The Positive Loop

Late-stage capitalism solved a problem nobody asked it to solve: how do you keep billions of people engaged when wages stagnate, housing is unaffordable, and traditional paths to success narrow every year?

The answer was dopamine. Not drugs. Feeds.

Social media, short-form video, infinite scroll, algorithmic recommendations. The entire consumer internet was optimized for one metric: time on screen. The more time you spend, the more ads you see, the more data you generate, the more precisely the next hit can be delivered. That is the documented business model of every major platform built since 2010.

The result is what The Gameable Society documented: a population fluent in the language of self-improvement but structurally prevented from acting on it. People who can explain periodization, compound interest, and dopamine regulation. Who know the meta. Who have never played the game.

This connects directly to what Preference for Legible Risk explored: when traditional paths feel closed, people gravitate toward whatever offers the most legible dopamine. Doom scrolling was the cheapest, most accessible version of that. Zero cost, zero effort, immediate reward, zero output.

We turned ourselves into dopamine-driven animals. Not because we are weak. Because the infrastructure was built to make us that way. Every pixel, every notification, every autoplay was designed by teams of engineers whose KPI was your attention.

The neuroscience is more specific than most people realize. Kent Berridge's lab at the University of Michigan spent decades proving that dopamine is not a pleasure chemical. It is a wanting chemical. Dopamine drives craving, motivation, the compulsive pull toward the next hit. Pleasure, what Berridge calls "liking," runs on a separate system entirely. Mice with artificially elevated dopamine pursued sugar more aggressively but did not enjoy it more. Every platform that hacked your dopamine did not make scrolling pleasurable. It made scrolling compulsive. You kept doing it not because it felt good but because your brain could not stop wanting to do it. That distinction matters for everything that follows.

The neural damage is measurable. A 2025 study in Neurocognitive Impact of Social Media Usage found a 22% reduction in prefrontal cortex beta power after just 20 minutes of social media engagement. The prefrontal cortex is the part of your brain responsible for decision-making, impulse control, and long-term planning. Twenty minutes of scrolling and a fifth of its capacity is suppressed. That suppression leads to heightened impulsivity, which leads to more scrolling, which leads to more suppression. The feed does not just waste your time. It degrades the neural machinery you need to stop wasting it.

And then something changed.

Picture someone who has never written a line of code. They wake up, open their phone, describe a product idea in plain English. By lunch, it is live on the internet. Not a mockup. A working product with real users accessing it.

That pipeline did not exist 18 months ago. It emerged that fast, and it works like this:

This pipeline did not exist 18 months ago. Not for regular people. Coding required years of learning. Design required specialized tools. Deployment required DevOps knowledge. Research required academic access or expensive databases. Automation required hiring engineers.

Now a person with zero technical background can:

• Run deep research studies that used to require a team of analysts, Bloomberg terminals, and weeks of work. Fetch real-time data, cross-reference sources, generate charts, publish findings. From a phone.
• Build functional products: websites, tools, dashboards, automations. Not mockups. Working software that serves real users.
• Automate repetitive work: data entry, report generation, monitoring, alerts. Tasks that used to require hiring someone or doing them manually every day.
• Translate creative visions into real output. "I want a portfolio site that looks like this" becomes a real site in an afternoon. "I need a script that does X every morning" becomes a working cron job.

The speed at which this became possible is the part nobody has processed. It is not that AI is improving gradually. The gap between "idea" and "deployed product" collapsed from months to hours, then hours to minutes. It happened so fast that most people still think "coding" is something you need a CS degree to do.

Think about this.

A video game with the TAM of every single human being on earth. Not a niche. Not a demographic. Everyone.

But unlike every other game, playing it actually does something for your life. It makes you money. Makes you more successful. Helps you translate creative visions into real output. Teaches you things you did not know. Solves problems you thought were too complex.

And there is a DLC dropping every day. New capabilities. New models. New integrations. Every morning you wake up and the game is slightly more powerful than yesterday.

And almost nobody is playing yet. 84% of professional developers use AI tools, but mostly to assist with code they already understand, not to build from scratch. Among developers who have heard of vibe coding, only 15% have tried it. 0.4% describe themselves as enthusiastic practitioners. Among the general population the number is effectively zero. The people who have discovered what this actually does are a rounding error.

The gaming comparison matters because it explains the adoption curve. Games spread through word of mouth, through watching other people play, through the feeling of "I could do that." AI adoption will be faster than any technology in history because the barrier to entry is not learning a new skill. It is typing a sentence.

The most powerful adoption curves happen when the cost of entry drops to near zero and the upside is visible. That is exactly what is happening here.

Technology has always been confusing. That was the moat. The reason most people never built anything was not a lack of ideas. It was the lack of a translation layer between "what I want" and "how computers work."

AI collapsed that layer.

You do not need to know what a server is. You do not need to know what React is. You do not need to know what an API endpoint does. You need to describe what you want. That is it. The AI translates human intent into technical implementation.

That is the black box that changes everything. You put an idea in. Something useful comes out. You do not need to understand the internals. You need to be good at having ideas and knowing what good looks like.

The closest comparison is the printing press. Before Gutenberg, making a book required being a scribe, understanding bookbinding, having access to materials. After Gutenberg, you needed to have something to say. The press handled the rest. AI is the Gutenberg press for software, research, automation, and design. The skill shifts from implementation to taste.

Notice what happens when you ship something with AI. The code runs. The page loads. Something that did not exist five minutes ago is now live on the internet. Your brain fires. Not the dull buzz of a liked post or a finished scroll session. A hit that says: I made that.

That feeling is the key to the whole thing. For the first time, there is an interface that is as addictive as doom scrolling but produces real output. The loop is not: scroll → content → brief pleasure → nothing. The loop is:

This has never happened before. Every previous internet dopamine loop was extractive: it took your time and gave you nothing durable in return. Likes, views, streaks. They vanish. You cannot compound a TikTok scroll session into anything.

The neuroscience of why this loop sustains goes deeper than dopamine. Mihaly Csikszentmihalyi spent 30 years studying what he called "flow," the state where a person is so absorbed in an activity that time disappears. His research across artists, athletes, surgeons, and programmers found the same conditions every time: clear goals, immediate feedback, and a challenge that matches skill level. AI building hits all three. You describe what you want (clear goal). You watch it get built in real time (immediate feedback). The complexity scales with your ambition (matching challenge). A 2024 neuroimaging study at Drexel by John Kounios and David Rosen found what happens in the brain during creative flow: the superior frontal gyri, the region responsible for conscious executive control, go quiet. The brain voluntarily releases oversight and lets trained circuits run. That is the opposite of what scrolling does, where the prefrontal cortex is suppressed against your will. In flow, the brain chooses to let go because the activity has earned its trust.

Edward Deci and Richard Ryan's Self-Determination Theory explains the rest. Their research, spanning over 40 years, found that sustained motivation requires three psychological needs: autonomy (choosing what you do), competence (getting better at it), and relatedness (connecting with others through it). Doom scrolling satisfies none of these. You do not choose the content, the algorithm does. You do not improve at anything. You do not connect, you spectate. AI building satisfies all three. You decide what to build. Each session makes you better at prompting, evaluating, directing. You ship things other people can use. Di Domenico and Ryan's 2017 neuroscience review showed that these three needs are subserved by dopaminergic systems in the ventral tegmental area and nucleus accumbens, the same reward circuitry Berridge studied. The difference: intrinsically motivated dopamine fires in service of growth. Scroll-driven dopamine fires in service of nothing.

AI loops are compounding. The website you built yesterday is still live today. The automation you set up last week is still running. The research you published is still being read. Each session deposits something real into the world that persists after the dopamine fades.

The key thing: the productive loop is just as addictive as the extractive one. The feedback is immediate. The reward is visible. The progress is tangible. The only difference is that when you come up for air, you have something to show for it.

The honest ceiling: a productive addiction is still an addiction. Clinical research on transfer addiction documents what happens when you replace one compulsive behavior with another. The output changes. The compulsive driver does not. People who quit substance abuse and shift to compulsive exercise produce visible results but the underlying pattern of avoiding discomfort through compulsive action persists. The positive loop has the same risk. Building things at 2am because you cannot stop is not automatically healthy just because the output is real. The difference between a productive loop and a productive compulsion is whether you can choose to stop. If you cannot, the dopamine is driving you the same way it drove the scroll. Berridge's wanting system does not care what it is pointed at.

The phone, the device that turned us into dopamine addicts, is now the device that can turn us into builders. Same addictive qualities. First time they are pointed at production. Whether that is liberation or a lateral move depends on the person using it.

Ask someone if they have used AI. They will say yes. Ask them what they built with it. Silence. Maybe they wrote an email. Maybe they asked it a trivia question. That is where most people are right now.

The gap between "I have used ChatGPT" and "I use AI to build things" is the gap between having a phone and building an app. Most people have not used AI for anything beyond a chatbot conversation. They have not built a product, run research, automated a workflow, or shipped something real. The speed of change is hiding the size of the change.

The barrier to entry feels annoying. That is the word people use. It feels like homework. It feels like learning a new tool when the last one was fine. But the growth underneath that friction is too exponential to sit out. The gap between people who conceptualize what AI can do and people who have not touched it yet is the largest capability mismatch in technology right now. And it is compounding. Every month the tools improve, the gap widens, and more people fall further behind without realizing how far behind they already are. Not because they are incapable. Because the on-ramp does not exist yet for how they think.

Think early internet. In 1995, most people thought the internet was for email and maybe some research. The idea that it would restructure commerce, media, communication, and culture was not on the radar for 99% of the population. We are in the equivalent of 1995, except the adoption curve is compressed from decades to years.

Low barriers, network effects, compressed time. These are the conditions that have always produced outsized outcomes. They are all present right now, producing productivity gains so quietly that most people and most institutions have not fully registered how fast things are moving.

Here is what this actually looks like in practice:

Every one of these follows the same structure: human intent → natural language → AI translation → real output. The translation layer is the breakthrough. Everything else follows from it.

The ceiling on this pipeline is quality. A METR study found that experienced developers were 19% slower with AI tools despite believing they were 20% faster. GitClear's analysis of 153 million lines of code found that code churn, the rate of code written and then deleted or rewritten within two weeks, doubled between 2021 and 2024 as AI adoption spread. Veracode's 2025 GenAI Code Security Report tested over 100 LLMs across four languages and found that AI-generated code introduced security vulnerabilities in 45% of all coding tasks, with flaws mapping directly to the OWASP Top 10. The pipeline works. The output is real. But the gap between "deployed" and "production-grade" is where the honest assessment lives. Vibe-coded products are real products. They are not yet reliable products at scale. That ceiling drops every quarter as models improve, but pretending it does not exist makes the argument weaker than acknowledging it.

It is 11pm. You are tired, stressed, done for the day. You reach for your phone. Normally this is when you open Instagram or TikTok and lose 45 minutes to a feed. But tonight you open a different app. You describe something you have been thinking about. Twenty minutes later, you are looking at a working prototype.

That is the substitution that changes everything. Doom scrolling persists because it is the path of least resistance to dopamine. Zero activation energy. Immediate reward. The cost is invisible (time, attention, mental health) and deferred.

AI tools have nearly matched that activation energy. Open your phone. Describe what you want. Watch it happen. The reward is almost as immediate. But "almost" matters. Doom scrolling requires zero cognitive effort. Zero. Content is delivered to you. AI building requires having an idea, articulating it, evaluating output, iterating. That is a fundamentally different cognitive demand. For people who already have something they want to build, the effort feels comparable. For people who do not, scrolling is still easier. The substitution only works when the person has a vision the tool can serve. The on-ramp is getting shorter every quarter, but it is not at zero yet for everyone.

For the people where it does work, the substitution is real. Instead of losing 45 minutes to a feed, you spent 45 minutes building something. The dopamine is comparable. The output is not.

There is something else happening that matters. The hardest part of doing any work has always been starting. The weight of the starting gap, sitting down, opening the blank page, grinding through the first hour before momentum kicks in. That weight lifts when the process is fun. You are not grinding through the hard part alone. You are having a conversation and watching things appear. The activation energy drops because the bridge between "I should do this" and "I am doing this" stops feeling like effort and starts feeling like play. That is the shift nobody expected. Not that AI makes work faster. That it makes starting work feel like something you want to do instead of something you have to force yourself through.

The behavioral shift is already visible in early adopters. People who discover vibe coding or AI-assisted building report the same compulsive usage patterns as social media: checking in constantly, thinking about it when away, feeling pulled back to it. Except the output is products, research, automations, and skills instead of scroll sessions and parasocial relationships.

This connects to the core idea in The Gameable Society: the spectator class exists because consuming optimization content was easier than doing optimization. AI eliminated that gap. The effort to do is now comparable to the effort to watch. For the first time, playing the game is as easy as watching someone else play it.

Look at what is already happening around you. Someone who used to spend their evenings scrolling built a side project last month. A friend who "isn't technical" shipped a website over a weekend. A teenager with zero coding experience launched a working product and has paying users.

These are not outliers. They are early signals of a structural shift in how humans spend their attention. Not gradual. Abrupt. Compressed by the same dynamics that compressed information and cultural change.

Technology reshapes behavior faster than institutions can adapt. Every time. AI is the most extreme version of this. It is not making existing processes faster. It is making previously impossible things trivial. And the people who internalize this earliest will compound advantages that look, in hindsight, like the early internet and early social media adoption curves combined.

The unknown is whether the loop sustains. Much of the reported compulsion around AI building may be novelty-driven. The tool is new. The capabilities expand weekly. The feeling of discovering what it can do is itself a dopamine trigger independent of the output. No longitudinal data exists on whether the productive loop holds after the novelty fades. The honest bet: it holds for people building toward something they care about, the same way creative flow states sustain for artists and builders across decades. For people chasing the tool's novelty rather than their own vision, the loop may fade the way every shiny technology fades. The difference between the two groups is purpose, not the tool.