, Jonathan Prentice, he pointed out that brands in an app marketplace are markers of trust. If a person trusts a brand, they may give that brand access to their credit card to facilitate future purchases. (Food delivery for instance.) Lesser known brands have a friction point of familiarity that would lead to more reticence in giving personal payment access. That can be assuaged by OS-level purchasing tools such as ApplePay or GoogleWallet flattening the competitive landscape across brands. "Intent is more powerful than brand," Jonathan asserts. Why develop an app at all if you can have direct conduit between customer and point of sale merchant, he illustrates this example by sharing the Apple Messaging direct contacts capability in Messages app. By having capability to express intents directly to merchants, there is no need for any app intermediary.
In the era of agentic AI, where small vertical representative agent applications on a web server can interface directly with user concierge apps like Siri, OKGoogle or Alexa we don’t actually need to stare at a mobile screen to make our intents known. In such a future world our eyes and fingers can be freed to concentrate on more human tasks than managing a cumbersome mobile phone interface.
Developers in the agentic context can focus on publicizing what can be done on the web with their model context protocol or Agent-to-Agent protocols rather than pushing pixels to win clicks on tiny surfaces one customer at a time.
When Amazon first launched their Alexa tower smart microphone, they asked developers to push “skills” into its memory that would make the tower actually be able to respond to questions their users might ask. That was a decade ago, and required a lot of user and developer fumbling to enable one by one talent extensions to the core device. Now we can use model context protocol (MCP) to advertise to the web of automated agents whatever our server’s basic skills and competencies are. Agentic layers can negotiate a lot of the decision layers without turning our humans into pixel prioritizers.
So while July’s launch of AI Studio is temporarily attempting to grease the skids of Android app development, ultimately what we are working toward is a web where voice to voice without bottlenecks arrives for us as people. There will be a reduction in conventional developer work as devices we rely on become more capable natively of solving needs and demands. But the more human user interface of voice and lowered friction in our daily lives is a good goal to work toward.
Comments
Post a Comment