DEV Community: Igor Defari

Movie Discovery Is Broken

Igor Defari — Sat, 13 Jun 2026 08:32:14 +0000

Movie Discovery Is Broken — and Streaming Services Are Doing It on Purpose

You're not bad at choosing. The system is designed to make you struggle.

It's a Tuesday night. You've finished dinner. You sit down, open Netflix, and start browsing. Forty minutes later, you've watched three trailers, scrolled past a hundred thumbnails, opened something, watched four minutes, closed it, and settled on rewatching a show you've already seen twice.

You didn't choose badly. You were never really given a choice.

This is the central paradox of the modern streaming era: we have access to more films than any human being in history, and we spend more time not watching them than any generation before us. Research shows that viewers spend nearly five days a year just deciding what to watch — time that could be spent actually enjoying content. The average user now spends 18 to 25 minutes per session browsing for something to watch, with more than half admitting that search time often exceeds the length of whatever they finally pick.

That's not a coincidence. It's a design choice.

The Algorithm Was Never Built for You

Here's what the streaming platforms don't tell you about their recommendation engines: they're not optimized to help you find the best film for your mood tonight. They're optimized to keep you on the platform as long as possible.

Those are related goals, but they're not the same goal — and the difference matters enormously.

An algorithm that maximizes engagement will surface what's trending, what's new, what has the highest completion rate across the user base, and what the studio paid to promote this week. It will learn your history and use it to funnel you toward more of what you already know you like — which feels personal but is actually the opposite of discovery. The majority of viewers believe their recommendations are tailored specifically for them, but under the hood, algorithms are designed to push engagement quotas, spotlight trending titles, and keep users clicking. You're not picking — you're being picked for.

The result is a kind of invisible conservatism built into every recommendation you receive. The algorithm knows you watched a thriller last weekend, so it shows you another thriller. You watch that one, so it shows you another. Over time, your recommendations narrow into a corridor — personalized in theory, limiting in practice.

Streaming's promise of personalization is seductive, but underneath, the algorithms prioritize engagement, not genuine exploration — producing a tightly controlled buffet of safe bets designed to keep you watching, not to broaden your cinematic horizons.

Too Many Services, No Single Map

The problem isn't just algorithmic. It's structural.

The average U.S. viewer now subscribes to four streaming services, each offering thousands of options — and decision time averages 10 to 20 minutes, while drop-off rates are climbing every year. Drop-off means giving up entirely — closing the app and doing something else rather than watching anything at all.

But the fragmentation goes deeper than the number of services. Studios are experimenting with hybrid distribution — licensing the same title to subscription services in one window, ad-supported platforms in another, and rental in a third. For viewers, this means decision fatigue. For platforms, it means missed engagement.

The film you want to watch might be on Netflix in your country and not available at all in someone else's. It might have been on Amazon Prime last month and rotated off. It might be available to rent but not to stream. The map of what's available where changes constantly, and no single platform has any incentive to show you where the content you want lives if it lives somewhere other than their service.

The companies that win in this environment won't be the ones with the biggest libraries — they'll be the ones with the best maps. Right now, nobody has the map.

The Hidden Gems Problem

There's a specific failure mode of the current discovery system that film lovers feel most acutely: the invisibility of everything that isn't new.

Streaming platforms are built around freshness. The homepage surfaces what just arrived. The trending section shows what everyone is watching this week. The algorithm learns from recent behavior and recommends recent content. The entire architecture of the platform points toward the present tense.

Which means the vast majority of what's available — the catalog, the back library, the decades of extraordinary cinema that predates the streaming era — is functionally invisible to most users. Not because it isn't there. Because the interface is designed to point elsewhere.

A viewer who doesn't know to search specifically for a 1973 French thriller will never be shown one. The algorithm has no mechanism for surfacing the film you didn't know you needed. It can only show you more of what it already knows you like.

Apps like IMDb, JustWatch, and Letterboxd have tried to help users navigate the chaos, but their models fall short: IMDb is popularity-driven, JustWatch is availability-focused, and Letterboxd is built around social reviews. Useful, yes — but none of them actually solve the problem. They offer more data, not better decisions.

This is the specific problem *MovieHunt** was built to solve — movie discovery that filters by mood, genre, decade, and hidden gem status, with a Spin Wheel for when you've filtered everything down and still can't commit. Free to browse, no account required.*

The Psychology Behind the Paralysis

The discovery problem isn't just a product design failure. It's a psychological one.

Choice overload is a cognitive phenomenon referring to the difficulty of making a decision when presented with numerous options — a complexity that leads to decision fatigue, dissatisfaction, or avoidance. Barry Schwartz called it the Paradox of Choice: beyond a certain threshold, more options don't increase satisfaction, they decrease it. The more you can choose from, the harder each choice becomes, and the less happy you are with whatever you eventually pick — because the awareness of everything you didn't choose lingers.

The psychological toll of this process is real: frustration, regret, and less enjoyment — even after you finally choose.

Streaming platforms have, possibly without intending to, created the worst possible conditions for enjoyment. They've maximized optionality — the number of things available to watch — while minimizing the tools for navigating it. The result is a product experience that feels abundant but functions as a cage.

The irony is that reducing choice actually increases satisfaction. Studies consistently find that curated selections — even arbitrary ones — produce more enjoyment than unlimited browsing. A single recommendation from a trusted friend beats an algorithm every time. A filtered list of ten films that match your mood tonight beats a library of fifty thousand.

What Good Discovery Would Actually Look Like

The solution isn't complicated in concept, even if it's been surprisingly difficult to execute in practice.

Good movie discovery starts with context, not catalog. What are you in the mood for right now — not what you watched last weekend, not what's trending, not what the studio paid to promote. What do you want to feel for the next two hours?

That question has dimensions that current platforms don't even ask. Mood. Energy level. How much attention you want to pay. Whether you want something familiar or challenging. Whether you want to see something recent or are open to any decade. Whether you want something with a short runtime or are ready for an epic.

AI-powered discovery in 2025 is moving toward context-awareness — systems that can consider factors like: "Give me a feel-good comedy under 90 minutes" or "suggest something like Prisoners" — blending human-like conversation with data-driven accuracy to actually reduce fatigue through small, guided decisions.

The other dimension good discovery addresses is serendipity. The films that stay with you for years are rarely the ones an algorithm recommended. They're the ones you stumbled onto — a recommendation from someone who knows you, a film festival lineup, a film studies course, a late-night channel-surfing accident. Discovery tools should create the conditions for that kind of accident. Not just show you what you already know you like.

Why the Platforms Won't Fix This

Here's the uncomfortable structural reality: the major streaming platforms have limited incentive to solve the discovery problem in the way viewers would want it solved.

A discovery tool that genuinely helps you find the best film for tonight — including films on other platforms — works against the business model. Platforms are competing for your subscription, not cooperating to serve your taste. JustWatch does the cross-platform availability search because it's a neutral aggregator with no stake in which platform you watch on. Netflix will never build a feature that tells you the film you want is on Amazon.

Beyond the competitive dynamics, there's a subtler misalignment. Platforms benefit from you spending time browsing, even if you don't watch anything. Time in app is time in app. The engagement numbers look better. The churn risk decreases. A user who spends 40 minutes browsing before watching something is, from a data perspective, a highly engaged user — even if from a human perspective they're frustrated and wasting their evening.

The incentives point toward more content, more recommendations, more thumbnails, more autoplay — not toward the simple, high-quality, mood-filtered list of five films that would actually make your Tuesday night better.

That misalignment is what *MovieHunt** exists to address — a discovery tool built by a film lover, not an engagement team. Mood filters, hidden gems, decade browsing, and a Spin Wheel for the genuinely indecisive. Built for people who love movies and hate choosing.*

The State of Play in 2026

The discovery problem is not getting better on its own. The number of streaming services is not decreasing. The volume of content being produced is not slowing down. The algorithmic incentives are not changing.

What is changing is the emergence of independent discovery tools — smaller, more opinionated products built by people who care about cinema and want to solve the problem the platforms won't. Products that start from the viewer's experience rather than the platform's metrics. Products that ask what you want to feel rather than what you watched last month.

The companies that will win the discovery space won't be the ones with the biggest libraries — they'll be the ones with the best maps, the movie metadata and availability intelligence to help viewers actually find what they're looking for.

The era of passive, algorithm-driven discovery is already starting to feel inadequate to the people who take film seriously. The question is whether the tools being built to replace it will reach them.

If you're done letting an algorithm decide what you watch, *MovieHunt** is worth trying. Filter by mood, genre, decade, or hidden gem status. Let the Spin Wheel decide when you can't. Built for the film lover who wants to find something good — not just something available.*

How I Built a Movie Website With No Coding Skills Using Bolt.new image

Igor Defari — Tue, 09 Jun 2026 09:38:09 +0000

I Built a Movie Discovery Website Without Knowing How to Code — Here’s Exactly How I Did It
No developer. No agency. No coding background. Just an idea, a tool called Bolt.new, and a lot of late nights.

A year ago, if you’d told me I would build and launch a fully functional movie discovery website — with user authentication, a subscription payment system, a Chrome extension, a newsletter integration, and a spin wheel feature for when you can’t decide what to watch — I would have assumed you were describing something that would cost thousands of dollars and require hiring a developer.

Instead, I built it myself. From Ukraine. Without writing a single line of code by hand.

The website is called MovieHunt — a movie discovery app built for people who spend more time choosing what to watch than actually watching. And the tool that made it possible is Bolt.new.

This is the honest story of how that happened, what worked, what didn’t, and what I’d do differently if I started over today.

What Is Bolt.new and Why Does It Matter

Bolt.new is an AI-powered full-stack web development platform created by StackBlitz. Unlike traditional AI coding assistants that generate code snippets for a developer to paste into their own environment, Bolt.new builds, runs, and deploys entire web applications directly in your browser — from a simple text prompt.

You describe what you want to build, and Bolt.new doesn’t just give you code to copy and paste. It actually creates a working application that runs in real-time, complete with a live preview, editable code, and the ability to deploy instantly.

Bolt works by converting natural language prompts into production-ready, full-stack applications. It requires no local setups or manual installs.

The platform runs on what StackBlitz calls WebContainers — technology that lets code that usually needs your computer run safely inside your web browser. In simple terms, they allow Bolt.new to install software packages, run development servers, and show a live website preview without extra setup — like a virtual workshop inside your browser where you can build and test your site without downloading anything.

For someone who doesn’t code, this matters enormously. There’s nothing to install. Nothing to configure. You open a browser tab and start describing what you want to build.

The Idea: A Movie Discovery App for Indecisive People
The concept behind MovieHunt was simple and personal. I watch a lot of films. I write about film. I have strong opinions about what makes a movie worth watching. And I spend an embarrassing amount of time staring at streaming service homepages unable to pick something.

The existing tools — letterboxd for tracking, TMDB for data, Rotten Tomatoes for ratings — all do their jobs. But none of them solved the specific problem of being in the mood to watch something right now, without knowing what. The decision fatigue is real, and it’s one of the most commonly complained-about experiences in modern streaming culture.

So I wanted to build something that filtered recommendations by mood, genre, decade, and streaming availability — and included a Spin Wheel feature for the moments when you genuinely cannot decide and just need the algorithm to pick for you.

What I didn’t want was to spend months learning to code or thousands of dollars hiring someone else to build it. That’s where Bolt.new came in.

**How the Build Actually Worked

**The process is less magical and more iterative than most “I built an app with AI” articles suggest. Here’s what it actually looked like.

Step 1: The first prompt. You start by describing what you want in plain language. Not technical specifications — just the idea. Something like: “Build a movie discovery web app that lets users filter films by genre, mood, and decade. It should have a clean, dark cinematic design, user accounts, and a subscription model.”

Bolt will quickly generate a basic structure covering the essential pages most sites need, providing a solid foundation to build on. What comes back isn’t a finished product — it’s a starting point. A real starting point, with actual pages, actual navigation, and actual code running behind it.

Step 2: Iteration through conversation. This is where the real work happens. You look at what Bolt built, identify what’s wrong or missing, and describe the next change in plain language. “Make the background darker.” “Add a loading spinner when the movie results appear.” “The search bar isn’t centered on mobile — fix it.”

The system excels at framework generation, UI component creation, and simple functionality implementation through prompt refinement. Each iteration gets you closer to what you actually want. The key skill — and it is a skill — is learning to describe what you’re seeing and what you want changed with enough precision that the AI understands.

Step 3: Integrating the real tools. A movie discovery app needs movie data. I connected MovieHunt to the TMDB API for film information and poster images. I integrated Auth0 for user authentication — so people can create accounts and save their watchlists. I connected Stripe for subscription payments, with a monthly tier and an annual tier. I added Beehiiv for newsletter sign-ups, and Google Analytics to understand how people were using the site.

None of these integrations required me to understand the underlying code. They required me to understand what I wanted to connect and to describe it clearly to Bolt.

Step 4: The features that made it distinct. The SEO Boost feature came later — a set of structured metadata and schema markup that helps MovieHunt’s film pages rank in Google for specific movie-related queries. Again: described in plain language, built by the AI, tested by me.

Press enter or click to view image in full size

*The Honest Challenges
*
I’d be doing you a disservice if I made this sound frictionless. It wasn’t.

Token limits are real. Bolt uses a token-based economy that affects both functionality and cost — larger applications need more tokens for each interaction, and paid plans begin at $20 per month. For a project the size of MovieHunt, with authentication, payments, multiple API integrations, and a Chrome extension, the token consumption adds up. Budget for it.

Complex bugs require patience. Users have burned through large numbers of tokens just to fix bugs. When something breaks in a way that isn’t visually obvious — a payment flow that fails silently, an API call that returns the wrong data — the debugging process through conversation is slower than it would be for someone who can read the code directly. You describe the symptom. The AI suggests a fix. You test it. Sometimes it works first try. Sometimes it doesn’t and you start again.

Prompt precision is everything. Bolt is fast, intuitive, and surprisingly capable, although not without a steep learning curve, especially when it comes to prompt precision. Vague prompts produce vague results. “Make it look better” produces something different from “Increase the contrast on the movie card titles, reduce the card padding to 12px, and make the hover state slightly lighter.” Learning to be specific — visually, functionally, structurally — is the actual skill this tool teaches you.

You still need to understand what you’re building. Bolt handles the code. It doesn’t handle the product decisions. What features matter? In what order do you build them? What does the user experience feel like? What problem are you actually solving? Those questions are still entirely yours. The tool removes the technical barrier. It doesn’t remove the thinking.

If you want to see what the finished product looks like — the movie discovery features, the subscription tiers — MovieHunt is live here. Free to browse, subscription for full access.

What I’d Do Differently

A few things I learned the hard way that would have saved me significant time and frustration:

Start with a detailed spec, not just a vague idea. The more specific your initial prompt, the better the foundation Bolt gives you. Spend time before you open Bolt writing out exactly what you want: the pages, the features, the user flows, the visual style. The initial build sets the architecture, and changing architecture later is expensive in tokens and time.

Handle legal and privacy infrastructure early. I set up the Privacy Policy and Terms of Service for MovieHunt through Termly relatively late in the process. Covering GDPR, COPPA, and disclosures for all third-party tools — Google Analytics, Stripe, Beehiiv, Auth0, TMDB — should have been one of the first things I did, not one of the last. If you’re building anything with user data or payments, do this first.

Test on mobile constantly. Bolt generates responsive code, but responsive code and a good mobile experience aren’t the same thing. Check every new feature on an actual phone, not just the browser’s responsive preview. The issues I found on mobile late in the build were more painful to fix than they would have been if I’d caught them earlier.

Use the version history. Bolt saves previous versions of your project. Use them. When an iteration goes badly wrong — and some will — being able to roll back to where things were working is invaluable.

The Bigger Point

MovieHunt took months to build properly. It wasn’t a weekend project. But it was built by one person, without coding knowledge, from a country with limited access to the major payment processors and development ecosystems that make this easier in other parts of the world.

That’s the real story of Bolt.new — not that it makes building software trivial, but that it genuinely moves the barrier. The thing standing between you and a functional web product is no longer technical knowledge. It’s clarity of vision, precision of communication, and willingness to iterate.

Building even a basic web application traditionally means hiring a developer or agency, waiting weeks, and spending a significant chunk of a small team’s budget. Bolt.new aims to put that power directly in the hands of founders, marketers, and product managers — turning what used to take weeks into a matter of hours.

For people with ideas and without technical backgrounds — creators, entrepreneurs, educators, writers — that shift is enormous. It was for me.

What MovieHunt Is Now

MovieHunt is a movie discovery web app built for people who love film and hate decision fatigue. It has:

Mood and genre filtering — narrow down by what you’re actually in the mood for, not just broad categories
Decade filtering — because sometimes you specifically want something from the 80s and nothing else will do
Watchlist — save films for later, track what you’ve seen
A Chrome extension — get movie recommendations without leaving your browser (it’s gonna be later)
A newsletter — film picks, hidden gems, and recommendations by email
The subscription tiers are $4.99 per month. Free browsing is available without an account.

If you’re a film person — and given that you’re reading this, you probably are — it’s worth a look.