Predictions
Lets make some assumptions about the future:
- Agent’s contexts will be irrelevant. No “dumb valley”. No 1M tokens max.
- Token generation will be virtually free (100€/month/dev for unlimited tokens).
- Token generation will be virtually instant. Time for mid-effort tasks: 1 second max. Time for high-effort task: 20-30 seconds max.
- Current bottlenecks will be solved: code reviews, agentic code non-maintainability, security issues, deployment, error detection/automatic fixing, …
The following are my predictions. This will be a live post that will update as I have more ideas. In 2 year’s time (2028-10-26), I’ll check which ones are correct.
Companies will be forced to offer APIs
People will prefer companies and products that are accesible thru APIs, to build their own tools, or use tools that use those APIs.
Examples:
- Flights: I don’t want to say “I don’t want extra luggage” 6 times
- Flights: I want the check-in to be done automatically (why is not a thing already? Oh right. They try to sell extra luggage 2 more times )
- Flights/Public transport: I want to make sure that the late flight has a public transport connection to my home at that hour
- Bank: I want the AI to import and classify every expense in GnuCash
- Banks: I want a scanner that tells me which bank is getting me a better remuneration for a savings account
- Hotel Bookings: AI to find great deals with my already-defined criteria, while on motorbike trip. Offers to be presented at the optimal time: delaying the booking + monitoring that the price do not increase
- Taxes/Administration: Vehicle inspection reservation done ahead of time. Review that taxes are linked to the correct bank account. Monitor balance on bank account to make sure they will be enough funds.
- Gym App: Gym calendar should be in sync with my main calendar.
- Gym App: Adding an event on top of my gym session should reschedule it if possible. If not, tell me to reconsider if missing gym for the added activity is worth it.
- Gym App: Trainer that I don’t get along with? Reschedule.
Repositories of Ideas/Features for apps
Ideas of features will be gold. The implementation is cheap. There will be repositories of features, describing in markdown what they do. Then you can use those ideas to tailor them to your needs. The perfect app is the combination of those features that matches what you want. We’ll be able to “vote” on top ideas, like we do today with stars in Github. A free idea marketplace.
I want to build my own calendar app. What should it have aside from the basic calendar functionality? Let me browse the “Calendar ideas” repository, which is a just a collection of markdowns explaining each feature other calendars have. Let me search for the most popular ones.
I love the “weekly close-friend call scheduler”. It checks my calls/messages and finds a person who I appreciate that I haven’t talked with in a while, scheduling a call on my free weekly slot. Perfect.
Software as a Service is dead
Tech-savvy people will create their custom apps.
There will be app-creation AI platforms for the rest of people.
It will become so easy an inexpensive to build apps, that everyone will do so. Tailor made. New features added being dictated to your phone, and implemented and deployed async.
Subscription models are dead
Users never wanted to recurrently pay for software/services. Less so when they are riddled with bugs and evolve in ways you don’t agree with.
Software will be fast and simpler
Apps will be tailored made. No fluff. No tracking stuff. No random updates for features you don’t need. No, report if there is a bug. No asking for more permissions that you think are needed.
AI will use programming languages that compile to single binaries, statically linked, that weight a MiB.
No more “note apps” that are 700MiB because it packs an entire web browser inside. That was crazy and lazy to begin with.
Companies will be required to have high SLAs
Their customers will use their APIs more and more, and require them to have good uptimes. If your provider is failing you constantly, and switching is easy, you just do so.