Build a private, voice-first web app called Content Brain. Content Brain is a creative operating system for turning the ideas in my head into finished short-form video content I can record and edit myself. This is NOT a generic AI social media tool, content calendar, caption generator, or “viral hook generator.” The core problem: I often have a strong content idea in my head, but it comes out as a long, messy stream of consciousness. Inside that rant there may be: a great opening line several different stories mixed together factual claims that need proof personal anecdotes visual assets I could show screenshots / receipts / products / objects tangents that should become separate videos a larger point I’m trying to make language that sounds uniquely like me Today I often need another person — an editor, videographer, producer, etc. — to help turn that raw idea into something shootable. Content Brain should perform that editorial translation. The experience should feel like: I talk → Content Brain understands the idea → it identifies the actual story → it helps me strengthen it → it tells me what to record → I make the video. PRIMARY UX Make this a very clean, premium, minimal application. Desktop-first for now, but responsive. Dark interface. Serious creative-tool aesthetic. It should feel closer to a modern AI creative workspace than a social media management dashboard. Do not fill the homepage with cards, analytics, marketing copy, graphs, or fake features. When I open Content Brain, I should immediately be able to start creating. Main screen Primary screen contains: “What’s in your head?” Large microphone / voice input. Also allow typing. Allow drag-and-drop/upload for: images screenshots video clips PDFs/documents if useful reference material I should be able to ramble for several minutes without the system interrupting me or prematurely trying to structure the idea. During recording, show a live transcript subtly. Have a clear Finish Thought action. After I finish, transition into an analysis workspace. THE CONTENT BRAIN WORKSPACE The app should take the raw thought and intelligently turn it into a structured creative project. The most important behavior: Do not immediately rewrite my rant into a polished generic script. First understand what I’m actually trying to say. Generate these sections: 1. The Story Give me a concise interpretation of the strongest story inside what I said. Example: You’re not really making a video about expensive Counter-Strike stickers. The story is that you identified value in digital objects extremely early, held onto that conviction for 12 years, and now you’re connecting that instinct to what you’re building today. This should feel like an intelligent editor reflecting the idea back to me. Include: Core idea 1–3 sentences. Why this is interesting What creates tension, surprise, credibility, novelty, humor, etc. What the viewer should understand by the end 2. Story Separation Detect when I accidentally combined multiple videos into one rant. For example, my first Content Brain input might contain: “I won the Steam market” buying Katowice 2014 capsules very early owning rare Counter-Strike collectibles buying Ethereum below $7 screwing up crypto despite being early realizing that now I can build the future instead of only investing in it Content Brain should say something like: There are 3 potential stories here. Then show them as distinct ideas. Recommend one as the current story, but save the others into an Idea Bank so nothing gets lost. Do NOT generate 20 variations just to look productive. Strong editorial judgment > quantity. 3. The Hook Suggest a maximum of 3 genuinely different openings. Default to the one you believe best expresses the idea. Preserve my actual personality. If I say: “I think I won the Steam market.” do not transform it into: “Here’s how one decision changed my financial future forever…” That destroys the voice. The AI may tighten my language, but should protect unusual, provocative, funny, confident, profane, or personal phrases when they are the reason the content sounds like me. 4. Speaking Outline This is more important than writing a rigid script. Create a natural sequence of beats that I can use while talking directly to camera. Example: Opening “I think I won the Steam market.” Receipt Explain that I bought Katowice 2014 sticker capsules when they were extremely cheap. Context Explain in plain English what those were to someone who has never played Counter-Strike. The insight Why I bought them / what I saw at the time. The outcome What happened over the following 12 years. Bigger point What this taught me about digital ownership / markets / recognizing emerging behavior. Optional bridge “I also bought Ethereum under $7 and somehow still managed to fuck up crypto. That’s another story.” Give me the option to toggle: Outline / Full Script The Full Script should still sound spoken rather than written. 5. Proof / Claims / Receipts This is a major Content Brain feature. While analyzing the story, detect claims that would be substantially stronger with evidence. Create a panel called: Receipts Separate information into: Provided Things I have already uploaded or established. Needed Things that would materially strengthen the video. Verify Claims that should not be presented as objective fact yet. Example: If I say: “I owned the most valuable digital object in history.” Content Brain should flag that as a strong global superlative that requires verification. It can suggest: This may be better framed as a personal/provocative claim unless we can establish exactly what is being compared. This should NOT make the system timid. The goal is: Maximum confidence supported by maximum credibility. 6. Visual Plan For every story beat, tell me what the viewer should SEE. Create a simple timeline/table: MomentWhat I’m sayingWhat viewer seesHook“I think I won the Steam market.”Me talking directly to cameraReceiptBought capsules for $0.25Zoom into purchase historyObjectExplain early CS collectiblesPhysical collectible in my handValueExplain what happened laterRelevant marketplace/inventory visualMeaningBigger takeawayBack to uninterrupted camera The system should understand that the video does NOT need constant B-roll. Sometimes the most powerful visual is simply leaving me on camera. 7. Capture List Turn the plan into a literal checklist for me. Example: Record direct-to-camera main take close-up holding original Counter-Strike collectible optional second take of opening line Capture Steam purchase history showing original capsule purchases inventory screenshot relevant historical price/value reference Already Have [uploaded assets] This is what makes Content Brain operational instead of merely inspirational. 8. Edit Plan Create an edit blueprint designed for a person doing basic editing in tools such as CapCut. It should say things like: 0:00–0:03 Stay on face. No intro animation. 0:03–0:07 Cut to receipt while the audio continues underneath. 0:07–0:15 Return to camera. On-screen text Only use text when it materially helps comprehension. Suggest: cuts punch-ins pauses captions screenshots overlays where to leave the shot alone Avoid hyperactive generic TikTok editing. The visual style should serve the story. 9. Record Mode Create a simple Record This mode. When clicked, remove most of the UI. Display: current story beat optional opening wording the next beat large recording timer controls to move forward/back between beats This should function like a lightweight producer sitting next to the camera. I should be able to glance at it without reading a teleprompter. 10. Uploaded Take Review Build the interface for this even if the first implementation uses mocked analysis. After filming, let me upload the recorded video. Content Brain should analyze the take and eventually be able to identify things such as: strongest possible opening occurs later in the take repeated explanation section that drags sentence worth cutting missing evidence confusing claim visual opportunity strong natural line worth preserving Show feedback on a timeline. The architecture should anticipate real transcription/timecodes later. For the initial implementation, this can use AI analysis of transcript/video if available or a clearly working prototype. Do not fake precision if we do not have the underlying data. IDEA BANK / CONTENT MEMORY Include a lightweight Idea Bank. Every discarded tangent or secondary concept can become its own idea. Ideas should show: title raw thought current angle status related project/story available assets missing receipts Statuses: Idea Developing Ready to Record Recorded Editing Published Do not turn this into a giant Kanban/project-management interface. The priority remains creating content, not managing content. BRAIN / MEMORY Content Brain should gradually learn: how I naturally speak phrases I use subjects I repeatedly return to stories I’ve already told claims I’ve established assets I have available projects/products I’m working on published videos previous versions of an idea Create a simple Brain view where this accumulated context can be inspected. Important distinction: Something I said is not automatically verified fact. Represent memory intelligently using categories like: Personal story Claim Verified receipt Asset Topic Project Published content Allow editing/removing incorrect memories. FIRST CANONICAL TEST PROJECT Seed/demo the interface with this example so we can test whether the product works: Raw Thought “I want to make a video explaining that I basically won the Steam market. I was extremely early to Counter-Strike skins and collectibles. Around Katowice 2014 I bought sticker capsules extremely cheaply, including some for around 25 cents, and I’ve held some of this stuff for roughly 12 years. I also have physical early Counter-Strike collectible pins that I can show. I think there’s a larger story here about how early I was to understanding digital objects having real value. I was also extremely early to crypto — I have receipts of buying Ethereum under $7 — although I also majorly fucked up crypto in other ways. And now the interesting thing is that with AI I don’t just have to identify where things are heading or invest in them. I can actually build the things I believe should exist.” The expected Content Brain behavior is NOT to simply output a giant script containing all of this. It should recognize something close to: Video 1 I Won the Steam Market Early Counter-Strike / Katowice story. Video 2 I Bought Ethereum Under $7 and Still Fucked Up Crypto Being early ≠ executing perfectly. Video 3 This Time I’m Building The connection between seeing trends early and having AI tools that let me create directly. Focus the active project on Video 1. DESIGN The product should feel extremely polished. References: creative AI workspace premium editing software restrained typography dark neutral palette generous negative space large content surfaces very few borders subtle motion excellent transitions between thinking states No gradient-heavy AI aesthetic. No cartoon brain imagery. No purple glowing AI orb. No generic “AI assistant” avatar. The intelligence should be expressed through the quality of the output. Use the name: Content Brain Potential small descriptor: Turn what’s in your head into something worth publishing. Do not over-brand the interface yet. TECHNICAL DIRECTION Use Next.js / React and build this as a real functioning application. Use persistent storage for: ideas projects raw thoughts/transcripts generated analysis assets story beats receipts memory uploaded takes Build the information architecture properly instead of hardcoding everything into one demo screen. Use AI structured output so the model can return predictable objects for: story interpretation story candidates hooks speaking beats claims evidence requests visual beats capture checklist editing instructions Keep the AI orchestration modular so we can improve individual stages independently later. Do not make every stage a separate slow model call if it can be generated intelligently together. Time-to-result matters. MOST IMPORTANT PRODUCT PRINCIPLES Listen before writing. Find the story instead of merely formatting the rant. Protect the creator’s voice. Separate tangents instead of deleting them. Distinguish claims from evidence. Tell the creator what to physically record. Make the output executable. Do not confuse more AI output with better creative judgment. Do not chase a fake universal formula for virality. The goal is to reduce the distance between an idea in my head and something I actually publish. Build this as a working first version that I can immediately use to create the Counter-Strike video.