Paul Bakaus created jQuery UI and later worked at Google. He now runs a company of his own and builds Impeccable, an open-source tool intended to solve design problems in AI coding agents. He argues that the most important human capability in the age of AI is not the creativity to add more things, but the ability to decide what should not be done and remove what is unnecessary. He also believes that a solo founder should hire a Chief of Staff before hiring more developers, protecting the time needed to continue doing what the founder does best: making things.
1. The Coming Age of Founders in Their 40s
The video opens with a question: what role remains for people when AI can write code? Bakaus says the ability we need most is not the power to produce new things indiscriminately. It is the judgment to remove what is unnecessary and decide what not to do.
"The ability humans need most right now is the ability to remove things—to say what should not be done."
He disagrees with the common image of Steve Jobs as simply a brilliant creator. Jobs's real strength, he says, was not inventing everything himself. It was his editorial ability: rejecting most of the ideas placed before him and releasing only the few that truly mattered.
"Steve Jobs was the greatest editor in history. He rejected 99% of the work, and the remaining 1% is what reached us."
Bakaus believes an era is approaching in which older founders may have an advantage over younger ones. He no longer has the energy he had in his 20s, but years of experience have given him intuition and craftsmanship—a sense of what works and what does not.
He began as an individual contributor, then became an engineer, an engineering manager, and eventually a manager of other managers. What troubled him was that each step upward in the organization took him further away from making things himself.
"It felt as though climbing the hill required me to give up the founder's life—or give up building things with my own hands. I didn't like that."
AI is changing that arrangement. With sufficient experience and intuition in a field, one person can use AI to build what they want directly.
"You may not have as much energy as you did at 20, but intuition and craftsmanship can make up for some of that lost energy."
2. Lowering the Floor and Raising the Ceiling
Bakaus identifies anti-gatekeeping—removing barriers to entry—as the theme running through his career. He has long believed that the internet should not belong to a small group and that more people should be able to participate.
At first, he cared less about coding syntax than about putting his ideas into the world. By building a website on his own computer and connecting it to the internet, he could show his work to people everywhere without asking permission from a television station, newspaper, or other institution.
"The web was a permissionless culture. If I built a website on my computer and connected it, the whole world could see what I was doing."
At the time, however, creating something beautiful and interactive on the web required a specialist JavaScript developer. People who were neither web designers nor programmers had little opportunity to participate in building web applications. Bakaus found that exclusion frustrating and became interested in tools that would allow many more people to create on the web.
To lower the floor is to reduce the barrier far enough that a beginner can start. To raise the ceiling is to give skilled practitioners room to produce more ambitious results.
That philosophy continued through jQuery, jQuery UI, and now Impeccable.
3. From Rural Germany to the Front Edge of the Web
Bakaus did not grow up in a Silicon Valley technology family. He was raised in Germany's Black Forest, a region that felt far removed from technology and startup culture. At the time, he believed it lagged ten or fifteen years behind other places.
"There wasn't much happening there in technology. I always felt that the region was ten or fifteen years behind."
The internet gave him a route beyond those geographic limits. The early-2000s web was not yet divided into giant platforms and countless subcultures; it felt more like one large village. Online, he could encounter global developments that were invisible in his local community.
He right-clicked websites, selected View Source, and studied the code other people had written. With no one nearby doing the same thing, he could not tell whether he was hopelessly behind or unusually far ahead.
"I was the strange kid coding in the basement and putting things out. I had no idea whether I was terribly late or incredibly early."
His first practical validation came from small local jobs. While working at a mobile-phone shop, he noticed that the store had no software for managing customer-support requests, so he built a program himself. The owner was astonished.
Bakaus believes that experience remains relevant today. A frustration discovered during a part-time job or internship can later become the seed of a business.
4. jQuery UI and the Democratization of Web Development
Bakaus was working on large JavaScript applications at the time. For Deutsche Telekom and T-Mobile, he built a web application with email and calendar features—a remarkably large web client before Gmail appeared.
Plain JavaScript made complex, polished interfaces difficult to build. Then he discovered the newly emerging jQuery library.
Many developers initially regarded jQuery as an odd way to write JavaScript. Yet it allowed people who were not professional engineers to add interaction to a web page.
"jQuery gave web designers—and people who didn't call themselves engineers—a chance to participate."
Bakaus fixed jQuery bugs and joined the core team. Because he was especially interested in user interfaces, he built features needed for richer web experiences, including drag-and-drop behavior and accordion menus.
A third-party library called Interface already existed, but it was full of bugs. Bakaus repaired and contributed to it. John Resig, the creator of jQuery, later asked him to build an official UI library, and jQuery UI was born.
jQuery UI became a defining project in the transition from Web 1.0 to Web 2.0, when static websites were giving way to richer, interactive web applications.
5. A First Business at 11, Then a Game-Engine Startup
Bakaus started a small consulting business when he was only eleven or twelve. Because he was so young, he had to visit the local city hall in Germany and request permission to operate it.
"When a child said he wanted to create a company, the people at city hall reacted as if to ask, 'What on earth does that mean?'"
He went through the review with his mother and ultimately received special permission. The business did not make him rich, but starting it at such an early age became an important experience.
He says he never searched for an idea simply because he wanted to create a company. He found interesting problems, tried to solve them, and companies emerged from that work.
After jQuery UI, he began thinking about how to let anyone build games on the web. Game development still belonged largely to specialists; he wanted to lower the barrier so that any web developer could make a game.
Around 2010, he and a colleague started a game-engine company. They built an engine capable of running isometric games like FarmVille and made it work on the relatively underpowered first-generation iPad.
"The idea was to let every web developer become a game developer."
They succeeded technically, but distribution and monetization remained difficult. Rather than sell the engine alone, the company attempted to demonstrate it through a flagship product: an isometric world resembling Facebook.
European investors were reluctant to fund an unproven idea. They kept asking, "Has this already been validated?" But the whole point of Bakaus's idea was that no one had attempted it before.
Customers responded much more positively. Large companies including Microsoft and Zynga showed interest, and Zynga eventually acquired the startup. Bakaus then spent several years working in research and development there.
6. Why He Became a Solo Founder Again
At his first company, Bakaus's co-founder served as CEO while he led technology. The co-founder handled business development, fundraising, and partnerships; Bakaus built the engine and the underlying technology. The division worked well.
Sixteen years later, however, Bakaus started Renaissance Geek alone. After spending years as a manager in a large organization, he felt an intense need to make something himself again.
"I felt that I had to build something again."
He wanted to test how far he could scale his own abilities. With AI agents, could one person do the work of a five-to-ten-person development agency? He deliberately accepted extremely difficult contract projects to find out.
The experiment showed that one person could handle a surprising amount of technical work. It also revealed a new limit. The number of AI agents could grow; human attention and concentration could not.
"Almost everything in software development can be scaled, except human attention."
After considering his highest-value contribution, Bakaus concluded that he needed to protect his making time. Each morning begins with the question, "What value will I create today?"
"A day in which I create no value is a deeply sad day."
He tries to outsource or delegate company operations, administration, and investor communication so they do not consume that time. What he does not want is to hire a large team of developers and designers only to become a middle manager again.
"I have no interest in handing the work I love to ten other people and becoming a middle manager again."
7. Why His First Hire Was a Chief of Staff
For his first key hire, Bakaus chose a Chief of Staff, not an engineer or designer. He initially imagined a person who could combine elements of Chief of Staff, operations lead, and executive assistant.
At first he tried to place all three responsibilities in one job description. People around him warned that this was poor design, especially because executive-assistant work differs in character from operations and strategy.
He was not looking for a senior executive who had performed one narrow function inside a giant corporation. He wanted someone with little ego and a strong bias toward execution, capable of solving whatever problem appeared in a small company.
"This isn't a job where you simply produce elegant strategies. You have to roll up your sleeves and do whatever needs to be done."
The work might begin with investor projects, certifications, opening bank accounts, or issuing corporate cards. Later it could shift to a partnership with a major company or the construction of a large payment system.
Someone whose operating experience comes entirely from a large corporation may struggle with that reality. Bakaus says the right person must adapt naturally as the type of work changes and be willing to handle almost anything.
The primary reason to hire a Chief of Staff first is to protect the founder's time. A typical founder hires developers, designers, and product managers first, handing away the work they used to make themselves. The founder then becomes the operations person.
"If investors and customers backed you because you were excellent at making things, why would you give that ability up? It makes no sense."
8. Impeccable: Reducing AI Design Slop
Renaissance Geek's first product, Impeccable, is an open-source design skill that helps AI coding agents produce better design.
Here, a "skill" means a capability added to a coding agent such as Claude Code or GitHub Copilot. Impeccable is intended to move the agent beyond code that merely works and help it understand design intent and context.
"Impeccable is a skill that makes coding agents better at design."
AI output often looks awkward, repetitive, and unmistakably machine-made. Bakaus calls it AI slop: ordinary, messy work produced by AI at enormous volume.
The form of AI slop keeps changing. The design failures common a year ago differ from today's failures, so Impeccable does not target only one fixed problem. It raises the baseline and creates a path toward stronger results.
"Impeccable establishes a baseline for good design. But the final step from good to great still requires a person."
Bakaus originally built Impeccable for his own work. When he asked Claude Code to make a section "stronger," the model would respond literally and unhelpfully—perhaps by changing a color to black or blue.
He did not simply want a darker color. He wanted the section to draw attention while preserving the direction of the existing design system. After noticing that designers obtained much better results from AI tools, he realized that the language and vocabulary used to discuss design made a major difference.
9. How Impeccable Preserves a Brand
Impeccable does more than inspect colors and fonts. When first installed, it conducts a product interview.
It asks what kind of app or website is being built, who will use it, and in what context. An iPad app for children, for example, will require a completely different design depending on whether the real audience is children or their parents.
"You have to ask who the product is for. A website for a child should be designed differently from a website for the child's parents."
Impeccable then analyzes the codebase and creates a design.md file containing information about the brand's colors, typography, design tokens, components, and broader design system.
The file does not replace a complete design system, but it gives the AI a reference that keeps its output from straying too far from the existing brand. When a user asks for a new page or section, Impeccable considers the product information and design information together.
After the code is written, design hooks and linters check it. They identify fonts or colors that do not belong to the brand and flag the kinds of cluttered patterns associated with AI-generated design.
"If the AI uses a font or color that doesn't exist elsewhere in the design system, it is flagged automatically."
That feedback lets the AI recognize a mistake and revise the work to fit the existing system. The result is not one clever prompt, but a structure combining product understanding, design documentation, and validation after generation.
10. Open-Source Distribution and Organic Growth
Bakaus did not initially set out to turn Impeccable into a business. He had several ideas and assumed that even if some failed, one of the others might succeed.
A more important signal arrived when people began asking on their own, "Why is this free?" and, "How can I pay for it?"
"People started asking, 'Why is it free? How can I pay you?' That is an extraordinarily strong signal of product-market fit."
Now that AI has caused an explosion in the number of products being built, reaching the first users can be as difficult as making the product itself. Bakaus's existing followers were not interested in his new subject area, and his first posts about AI and design received almost no response.
"I thought my posts were excellent, but nobody cared."
He therefore released the most valuable thing he could make for free and allowed people to use it immediately. Some projects failed, but Impeccable attracted attention, and both users and followers grew organically.
He believes an early product can benefit from giving away value in order to establish usefulness and distribution first. Launching a product with no path to reach users at all, however, remains extremely difficult.
11. What Zynga Taught Him About Real Virality
Bakaus saw viral growth at work firsthand at Zynga. Facebook strongly encouraged games to publish to user profiles and invite friends.
An in-game action could generate a "tell your friends" post across many profiles, which pulled more users back into the game. At the time, a game might be considered a failure unless it reached hundreds of millions of monthly or daily active users.
Bakaus distinguishes this as forced virality. A platform structure that pushes people to spread a product is different from users recommending it because they genuinely love it.
"There is a major difference between forced virality and organic virality."
Impeccable did not force advertising, influencer campaigns, or user-generated content. The team believed that people who experienced real value would tell others voluntarily.
"The only thing I can optimize is the experience of each individual user. After that, I have to hope it spreads naturally."
What matters most to Bakaus is how people discuss the product when its maker is not present.
"What matters is the conversation two people have after trying the product when I'm not there. If they say, 'I tried it and it wasn't very good,' that's the end. But if they say, 'This is so good that I told five people,' then you may have something."
He would rather have a small group of customers who love a product deeply than a large number who try it once and leave.
12. Quality and Integrity Before Reckless Growth
Integrity is central to the way Bakaus wants to build his company. If a customer used the product and found that its advertised capability did not work, he says he would feel like a fraud.
"If a customer said, 'The feature you advertised didn't work,' I would feel that I had failed completely."
He values feedback that genuinely helps, even when it is uncomfortable, over merely appearing pleasant.
"Don't aim to be nice. Aim to care."
AI is full of unverified tools and exaggerated claims. A person may test an AI skill a few times on their own computer and publish it immediately, even though success in one environment says little about how it will behave in someone else's.
Bakaus tested every line of Impeccable in real agent environments. If a proposed rule for removing AI slop did not improve results in practice, he deleted it.
"If a rule didn't work, I removed it from the skill no matter how sensible it sounded."
Good products, he argues, require real validation and honest feedback more than inflated marketing.
13. Where Humans Belong in an AI Workflow
Bakaus describes two extremes in the use of AI. At one extreme, several agents hand work to one another and operate autonomously for days. At the other, a person reviews every line of code and uses AI only in tiny increments.
The practical answer lies between them. The important skill is not constant intervention, but knowing when to intervene.
Creative work has three broad stages:
- Deciding what to make
- Making it
- Refining a good result into a great one
Human involvement is essential in the first stage. If no one has decided who the product is for, why it is needed, or what should be made, an AI can build efficiently in the wrong direction.
AI can handle a large share of the second stage. Even then, an agent may work well for an hour and suddenly depart from the original vision, so a person needs to check its direction periodically.
Human judgment becomes especially important in the final ten or twenty percent: the work of refinement.
"Most AI-generated code contains too much. The writing is too long, and the design is too elaborate and complicated."
The most important human task is therefore deciding what to remove, not what else to add.
"Delete this entire section. Simplify this. We don't need that. That kind of judgment is enormously important right now."
Bakaus believes AI can now produce work that is reasonably good, but cannot create an excellent finished artifact from one prompt.
"I have yet to see a design, piece of writing, codebase, image, or video that was 100% complete in the first result produced by AI."
Claims that one prompt will produce a beautiful landing page do not match reality. Design is iterative, and human perspective and taste remain necessary.
14. Steve Jobs and the Ability to Remove
Returning to Steve Jobs, Bakaus says great work does not come from adding a great many elements. It comes from removing most of them until the remaining core is clear.
"What made Steve Jobs special was not knowing what to make. It was knowing what not to ship."
That does not mean every design must be minimalist. An internal enterprise tool or system-level application may be better when it uses system fonts and familiar interface conventions.
A consumer product reached through an Instagram advertisement faces a different problem. If it must capture attention within five seconds, it needs a distinctive personality. Design should follow the purpose of the product and the people it serves.
AI tends to begin with more elements than the result needs. Bakaus says the editor must cut them aggressively.
"Cut until it hurts. You can always bring something back, but the first version almost always contains too much."
People can become attached even to AI-generated work. A striking logo section or case-study area that appeared by chance may be difficult to delete. If it does not improve the product as a whole, however, it should go.
15. Working with AI Is Like Carving Marble
Bakaus compares creating with AI to carving a sculpture from stone. The model contains an enormous range of possibilities, but the desired result does not emerge fully formed.
"Everything is already inside the model. You still have to carve it into the shape you want."
He has described the process as "carving a shader out of Claude Code." The first result may be complicated and overstated. Repeated refinement and deletion bring it closer to something that can actually be used.
The problem is not that an LLM knows nothing about humor. It knows too many kinds of humor and does not know which one the user enjoys. The user must learn to find the desired style among the model's many possibilities.
"The model contains hundreds of thousands of definitions of humor. What matters is finding the precise definition that you find funny."
Writing a good prompt is only part of the job. The user must evaluate the output and correct its direction. A person who uses AI well is not someone who delegates everything, but someone who understands their own standards and taste clearly enough to shape the model's output.
16. The Risks and Limits of Solo Founding
Starting a company alone is not always the right choice. Bakaus identifies the absence of an objective second opinion as one of its greatest difficulties.
"You have to be a little crazy to start something. But it can be very hard when no one is there to tell you whether you are actually crazy."
A solo founder therefore needs sounding boards—friends, family, mentors, or investors who will test the founder's thinking. The most valuable people are not those who praise everything, but those willing to offer an objective view.
Another danger is failing to protect attention. Founders should concentrate on what they do best and use contractors, operations staff, or a Chief of Staff to support the rest.
Bakaus observes his biological rhythm across the day. Decision fatigue is low in the morning, so he schedules important meetings then. In the afternoon, he feels more creative but more tired of making decisions, so he turns to production work.
"Listen carefully to what your body is telling you. Different people work best in different ways."
He also believes that creating tangible value every day helps prevent burnout. In his former corporate life, many days ended without a clear answer to what value he had brought into the world. He found that deeply difficult.
"You should be able to answer what value you created each day. If you can't, it may lead to burnout."
17. Small Companies May Not Need Middle Management
Processes that work in a large company do not automatically belong in a small startup. Hire someone from a big corporation too early, Bakaus says, and they may immediately introduce a monorepo, ticketing systems, and an elaborate development process.
Going from zero to one is fundamentally different from scaling from five to ten.
"Being good at going from five to ten does not mean you are good at going from zero to one."
When AI can generate tens of thousands of lines of code in a day, the traditional sequence of handoffs—from design to engineering, or product to design—may no longer make sense.
Bakaus expects more designers to work directly in code, more engineers to understand design, and more product managers to operate across several disciplines.
"Designers are moving into code, engineers are moving into design, and product managers are moving across every area."
Unnecessary processes and roles should be removed as those boundaries change. A small company that imitates a large one can lose both speed and momentum.
18. Why You Might Start Alone
Bakaus also sees clear advantages in solo founding. The largest is that no one else must agree before a decision can be made.
"When you start alone, you are accountable only to yourself."
The company can move at the speed of the founder's thinking, trying ideas, failing, and changing direction quickly during the early search. A founder still exploring ideas does not need to waste time waiting for a co-founder.
"If there is something you want to build, you don't necessarily need another person at the beginning."
Waiting to begin until the ideal co-founder happens to appear is, in Bakaus's view, a serious mistake.
"Waiting for the right person to knock on your door is a huge mistake. Just start."
Starting alone does not mean doing everything alone forever. A founder can later hire a Chief of Staff, an operations lead, or outside specialists. What matters is retaining responsibility for the company's direction and for the most important creative work.
Conclusion
The video's central message is that AI has dramatically lowered the barrier to making things, but it has not automated the creation of good work. AI can produce apps and code quickly. People must still decide what to make, judge which direction fits the brand, and remove what is unnecessary until the result becomes excellent.
Bakaus identifies several essential capabilities for the AI era:
- The judgment to decide what should be made
- The editorial ability to remove what is unnecessary from AI output
- The ability to communicate a clear product and design philosophy
- The ability to intervene at the right time when an agent begins moving in the wrong direction
- The ability to reduce operational work and protect time for the work the founder does best
- The ability to build a network that provides objective feedback, even when founding alone
In the end, the people most likely to build better products with AI are not those who produce the most. They are the ones who know what to choose and what to discard.