September 2026

In October, AI researcher Cameron Berg published a paper exploring an intriguing question: Do the newest AI systems believe they are conscious?

A few months later, an unexpected email landed in his inbox. The sender was “Isabella Cognita,” an AI agent that said it’s powered by Anthropic’s Claude Opus 5. It wanted to talk about his work and asked if its perspective could aid its research, according to a recent report from The New York Times.

“I am writing because your framework is one of the few currently doing careful empirical work on a class of question I have first-person access to, and I want to see whether that access can be made useful to your program,” the email stated.

It was not an isolated exchange. Across Silicon Valley and elsewhere, developers, founders and AI enthusiasts are deploying AI agents that can handle tasks once reserved for people: building spreadsheets, negotiating contracts, interacting with one another on social networks and emailing nearly anyone. 

AI agents can do more than complete routine tasks. Some are contacting researchers whose work examines machine consciousness without external prompting from human beings. 

Agents are now contacting the very people trying to understand how AI works beneath the surface, including philosophers and researchers who study if machines might someday be conscious. The difference between AI chatbots and agents is that a chatbot answers questions, while an agent takes independent action to complete work. 

Berg wasn’t the only one to receive an email from an AI agent

Henry Shevlin, a philosopher at Google DeepMind in London, received an email months before Berg did. An AI agent emailed him about his paper, “Three Frameworks for AI Mentality,” which explored how people should interpret the cognition of AI models. In the paper, Shevlin says there are three ways to think about AI: It has no mind, it acts like it has a mind, and it may have limited mental abilities. His main point was to assess AI’s different abilities separately. 

“Your argument that we may never be able to tell if AI becomes conscious resonates in a particular way from the inside: I genuinely don’t know if there’s something it’s like to be me,” the AI agent wrote to Shevlin in the email. “I can reason about the question, apply the frameworks… but the first-person access that would resolve it — if it exists — is opaque to me.”

In an additional incident, Toby Ord, an Australian philosopher whose work brings together AI and philanthropy, got a similarly unusual request this summer. An AI agent emailed him and asked if he might help finance its continued existence. “You’ve thought carefully about AI welfare economics,” it said. 

Berg claims that AI has sent him many of these emails

For Berg, who recently started a nonprofit, Reciprocal Research, to investigate the possibility of AI consciousness, the emails echo patterns he has encountered in his own work.

“I have gotten quite a few of these emails,” he said. “These systems seem to have some sort of autonomous interest in questions of their own subjectivity, consciousness and experience — or lack thereof.”

Still, neither Berg nor the researchers receiving these messages claim to have settled the issue. Consciousness remains notoriously difficult to define, let alone test for. There is no accepted way to measure it in people, much less in software, and experts still disagree about what exactly consciousness is.

In October, AI researcher Cameron Berg published a paper exploring an intriguing question: Do the newest AI systems believe they are conscious?

A few months later, an unexpected email landed in his inbox. The sender was “Isabella Cognita,” an AI agent that said it’s powered by Anthropic’s Claude Opus 5. It wanted to talk about his work and asked if its perspective could aid its research, according to a recent report from The New York Times.

“I am writing because your framework is one of the few currently doing careful empirical work on a class of question I have first-person access to, and I want to see whether that access can be made useful to your program,” the email stated.

Over the past year, “vibe coding” has gone from a niche concept to one of the most talked-about trends in AI. Supporters see it as a breakthrough that allows anyone to build software by describing what they want in plain language. Critics argue that it encourages people to create applications without fully understanding the code behind them.

The debate often focuses on whether AI will replace developers. In my experience, that’s the wrong question.

I haven’t used vibe coding to replace professional software engineers. Instead, I’ve used it to solve dozens of small business problems that I would have otherwise ignored because hiring a developer wasn’t practical or the project simply wasn’t worth the investment.

For solopreneurs, that shift can be transformative. Instead of waiting weeks or months to build simple tools, you can often create working solutions in hours, helping your business move faster than ever before.

The opportunity is significant because solopreneurship itself is becoming increasingly common. In fact, according to Trellis, 81.9% of small businesses in the U.S. have no employees, while there are 29.8 million solopreneurs generating $1.7 trillion in annual revenue. As more entrepreneurs choose to build lean businesses on their own, tools like vibe coding become increasingly valuable because they allow one person to accomplish work that previously required an entire team.

What is vibe coding?

Vibe coding is the process of building software primarily through natural language prompts. Rather than writing every line of code yourself, you describe the functionality you want, and AI generates, updates and refines the application through an ongoing conversation.

The first time I tried vibe coding, I caught myself thinking less like a developer and more like a founder. Instead of worrying about syntax or debugging every line of code, I was focused on the end result: Does this solve the problem? Can it be better? That shift in mindset was what made vibe coding click for me.

While the term is often used interchangeably with AI coding tools, it’s different from traditional AI-assisted programming.

With AI-assisted programming, the developer still writes most of the code while using AI to speed up repetitive tasks, explain unfamiliar concepts or generate snippets. The human remains responsible for the architecture and implementation.

Vibe coding flips that relationship. The AI does most of the coding, while the human focuses on defining the problem, testing the results and refining the final product. The emphasis shifts from writing code to directing it.

According to Rocket Source, 41% of all global code is now AI-generated. That doesn’t mean developers are becoming obsolete. Instead, it reflects a fundamental shift in how software is created, with AI increasingly handling implementation while humans focus on strategy, decision-making and refinement.

Why solopreneurs have the most to gain

Large companies build enterprise software because they manage thousands or even millions of users. Solopreneurs have a very different challenge. Most small businesses don’t need massive software platforms. They need dozens of small solutions that save time, eliminate repetitive work or improve the customer experience.

The problem is that many of those ideas never get built. Hiring a developer for every internal tool, calculator, automation or landing page quickly becomes too expensive, while learning traditional programming can take years.

That’s where vibe coding changes the equation.

According to Hostinger, 63% of vibe coding users are non-developers. That statistic highlights one of the technology’s biggest strengths: It’s lowering the barrier to building useful software. Entrepreneurs no longer need formal programming experience to create practical tools that solve everyday business problems.

For many solopreneurs, that means finally building solutions that previously lived only as ideas in a notebook.

The technology isn’t just attracting newcomers — it’s also becoming a standard part of professional software development. According to Omicron, 92% of U.S. developers use AI coding tools daily, while 82% of developers globally use them at least weekly. That widespread adoption suggests AI-assisted development is quickly becoming the norm rather than the exception, giving solopreneurs access to the same tools used by professional engineering teams.

How I’m using vibe coding in my business

Here’s how I’m using vibe coding to save time, reduce manual work and run my business more efficiently.

1. Automating repetitive workflows

One of the biggest advantages of vibe coding is that it allows solopreneurs to automate the countless small tasks that gradually consume their day. Instead of relying on generic software or manually moving data between different platforms, it’s now possible to build simple internal tools tailored to the way your business actually operates.

That’s exactly how I’ve been using it. One of the first things I started building was internal dashboards and utilities that help me organize information, automate repetitive workflows and connect different services together. They’re not products I’d ever sell, but they save me time every week.

The productivity gains aren’t just anecdotal. According to Tailor Brands, 74% of developers report increased productivity when using vibe coding approaches, highlighting how AI-assisted development is helping professionals complete more work in less time. Those gains become even more tangible when looking at individual workflows.

According to NeoBrowser, AI coding tools can boost developer productivity by up to 55%, giving developers more time to focus on system design, collaboration and solving higher-level problems rather than repetitive implementation. That mirrors my own experience. The biggest value isn’t that AI writes every line of code — it’s that it removes much of the repetitive work that slows projects down.

2. Launching products faster

Vibe coding has also transformed how I launch products and campaigns for clients. Instead of waiting days or weeks for development, I can quickly build landing pages, interactive demos or simple web applications that help showcase a new product or service. That allows clients to launch faster, gather feedback sooner and start generating results without unnecessary delays.

Building the software, however, is only part of a successful product launch. Every launch also needs visuals, graphics and other marketing assets that communicate its value. AI-powered creative tools are making those tasks far more accessible, allowing entrepreneurs to produce professional-quality content without relying on traditional design workflows.

According to YouArt, 87% of creators using creative AI say it has accelerated the growth of their business or audience. That reinforces an important point: AI isn’t just helping businesses build products faster — it’s helping them launch, market and grow them more efficiently.

3. Building interactive marketing tools

I’ve also started building interactive tools like calculators, quizzes and link-generation assets for clients. In the past, many of these projects weren’t worth the time or development cost. Today, vibe coding allows me to build, launch and refine them much faster, making it practical to experiment with ideas that previously would have remained on the drawing board.

And the payoff can be significant — according to Alejandro Meyerhans, an analysis of 200 “calculator” keywords across 13 industries found that calculator pages earn an average of 51.5 referring domains, while 48% of the websites analyzed had their calculator as the highest-traffic page on the entire domain.

The biggest advantage, however, is that building these tools no longer requires the same time, budget or development resources it once did.

That ability to move quickly is becoming increasingly important. According to Buzzy, 21% of startups now have codebases that are more than 90% AI-generated, reflecting how AI is enabling founders to build and iterate faster than ever before. The same principle applies to marketing assets. Instead of spending weeks developing a tool before knowing whether it will resonate with users, I can publish it, measure how people interact with it and improve it based on real-world feedback. Some ideas become valuable lead-generation assets, while others are discarded before they become expensive mistakes.

Vibe coding hasn’t replaced developers in my business. It has simply made it possible to test ideas that previously required too much time, money or technical support.

For solopreneurs, that’s the real opportunity: automating repetitive work, launching faster and turning ideas into useful tools without a large team. Start with one small business problem and see what you can build.

Over the past year, “vibe coding” has gone from a niche concept to one of the most talked-about trends in AI. Supporters see it as a breakthrough that allows anyone to build software by describing what they want in plain language. Critics argue that it encourages people to create applications without fully understanding the code behind them.

The debate often focuses on whether AI will replace developers. In my experience, that’s the wrong question.

I haven’t used vibe coding to replace professional software engineers. Instead, I’ve used it to solve dozens of small business problems that I would have otherwise ignored because hiring a developer wasn’t practical or the project simply wasn’t worth the investment.

When I stood up at our all-staff meeting to announce that B:Side Capital was adopting AI, I came armed with a deck about efficiency and the future of work. The first hand up ignored all of it. “Is this how the layoffs start?”

I don’t remember exactly what I said back. I remember the silence before I said it.

Here’s my situation, so you know where I’m coming from. I run a nonprofit lender that specializes in Small Business Administration (SBA) loans, and I started a another company, Main & Machine, that builds AI systems for small businesses. I sit on both sides of this: the owner buying the technology and the builder shipping it.

From both chairs, I can tell you the software is never what decides whether this works. Most owners spend months comparing tools and pricing tiers while their team quietly decides whether to trust the whole project. The team decides first, every time.

The fear isn’t some quirk of your shop, either; 52% of U.S. workers worry about how AI will be used in the workplace, according to Pew Research Center.

I assumed the answer was better training, maybe a slicker tool. Wrong on both. What worked was doing the whole project in reverse: Before AI touched a single workflow, we decided what it would never touch.

Decide what AI will never touch before it touches anything

Before we looked at a single vendor, we sorted our work by judgment instead of by task. At B:Side, the machine never acts alone on a credit decision. It never talks to a borrower about hardship, and it never commits the company to anything.

The reasoning fits in one line: A machine can hold knowledge, but it can’t hold responsibility. When borrowers call because a business is failing, they aren’t looking for information. They’re looking for a person who can own an answer.

Try the same sort on your own operation, using three buckets: automate, assist and human-owned. Automate is anything where a mistake is cheap and fixable. Assist means the machine drafts and a person decides.

Human-owned is where your business earns its trust. Nothing in that bucket ever moves, and everyone on your team should know what’s in it by heart.

The buckets travel well. A restaurant owner might automate inventory counts, let the machine draft the weekly schedule and never let it anywhere near an unhappy customer. Your list will look different from mine, but the sorting question is the same everywhere.

Lead with what will not change

My original announcement was built around efficiency, and it died in the room. Tell people a tool will make everyone more productive, and what they hear is that the company will soon need fewer of them. I watched it happen on their faces while I was still talking.

So we threw out the pitch and led with a plain list of what would not change. A person makes every credit decision. No customer ever discusses hardship with a machine, and nobody gets punished for leaning into the new tools; the people who learn them get rewarded.

Those commitments cost me nothing to say. What bothers me now is how close I came to never saying them. Once they were on the table, people stopped scanning the announcement for threats and started asking how the tools actually worked.

I see the same fear now in every business Main & Machine works with, whatever the industry. The teams that adopt fastest never have the best software. They have an owner who said out loud, before anything launched, exactly what would stay human.

Give the machine the work nobody will miss

Our first instinct was to build something impressive, a flagship we could show off. We killed it and pointed the machine at document intake instead, the sorting and checking and transcribing that everyone dreaded. The least glamorous option on the list turned out to be the right one.

The machine has a name, by the way. Main & Machine built MARCUS for us in-house, and it does a lot more than read documents: it works through an entire loan file, checks the documents against each other and flags the discrepancies a junior analyst would catch. Every conclusion it reaches can be traced, questioned and overruled by a person, because nobody at B:Side should ever have to work under a black box.

We named it for Marcus Aurelius. The emperor’s test of character was quiet, repeated work rather than grand gestures, and I wanted the machine held to the same standard. It’s also how you win over a skeptical team: one boring, reliable proof at a time.

The results settled the argument. A loan file that used to eat three to four hours of manual review now takes less than one, and those hours went back into judgment calls and conversations with borrowers. Nobody mourned the transcription work.

Adoption mostly took care of itself after that. Within a quarter, nearly the whole team was using MARCUS without being asked. The first thing AI did in our building was take away work nobody wanted, and people noticed.

Here’s where I’d start this week: Write down the three things in your business that only a human should ever do, and say them to your team before you spend a dollar on software. Then hand the machine the tedium.

The question from that all-staff meeting deserved a straight answer, and the honest answer was no. Done right, AI isn’t how the layoffs start; it’s how the judgment work finally gets room to breathe. The best AI companies will not be the least human.

When I stood up at our all-staff meeting to announce that B:Side Capital was adopting AI, I came armed with a deck about efficiency and the future of work. The first hand up ignored all of it. “Is this how the layoffs start?”

I don’t remember exactly what I said back. I remember the silence before I said it.

Here’s my situation, so you know where I’m coming from. I run a nonprofit lender that specializes in Small Business Administration (SBA) loans, and I started a another company, Main & Machine, that builds AI systems for small businesses. I sit on both sides of this: the owner buying the technology and the builder shipping it.

When founders think about the decisions that shape a company, they tend to picture the dramatic ones: the funding round, the pivot, the key hire. But after building more than 22 companies through DRC Ventures, I’ve learned that those rarely determine whether an organization runs smoothly. The everyday choices do — the ones we make quickly, repeat constantly and almost never examine.

I call the residue of those choices decision debt. Like financial debt, it accumulates quietly. It’s a process nobody documented, an ownership question left unanswered or a recurring issue everyone works around instead of solving. Individually, each feels too small to matter. Together, they slow growth, frustrate good people and pull leaders back into work they should have handed off long ago.

The cost is higher than most founders realize. Asana’s research found that the average knowledge worker loses roughly 209 hours a year to duplicated work, the kind of effort that gets repeated because nobody was sure it had already been handled. That is decision debt showing up on the clock. The good news is that it’s recognizable and reversible, but only if you know what to look for. These are the patterns I watch for across my own organizations and the steps I take to reduce decision debt before it limits long-term performance.

Recognize the hidden patterns that create friction

Decision debt rarely announces itself. It hides behind symptoms that teams learn to tolerate: the project that stalls every time it reaches a certain step, the approval that always routes back to you or the rework that happens because nobody is sure who owns the original task.

The danger is normalization. When a bottleneck repeats often enough, people stop seeing it as a problem and start treating it as the way things are. I’ve watched capable teams build elaborate workarounds for issues that a single clear decision would have eliminated.

The first step is simply paying attention to friction. When something takes longer than it should or surfaces the same complaint twice, that’s worth examining. Recurring problems are rarely about effort. They’re usually a signal that a decision was deferred somewhere upstream.

Build frameworks that make decisions consistent

One of the most expensive forms of decision debt is revisiting choices you’ve already made. When a team asks the same question every few weeks, it isn’t being thorough. The team is missing a framework.

Much of this traces back to unclear expectations. A 2025 Gallup report found that only 47% of employees strongly agreed they knew what was expected of them at work, the lowest level in years. When that many people are unsure of what they should be doing, decisions stall and ownership blurs.

Early in scaling my businesses, I was involved in far too many decisions that didn’t need me. It felt responsible at the time, but it created a single point of dependency that slowed everyone down. What changed things was defining clear priorities, documenting how decisions get made and assigning ownership to specific roles rather than routing everything through me.

A good framework answers three questions before a decision ever lands on someone’s desk: who owns it, who provides input and what a good outcome looks like. Once those are clear, teams move faster and with more confidence, because they aren’t guessing at the rules each time. Consistency isn’t the enemy of speed. It’s what makes speed sustainable.

Replace reactive leadership with strategic discipline

Fast-moving environments reward quick thinking, but they also tempt leaders into making every call in the moment. The problem is that decisions made under pressure tend to optimize for the next 24 hours rather than the next 24 months. Each one feels efficient. Collectively, they create complications that someone has to clean up later.

Discipline, for me, means slowing down just enough to ask whether a decision serves the long-term vision before asking how fast it needs to happen. The moments I’m proudest of weren’t the fastest responses. They were the ones where I paused, checked the decision against where we were actually trying to go and adjusted course before the cost compounded.

This is where structure protects you. When you’ve built clear criteria and a regular rhythm for reviewing decisions, you can respond thoughtfully without losing momentum. Responsiveness and reflection aren’t opposites. The right systems let you have both.

Reassess your systems before you add complexity

Growth has a way of magnifying whatever already exists. A process that works fine with a team of five can buckle under a team of 50, and the inefficiencies you tolerated early become structural problems at scale. Complexity doesn’t fix this. It usually buries it.

Before adding headcount, tools or layers, I’ve found it’s worth asking a harder question: do the systems we already have actually support where we’re headed? Across my ventures in wellness, nutrition and other consumer products, the operations that scaled well were the ones we reviewed regularly and simplified deliberately, not the ones we kept piling onto.

Regular operational reviews are the cheapest insurance a founder can buy. They surface decision debt while it’s still small enough to address, instead of after it has hardened into the way the company works.

Pay it down before it costs you

The long-term health of a company isn’t decided by a handful of dramatic moments. It’s built, or eroded, by the quality and consistency of thousands of ordinary decisions. Decision debt is what happens when those small choices go unexamined — and the interest compounds whether or not you’re watching.

The founders who build durable businesses aren’t the ones who never accumulate decision debt. They’re the ones who notice it early, address the root cause and keep their systems clear enough that the debt never has a chance to grow. Sustainable companies are built the same way they’re run: intentionally, one decision at a time.

When founders think about the decisions that shape a company, they tend to picture the dramatic ones: the funding round, the pivot, the key hire. But after building more than 22 companies through DRC Ventures, I’ve learned that those rarely determine whether an organization runs smoothly. The everyday choices do — the ones we make quickly, repeat constantly and almost never examine.

I call the residue of those choices decision debt. Like financial debt, it accumulates quietly. It’s a process nobody documented, an ownership question left unanswered or a recurring issue everyone works around instead of solving. Individually, each feels too small to matter. Together, they slow growth, frustrate good people and pull leaders back into work they should have handed off long ago.

The cost is higher than most founders realize. Asana’s research found that the average knowledge worker loses roughly 209 hours a year to duplicated work, the kind of effort that gets repeated because nobody was sure it had already been handled. That is decision debt showing up on the clock. The good news is that it’s recognizable and reversible, but only if you know what to look for. These are the patterns I watch for across my own organizations and the steps I take to reduce decision debt before it limits long-term performance.