Inspiration
Orislop started the night I decided to kill my previous project, Currata. Currata was focused on regulatory updates for law firms, but I realized I was building something that did not feel urgent enough to me. After dropping it, I tried to relax by playing Roblox and scrolling online, but everything felt flooded with brainrot, spammy AI content, engagement bait, and low-effort videos designed to keep people watching without giving them anything valuable.
That made the problem feel obvious: the internet is getting filled with content that is not just low quality, but actively optimized to waste attention. A lot of tools focus on detecting whether something is AI-generated, but that is not the full issue. Some AI content can be useful, and some human-made content is still complete slop. I wanted to build something that judges content based on whether it is actually worth your time.
That became Orislop: a tool to detect online slop before it wastes your time.
What it does
Orislop is an early browser-based prototype that helps users detect low-quality online content before they get stuck watching it. It analyzes signals from videos and online content to estimate whether something looks like “slop,” including brainrot, spammy AI-generated clips, engagement bait, repetitive edits, and low-effort content.
The main idea is simple: before a user wastes time on a video, Orislop gives them a warning. Instead of just saying “AI” or “not AI,” Orislop gives a more useful recommendation, such as whether the content looks worth watching, questionable, or likely worth skipping.
A major part of Orislop is the temporal detector. Instead of judging a video from one frame, the temporal detector looks at how the video behaves over time. That matters because a lot of slop is not obvious from a single screenshot. The problem often shows up through repetition, pacing, low-effort visual structure, sudden pattern changes, AI artifacts, or the way a video is edited to trap attention.
The hackathon version is still a prototype, but it shows the core idea: use AI and scoring systems to help people defend their attention from junk content.
How we built it
I built Orislop as a solo project using a mix of AI engineering, video analysis, frontend development, and product design. The AI side uses a temporal video detection approach, where the system looks at video behavior across time instead of relying only on static image detection. I worked with Python, PyTorch, CUDA, OpenCV, NumPy, FFmpeg, Hugging Face, and model checkpoints to test and build the detection pipeline.
I also worked on multiple scoring signals beyond just the temporal detector. Orislop looks at patterns like clickbait language, repetitive captions, AI-style video structure, engagement-bait wording, low-information content, and other signals that commonly show up in slop. The goal was to combine multiple weak signals into a stronger overall judgment instead of depending on one perfect model.
For the prototype experience, I built the frontend using web technologies like JavaScript, TypeScript, HTML, CSS, Electron, and Node.js. The interface is designed to show a slop score, a recommendation, and reasons behind the decision so the user is not just given a random number with no explanation.
I also focused on making the project understandable as a product, not just a model. Orislop is not only an AI detector. It is meant to be a warning layer between the user and the content trying to steal their time.
Challenges we ran into
One of the biggest challenges was defining what “slop” actually means. It is easy to say that the internet is full of brainrot, but turning that into something a system can detect is much harder. Slop is not just AI-generated content. It can be repetitive, manipulative, low-effort, misleading, over-edited, or designed mainly to keep people scrolling.
Another challenge was working with video over time. A temporal detector is more complex than a normal image classifier because it has to understand motion, pacing, repetition, and changes across frames. A single frame might look normal, but the full video can still be low-quality or attention-baiting.
I also had to balance usefulness with accuracy. If Orislop flags too much, users will stop trusting it. If it flags too little, it does not solve the problem. Since this is still a prototype, the challenge was to build something honest and useful without pretending it is already a perfect filter for the entire internet.
Another challenge was scope. Orislop could become a browser extension, a desktop app, a website, a video detector, a feed filter, a feedback system, and a full attention-defense platform. For the hackathon, I had to focus on proving the core idea instead of building every possible feature at once, which is tragic because scope creep is apparently the default human condition.
Accomplishments that we're proud of
I am proud that I turned a real frustration into a working prototype. Orislop came from an actual moment where the internet felt unusable because everything was flooded with low-effort content. Instead of just complaining about it, I started building a tool to detect it.
I am also proud of building around temporal detection. Most simple AI detectors look at one image, one frame, or one piece of text. Orislop’s approach is more focused on how video content behaves over time, which fits the problem better because modern slop is often about pacing, repetition, and attention manipulation.
Another accomplishment is that Orislop is not just a technical experiment. It has a clear user problem: people do not want to quit the internet completely, but they do want help avoiding the junk that wastes their time. The prototype shows how AI can be used defensively, not just to generate more content.
I am also proud that I built this as a solo project. I had to think like an AI engineer, frontend developer, data analyst, product designer, and tester at the same time. That made the project harder, but it also forced me to understand the whole system instead of only one piece of it.
What we learned
I learned that detecting low-quality content is much harder than detecting AI-generated content. AI detection is only one part of the problem. The bigger question is whether the content is actually valuable or whether it is just optimized to waste attention.
I also learned that temporal analysis matters a lot for video. A single screenshot does not always show whether a video is brainrot or spam. Looking at the way the video changes over time gives more context about repetition, pacing, editing style, and structure.
I learned that product design matters just as much as the model. A detector is not useful if users do not understand it or trust it. Orislop needs to explain why something was flagged, give clear recommendations, and avoid being too aggressive.
Most importantly, I learned that the internet’s content problem is not just about bad content existing. It is about how platforms reward content that captures attention, even when that content has almost no value. Orislop is my attempt to build a tool that pushes back against that.
What's next for Orislop
The next step for Orislop is turning the hackathon prototype into a real MVP. The prototype proves the core idea: online videos can be analyzed over time and scored for patterns commonly found in low-quality slop. Now the focus is making that experience usable, reliable, and simple enough for real users.
For the MVP, I want Orislop to work as a browser tool or website where users can analyze YouTube videos and get a clear recommendation before watching. The basic flow would be simple: paste or open a video, run Orislop, receive a slop score, see the reasons, and decide whether to watch or skip.
I also want to improve the temporal detector so it can better recognize repetition, AI-generated video patterns, low-effort edits, and engagement-bait pacing. I plan to add stronger caption analysis, comment signals, user feedback, and better calibration so Orislop improves over time.
The long-term goal is to turn Orislop into a personal filter for the internet. Instead of blocking everything or forcing people to quit scrolling completely, Orislop would help users avoid the worst content and spend more time on things that actually matter.
Log in or sign up for Devpost to join the conversation.