If Anyone Can Build It, What Does a Portfolio Prove?

Two weeks. Eight projects. You stuff them one by one into your portfolio page, each with a screenshot, each clickable and runnable. Landing pages, AI note-taking tools, data dashboards, a small game—your tech stack is packed full. You look at this lineup and feel a sense of accomplishment, thinking this should finally get someone’s attention.

You send out over thirty applications. Two respond, both with generic rejection templates. The rest are either left on read, or not even read at all.

You’re not convinced. You go look at portfolios of peers who received offers, trying to see what makes theirs better. After checking, you’re even more confused—their projects look almost identical to yours. Same card layouts, same hover effects, same clean color schemes, even the same project types: landing pages, AI tools, dashboards. You even think some of yours are more detailed.

So why them, and not you?

You stare at the screen for a while, and an even more chilling thought hits you: these projects—anyone could build 8 in two weeks. You could. And someone who doesn’t know a single line of code could open Lovable, chat with it for one evening, and deliver 8 projects by tomorrow.

If anyone can build it, what does building it actually prove anymore?

Your Projects Are “Good Enough”

The problem isn’t that they’re not good enough; it’s that they’re too good. So good that anyone could make them.

In early 2025, Andrej Karpathy tweeted about “vibe coding”—writing software by chatting with AI based on intuition. Back then, people treated it as a joke. By November that year, Collins Dictionary named “vibe coding” its word of the year. In less than a year, the wall between describing an app and owning one collapsed. A person with zero coding skills could open Lovable or Bolt, describe in plain language “I want a to-do app with login, payments, and a backend,” and within minutes see a runnable, deployable, shareable product.

Many people haven’t fully grasped what this means for portfolios.

The entire logic of a portfolio rests on an implicit premise: creating a decent product has a barrier to entry, so the fact that you can do it proves something. A designer’s Dribbble, a developer’s GitHub, an indie maker’s demo list—they all filter for this: do you have the craft to turn what’s in your mind into something usable on screen?

This premise barely holds by 2026.

AI has flattened product quality to an average baseline. A landing page with clean design, smooth animations, and pleasant colors used to indicate taste and skill; now it’s the default output from tools like v0. A full-stack demo that handles login, registration, and a database used to show engineering chops; now it’s an afternoon’s work with Cursor. When everyone can consistently produce B-level work, a B grade no longer distinguishes anyone. It degrades from a signal into background noise.

Worse, this averaging has a distinct flavor, like AI-generated résumés. In June 2025, The New York Times interviewed hiring managers who said incoming résumés looked suspiciously similar—so many that they couldn’t tell who was truly qualified or truly interested. A Resume-Now survey of 925 HR professionals in March that year found that 62% of employers would outright reject a résumé they suspected was AI-generated. The rejection wasn’t for poor quality, but for being too much like everyone else’s.

Portfolios are treading the same path résumés did. When ten thousand people use the same tools and default templates to produce work, the marginal information in each piece approaches zero. A recruiter flips through your portfolio and starts swiping past by the third project because they know the next five probably look the same—and they do.

By cramming eight projects into two weeks, you thought you were adding weight; in reality, you’re diluting your portfolio. Each average project drags down your overall distinctiveness. Recruiters aren’t ignoring your effort—they’ve already assumed AI could handle that part for you. Their attention has quietly shifted elsewhere.

Five Centuries of Portfolios Have Always Filtered for the Same Thing

To understand where their attention has shifted, we first need to understand what portfolios have always been doing.

The word “portfolio” comes from the Italian portafoglio—portare meaning “to carry” and foglio meaning “paper”—a folder for carrying papers. The first to use them were painters and architects. When meeting a patron, an artist couldn’t bring their entire studio; they carried a sheaf of drawings to lay on the table. The patron flipping through wasn’t just checking if you could draw hands or handle lighting—they were looking for something in your work that resonated with what they wanted.

From the start, portfolios filtered not for whether you could do something, but for how it was done. In the Renaissance, being able to draw hands was the entry-level bar; patrons assumed you could. What they compared was everything above that baseline.

This pattern repeated over the next five hundred years: every time technology lowered a barrier, the filtering criteria moved up a notch.

Before photography, accurately depicting a person’s likeness was a scarce skill—portrait painters made a living from it. Once cameras became common, painting realistically instantly lost its value. Painters were pushed upward, forced to capture what photos couldn’t: Impressionism, Expressionism—these movements emerged from being squeezed out. When machines flattened the ability barrier, people had to retreat to where machines couldn’t reach and reestablish their footing.

The software industry replayed the same story with a different skin. In the early days, being able to write HTML and make a webpage work was a skill in itself. Then frameworks, templates, and no-code tools stacked up, steadily lowering the barrier to building a working website. So GitHub took off—just having a product wasn’t enough; people needed to see what your code looked like and if your commit history was healthy. Dribbble and Behance flourished—just having work wasn’t enough; you had to show a distinct visual language. Each time, platforms helped recruiters push the filtering criteria one level higher.

Interestingly, until AI arrived, this system increasingly leaned toward valuing finished products. In Stack Overflow’s 2024 developer survey, 73% of hiring managers said a strong portfolio mattered more than a perfect résumé for developer roles. That number was correct at the time—it assumed building a strong portfolio still had a barrier to entry, so portfolios could still function as filters.

That statement’s expiration date was the day “vibe coding” became word of the year. When machines flatten a barrier, humans are forced to retreat beyond the machines’ reach. Cameras flattened realism; painters retreated into painting with ideas. AI now flattens the creation of functional products. For portfolios to remain useful, what they filter must move up another level—to where AI still can’t reach.

Six Places AI Can’t Reach

Where is that? It’s in the things AI can’t generate for you. These share a common trait: they’re not the finished products, but the traces behind them. AI can fabricate an entire portfolio in an afternoon, but it can’t fabricate your twelve months of public thinking.

First, a record of judgment and trade-offs.

The most valuable part of a project is often not what it includes, but what it excludes. You build a to-do app and cut the collaboration feature—why? You explain: I evaluated it and found my target users are heavy individual users; adding collaboration would triple the data model complexity, and I observed their real pain point was having too many tasks to manage. So I spent those two weeks on smart sorting. AI can’t write this because it doesn’t know what you observed, what you agonized over, or which side you ultimately bet on. It’s like watching a chef: anyone can taste whether a dish is good, but only by hearing that they originally planned to use butter, tried three times and found it masked the Sichuan pepper’s aroma, and finally switched to light oil, do you know they’re a chef with judgment, not just someone following a recipe. The dish is the recipe’s product; judgment is the chef’s product.

Second, the choice of problem.

AI can solve any problem you throw at it, but it won’t choose the problem for you. Which projects you include in your portfolio exposes the boundaries of your taste. Ten people build an AI chat wrapper; the eleventh builds a tool that helps hearing-impaired friends transcribe meetings in real-time and label who’s speaking. Even if the latter is rougher, the signal is stronger because choosing that problem required you to truly see a specific person and a specific struggle. The source of problems can’t be scaled—it comes from your life, and only you have your life.

Third, process traces.

Code can be AI-generated, but commit history is hard to fake throughout. A real project on GitHub shows sporadic commits, late-night hotfixes, “fix typo,” “revert last revert,” and a README that’s been revised multiple times. These messy traces are proof that a real person did the work. You can also include key discarded drafts, rejected versions, or even the prompts you used to wrestle with AI. Discarded drafts are the best anti-AI verification because AI only gives you the final version—it won’t proactively leave the corpse of “I tried this path and it failed.” Displaying those corpses proves you walked that path.

Fourth, consistency of taste.

A single beautiful product isn’t valuable, but a consistent preference threaded through a series of works is. If all your projects reflect a leaning toward minimalism, a dislike of pop-ups, and high information density, that consistency coalesces into a persona. AI’s outputs are independent and memoryless each time; statistically, they’re the mean. Your biases, however, are directional deviations. Recruiters aren’t hiring a tool that produces averages—they could just buy an AI subscription for that. They’re hiring a person with a stable direction.

Fifth, delivery under real constraints.

“I built a working first version in three weekends, with zero budget, in a domain I was completely unfamiliar with.” The constraints here are the signal. Constraints are what AI can’t experience. It doesn’t know you only had three days, that you couldn’t afford API credits so you built your own caching layer, or how you coped when a client changed requirements midway. Clearly stating the constraints re-evaluates the product’s worth.

Sixth, verifiable impact.

A single number, if true and checkable, is worth more than ten screenshots. “Live for four months, 387 real users, 41% weekly retention,” or “This recap was retweeted 600 times on X, with two frontline engineers debating thirty comments below.” Verifiable means others can click to confirm. AI can fabricate pretty numbers, but it can’t fabricate a link that actually opens and has real people commenting underneath.

These six layers are not about making the finished product prettier. They’re about making the decision-maker behind the product visible.

You’re Comparing Craft; They’re Comparing Credibility

Here’s the misalignment—and the real reason most people’s applications go unanswered.

Job seekers stand on their side, thinking it’s a competition of product quality, so they pour all their energy into making demos smoother, screenshots prettier, and projects more numerous. They imagine recruiters will be wowed by their polished output.

Recruiters stand on the other side, seeing a completely different game. They review dozens, sometimes hundreds, of portfolios daily, carrying a pre-installed default assumption: these products were likely built with AI. So when they look at your work, they’re not evaluating whether you could make this—they’ve already answered that for you: yes, anyone could. They’re looking for something else: Is this person reliable? Do they have judgment? Did they actually do the work?

The two sides aren’t comparing the same thing. You’re comparing craft; they’re comparing credibility.

This misalignment has been magnified by the flood of AI content. The 2025 tsunami of AI-generated résumés forced recruiters to use AI tools to filter AI résumés—what Mashable called an ouroboros, a snake eating its own tail. In that environment, something counterintuitive happened: the more the world fills with AI-generated polished content, the higher the price of verifiable, human-traced authenticity.

A product you can click into, with 387 real users, with commenters complaining about bugs, is more credible than ten perfect but untraceable demos. Built In’s article on developer hiring in the AI era put it directly: a candidate with moderate coding ability but strong system design intuition and explainability is often more valuable in many roles than a pure coder. Because companies already have AI for pure coding; what they lack are people who can judge. The design world says the same thing. A design recruiter with over a decade of experience once wrote something to the effect of: your portfolio is a performance; your public work is the evidence. AI can perform a beautiful portfolio in an afternoon, but it can’t perform the authentic thinking you’ve left in public over the past year.

Of course, this isn’t absolute. There’s still a category of work where only the finished product matters: outsourcing, rapid delivery, pure execution projects. Clients just want a working landing page by tonight; they don’t care how you think—run, look good, be cheap, be fast. If that’s your line of work, stacking volume, stacking products, stacking speed still holds; AI even makes you more competitive.

But if you’re after good positions, good clients, roles where others are willing to hand you decision-making power, the game has changed. At that table, the finished product is the entry ticket, not the chips. Everyone has an entry ticket. The real bet is whether, once inside, you can convince the other side that you’re irreplaceable.

Stop Adding Projects

Back to you—the one who built eight projects in two weeks and heard nothing back.

Your likely impulse right now is to build a few more, make them prettier. Stop. The issue was never quantity; building eight more just adds eight drops to the same average pool. What you need isn’t more projects—it’s to add a judgment layer to the ones you have.

Pick the three you’d least want to delete from your eight; remove the rest. Fewer and truer beats more and average. Then, for each remaining project, answer four questions and write the answers in.

Why did I build this instead of something else? If the answer is “a tutorial told me to” or “I saw someone else’s went viral,” that project probably should go too. Keep those where you can name a concrete reason: a specific person you encountered, something you genuinely would use.

What did I cut, and why? Write about the features you dropped, the approaches you rejected, the detours you took—especially detours. “I used A in the first version, but after launch, I noticed users weren’t clicking it at all. Three weeks later, I scrapped it and rebuilt with B.” Such statements are worth more than any success description because they prove you’re making real judgments, not performing a one-shot masterpiece.

Is there a real, verifiable number or trace? If you have real users, write the user count; if there’s public discussion, post the link; if there’s commit history, let people browse it. Even if the numbers are small—“23 users, 4 still active”—it’s worth a million times more than “industry-leading experience,” because the former can be clicked and verified, while the latter is obviously scripted.

Can these three projects, together, reveal what kind of person I am? If not, your taste hasn’t surfaced yet. Think about the recurring preference in your work: do you always trim features, always fight pop-ups, always want to make responses faster? Name it. Make the person viewing your portfolio, by the end of the third project, picture a person with direction, not an average-output machine.

After these four steps, do the most anti-AI thing: start writing publicly. It doesn’t have to be long or polished. Continuously write and share your real struggles, pitfalls, and moments of changing your mind while working on these projects. AI can mimic a portfolio in an afternoon; it can’t mimic your public traces over a year. A year later, those traces themselves will become the hardest part of your portfolio—so hard that no person, no model, can replicate them.

The finished product is now the entry ticket. Everyone has one.

The chips you can truly bet on are the you who makes decisions. Your eight projects got no response not because they weren’t pretty enough, but because they were pretty in the same way as everyone else’s, while the only different thing about you—you—was hidden behind your screenshots.

Bring him to the front.