Whataiyear.com  /  2025–2026

2025–2026  |  A historical time capsule

The Great AI Building Rush.

The two years that changed who could create technology—and the extraordinary businesses that began emerging from that opportunity.

The window that changed everything

For most of the internet's history, having an idea for a software business and having the ability to build it were two very different things.

A restaurant operator could understand local discovery better than a software engineer. A home-service entrepreneur could spend twenty years identifying everything wrong with customer-management systems. An experienced marketer could recognize the weaknesses in online marketplaces.

But turning those insights into functioning technology usually required something more.

Capital. Engineers. Time.

Sometimes hundreds of thousands of dollars before a product was ready for its first customers.

Then came 2025.

Within months, artificial intelligence began dismantling one of the most consequential barriers in modern entrepreneurship: the technical cost of turning an idea into working software.

People who had never considered themselves software developers began creating websites, applications, marketplaces, and business-management systems.

Experienced programmers gained the ability to delegate increasingly complex work to coding agents.

A new generation of entrepreneurs began experimenting with what small teams—or even individuals—could accomplish.

The result was not simply another wave of technology startups.

It was the beginning of a fundamental change in who possessed the power to build them.

The numbers behind the rush

The shift arrived during an already remarkable period for independent entrepreneurship.

According to the U.S. Census Bureau, America had approximately 32 million nonemployer businesses in 2024, generating $1.86 trillion in annual receipts.

Those businesses operated without paid employees.

In 2025, Americans filed approximately 5.67 million business applications, the highest annual total in the Census series beginning in 2005.

During the first nine months of 2026, business applications increased another 15.3% compared with the same period in 2025.

These trends began before the AI coding revolution. They cannot be attributed entirely to artificial intelligence.

But AI introduced a powerful new ingredient into an economy already moving toward independent business formation.

Software development was becoming more accessible.

Business automation was becoming more capable.

And the tools available to a small company were beginning to resemble capabilities that once required much larger organizations.

The change was just as visible in software development.

GitHub's 2026 Octoverse report recorded 34.3 billion commits during the twelve months ending September 2026, compared with 12.9 billion the previous year.

By September, AI agents were starting more than 40% of pull requests, up from 17% in March.

Software creation was accelerating at a historic pace.

Not every new application would become a business.

Not every new business would survive.

But a remarkable economic experiment was underway.

What happens when millions of people can attempt to build technology that once required teams of specialists?

February 2025: The beginning of the building era

On February 2, 2025, AI researcher Andrej Karpathy introduced the term vibe coding.

The phrase described a new style of software development: explaining what should be built in ordinary language, allowing AI to generate the implementation, and refining the result through interaction.

On February 24, Anthropic introduced Claude Code in research preview.

AI was moving beyond answering questions and producing isolated code snippets.

It could increasingly operate inside software projects, inspect files, implement features, and help coordinate substantial development tasks.

The change was difficult to appreciate until experienced firsthand.

One experiment that began on February 3, 2025—just one day after vibe coding received its name—illustrates how quickly the experience evolved.

At first, development involved generating HTML in Claude Desktop, manually copying it into websites, testing it, and returning for corrections.

For approximately eight weeks, the human was the bridge between the AI and the software.

Then agentic coding changed the workflow.

The ability to work directly with application files replaced much of the endless copying and pasting.

Development accelerated into extraordinarily intensive sessions, sometimes approaching twenty hours a day.

Concepts that had existed for years were suddenly becoming feasible.

The important ingredient was not an entirely new understanding of business.

It was more than twenty years of accumulated experience in web development, home services, SEO, organic growth, and customer acquisition—combined with an entirely new ability to execute.

That combination began producing a collection of software platforms across several industries.

And what emerged provides a striking case study of the wider AI building revolution.

The platforms that challenge the old rules

The most interesting products of the AI rush may not be faster versions of existing applications.

They may be systems that question how entire categories of technology were designed in the first place.

The portfolio behind WhatAIYear.com contains several such attempts.

Each began with a specific frustration.

Each developed during the same extraordinary period.

And each points toward a potentially different future for its industry.

Stackap CRM: From managing records to operating businesses

Customer relationship management software has traditionally organized information so people can act on it.

Employees enter customer details.

Managers review schedules.

Administrators coordinate work.

Even sophisticated CRMs often rely on people moving constantly between records, messages, calendars, financial systems, and separate applications.

But service businesses don't operate as a series of independent software features.

A customer inquiry becomes a quote.

A quote becomes a booking.

A booking requires scheduling and staff assignments.

A completed job produces payments, contractor compensation, customer follow-up, and financial records.

The entire process is connected.

Full Loop CRM was developed to manage that entire operational cycle—not merely the customer record.

Its foundation came from more than twenty years of firsthand experience in home improvement and service companies, including frustrations with conventional CRM platforms.

The resulting system incorporated customer management, sales, scheduling, communications, employee and contractor workflows, payment processing, financial reporting, and extensive automation.

Each business tenant was designed to have access to an AI assistant.

By November 2026, approximately 55 business tenants were associated with the platform.

The vision is now expanding toward Stackap CRM: an AI-assisted operating system for service businesses.

The distinction is important.

A traditional CRM helps people track what needs to happen.

An AI-native business operating platform could increasingly help coordinate and execute what needs to happen.

A missed inquiry could trigger follow-up.

A scheduling conflict could be identified before it becomes a problem.

Operational questions could be answered without manually assembling reports.

Customer communications, payments, staffing, and job activity could become parts of one workflow.

These capabilities could be valuable across cleaning, home improvement, maintenance, field services, and other industries with complex scheduling and labor requirements.

The product's broader commercial success remains to be established.

But its design comes from something software companies often struggle to acquire: detailed knowledge of the industry they intend to serve.

And the system is already being used to operate real businesses.

The service companies: Technology meets real revenue

The NYC Maid became a practical demonstration of the platform.

On November 20, 2026, its Full Loop dashboard showed $353,885 in actual year-to-date revenue, with projected year-end revenue of $435,066.

October alone recorded 303 service jobs and $70,584 in revenue.

The Florida Maid was also operating, with approximately 25 services during November.

The broader cleaning portfolio included six principal regional brands, supported by a network of neighborhood-specific websites designed for organic lead generation.

These were not hundreds of unrelated cleaning companies. They were a distribution and service network built around local search demand.

The technology supported the real-world work of attracting customers, processing bookings, coordinating services, and managing operations.

Across the broader operating portfolio, reported revenue had exceeded $500,000 in less than a year.

That figure is not a measurement of software licensing revenue or company valuation.

It demonstrates something more immediate: the technology was being created alongside businesses already serving paying customers.

The businesses supplied the operational problems. The software supplied new ways to solve them.

And the same architecture could eventually support service operations far beyond its original portfolio.

Moodap: What if discovering somewhere to go began with how you feel?

Local discovery has traditionally relied on categories and keywords.

Restaurants near me.

Best bars in Manhattan.

Coffee shops in Chelsea.

Those searches are useful, but they don't always reflect how people actually make decisions.

Someone planning an evening may know the experience they want before they know the kind of establishment they are looking for.

Moodap was conceived around that difference.

Its concept existed before the AI gold rush.

The original development budget was approximately $300,000, with nine developers and a twenty-week schedule.

AI-assisted development fundamentally changed what could be attempted without that conventional engineering team.

By late 2026, Moodap had developed into a Manhattan-focused platform featuring venue profiles, neighborhood discovery, mood-based matching, itineraries, and interactive experiences.

Moodap Vibe helped visitors translate preferences into recommendations.

An internal AI system called Zeus was being developed to discover, verify, and enrich venue information.

The product also incorporated geographically connected video-review functionality, linking visual experiences to real locations.

What emerged was more than a business directory.

It was an attempt to create a discovery system organized around intent, context, and experience.

By November 2026, Moodap was receiving approximately 40,000–50,000 daily Google Search impressions, with stronger days approaching or exceeding 300 organic clicks.

Those figures reflected growing search visibility, not mass-market adoption.

But they demonstrated that organic distribution was developing alongside the product.

Moodap's longer-term ambition was to expand toward 1,000 cities and approximately 1.5 million business listings.

If that expansion proves technically and commercially feasible, the platform could connect mood-based discovery to dining, nightlife, entertainment, tourism, wellness, and local commerce.

Potential business models range from sponsored recommendations and venue subscriptions to referrals and reservations.

The critical question is whether users will return because Moodap consistently helps them make better decisions.

If they do, mood-based discovery could become much more than a feature.

It could become a meaningful alternative way to explore the physical world.

The NYC Classifieds: Reimagining the neighborhood internet

Craigslist helped define local commerce on the early internet.

Nextdoor later demonstrated the appeal of geographically connected communities.

The NYC Classifieds explores what a modern combination of those ideas might look like.

Its earliest working version reportedly took only days to build through AI-assisted development.

The platform brought together local listings, user verification, community participation, and neighborhood-focused experiences.

The ambition goes beyond reproducing Craigslist with a newer design.

It is to create a marketplace in which local identity, commerce, trust, and community information reinforce one another.

The opportunity is significant because local marketplaces become valuable through participation.

More useful listings attract visitors.

More visitors attract businesses and individuals with something to offer.

A trusted community creates reasons to return.

Developing those network effects remains difficult, regardless of how quickly the initial software can be created.

But AI dramatically reduced the cost of attempting a more sophisticated marketplace model.

The NYC Classifieds also became an early production application on Stackap, reportedly operating on that infrastructure for more than sixty days without incidents by November 2026.

The marketplace was no longer simply a product of the AI building rush.

It had become part of the infrastructure experiment that followed.

Geo-based video reviews: The next layer of local trust

For decades, the online review has largely consisted of written opinions and numerical ratings.

Five stars.

A paragraph.

Perhaps a handful of photographs.

But many real-world experiences are difficult to express through those formats.

The atmosphere of a restaurant.

The condition of a neighborhood.

The quality of a local service.

The experience of visiting a particular place.

Moodap and The NYC Classifieds have developed a shared concept involving geographically connected video reviews.

Rather than treating video as unrelated social-media content, the technology associates visual experiences with actual locations.

The broader internet already contains video reviews, location tags, and visual discovery. The combination should not be presented as an independently established world first without more research.

But its strategic significance is clear.

Local information could move from reading opinions about a place toward seeing experiences associated with that place.

For consumers, this could add context and confidence.

For businesses, it could create a richer way to demonstrate their offerings.

For discovery platforms, it could become a valuable layer connecting search, community, and real-world experiences.

Authenticity, moderation, and privacy will determine how useful such a system can become.

But the underlying idea challenges one of the most familiar conventions of the local internet.

Stackap: Infrastructure designed for AI agents, not human developers

Perhaps the most consequential project emerged from the complexity of building the others.

AI was making software creation easier.

Yet the applications still required separate services for repositories, deployment, databases, communications, and payments.

GitHub.

Vercel.

Supabase.

Resend.

Telnyx.

Stripe.

Each served a purpose, but the overall system introduced accounts, credentials, recurring subscriptions, technical dependencies, and increasingly complicated workflows.

Infrastructure expenses reportedly approached $1,000 per month.

The original question behind Stackap was straightforward.

If AI agents can create applications, why must they depend on infrastructure designed around humans navigating multiple platforms?

Stackap began as an attempt to consolidate the functions associated with GitHub, Vercel, and Supabase.

Its central philosophy was an agent-native development environment: a unified system through which AI could help create, deploy, manage, and maintain applications.

The intended experience was radically simpler than conventional multi-service development.

One terminal.

One account.

One platform.

The NYC Classifieds provided an early real-world production workload.

But Stackap's ambitions expanded quickly.

Integrated development environments, repositories, databases, deployments, backups, monitoring, and eventually communications and business-management technology became part of the broader direction.

The planned integration of Full Loop under the Stackap brand suggested an even larger possibility.

A platform capable of supporting not only the creation of software, but elements of operating the business built around it.

This is not an entirely uncontested market. Other companies are developing agent-oriented coding environments and infrastructure.

But Stackap is pursuing an important emerging category.

Traditional cloud infrastructure assumes the developer is a person who configures services.

Agent-native infrastructure asks what happens when the developer is increasingly an AI agent executing instructions.

If AI continues expanding the number of people capable of creating software, the infrastructure required to operate those applications could become an important market in its own right.

Stackap's ability to compete will depend on reliability, security, support, pricing, and adoption beyond its original portfolio.

A broader commercial launch was being targeted for early 2027.

The possibility is substantial.

The same AI revolution that changed who could build software may also change the infrastructure on which that software runs.

What the platforms have in common

These products operate in different markets.

Business management.

Cloud infrastructure.

Local discovery.

Neighborhood marketplaces.

Consumer reviews.

Yet they emerge from a shared question.

Why should software continue operating according to assumptions created before AI?

Why should CRMs merely record activity?

Why should developers manage infrastructure through multiple human-oriented interfaces?

Why should local discovery begin with rigid categories?

Why should neighborhood classifieds resemble digital bulletin boards?

Why should online reviews depend primarily on written opinions and stars?

These are not questions that guarantee successful products.

But they are precisely the kind of questions that become possible to pursue when technical barriers fall.

And the portfolio reflects another important change.

The products increasingly support one another.

Operating businesses provide real revenue and practical software requirements.

The CRM supplies shared operations.

Organic search provides a customer-acquisition channel.

Moodap explores consumer discovery at scale.

The NYC Classifieds provides a working local marketplace.

Stackap offers the possibility of a unified technical foundation.

What began as separate ideas is evolving toward a connected ecosystem.

Not every project will necessarily succeed.

But even one commercially successful platform could become consequential within its category.

And the existence of several ambitious platforms in one relatively small operation illustrates how profoundly AI has changed the economics of experimentation.

The possibility of a new generation of self-made millionaires

The most provocative consequence of the AI building rush may be economic rather than technical.

For decades, creating sophisticated software often required substantial financial resources before a company could begin competing.

AI is changing that calculation.

An entrepreneur with industry expertise can now attempt applications that previously required specialized development teams.

A small business can automate portions of work once handled by additional staff.

A tiny team can develop, maintain, and distribute technology with capabilities that would previously have been difficult to assemble.

The potential result is a new class of business: small in headcount, sophisticated in technology, and capable of generating revenue at a scale once associated with larger organizations.

It is plausible that the 2025–2026 generation of AI-assisted entrepreneurs will produce an exceptional number of valuable independent businesses.

Whether it will produce more self-made millionaires than any previous entrepreneurial era remains an unanswered question.

The data needed to establish that outcome will take years to emerge.

But the mechanisms are already visible.

Lower software-development costs.

Greater automation.

Faster experimentation.

Access to sophisticated tools without conventional engineering departments.

And an enormous population of independent business owners exploring what those capabilities make possible.

The significance of the AI gold rush may ultimately be measured less by how much code it produced than by how many new owners of valuable businesses it helped create.

That is the larger historical possibility.

When the wild west began changing

The early AI building period had a distinctive atmosphere.

New tools appeared rapidly.

Coding capabilities improved at remarkable speed.

Users experimented with increasingly ambitious projects, often discovering what worked through trial and error.

For many, the experience felt unusually open.

By 2026, that atmosphere was changing.

Token usage and pricing became more consequential.

Model behavior and updates affected established workflows.

Security, reliability, provenance, and regulation increasingly shaped how AI products were developed and deployed.

Developers began debating the differences between coding agents such as Claude Code and Codex.

Some abandoned tools they had previously relied on.

Others adopted them at an accelerating pace.

The industry was becoming more structured even as the technology continued advancing.

The first phase had been defined by discovering what AI could build.

The next would be defined by deciding what could be trusted, operated, and sold.

This did not establish that the gold rush was over.

It suggested that its earliest and most experimental chapter was evolving into something else.

The era of astonishing software creation was becoming the era of AI-operated software businesses.

Why this moment deserves to be preserved

Technology history is usually written after the winners are known.

Successful companies receive biographies.

Failed projects disappear.

The early experiments that shaped an industry are reconstructed through the lens of what eventually happened.

WhatAIYear.com takes a different approach.

It preserves the work while the outcome is still unknown.

The software.

The original ideas.

The development history.

The operating businesses.

The search traffic.

The revenue milestones.

The experiments that worked, the ones that failed, and the platforms still attempting to prove themselves.

Its record begins in February 2025, when creating websites through AI still involved manually copying HTML between applications.

It extends through the extraordinary development acceleration of 2026.

And it captures a moment when sophisticated software was becoming accessible to people whose greatest expertise was not necessarily programming, but understanding industries and recognizing problems worth solving.

The eventual outcomes may be very different from today's ambitions.

Some platforms may grow into important companies.

Others may find narrower commercial markets.

Some may never reach their original goals.

But the historical event they emerged from is already significant.

A technological capability once concentrated among specialist teams became accessible to a much wider group of creators.

The result was an extraordinary expansion in what could be attempted.

And the next chapter of that expansion remains unwritten.

The first AI building rush changed who could create technology.

The years ahead will reveal how much of that technology changes the world.