Improve the Content Idea Engine

Open the Idea Engine in this folder. We are improving it, not rebuilding it. Keep everything that works.

The ideas cite sources that barely relate to them. Fix that.

Clean the sources first. Before the idea writer sees anything, drop posts that are not in English, older than 90 days, or without an outlier badge. Then use one quick Claude call to check each remaining post, thread and comment against my DNA and drop anything that is not really about my niche. Show me how many were kept and dropped.

Then break down every source that survives. For each one, have Claude write a short breakdown: the hook that stops people, the format, the triggers it pulls, why it performed, and what is worth borrowing. Do it in one batch call, save each breakdown so a post is only analyzed once, and show it when I click a source. The writer gets these breakdowns, not just headlines.

Then make the connection explicit. Every time an idea cites a source, the writer writes one line saying what it borrowed, named from that breakdown. If it cannot name it, drop the source. If nothing is left, the idea is original. Show that line under Sparked by, and stop cutting off the source titles.

A few rules. No source sparks more than two ideas a day. A trend needs at least two sources. Never force a source type: if the evidence is not there, write fewer of that kind, and the board does not have to show every type. If fewer than fifteen clean sources survive, run a fresh scan on my DNA phrases first instead of stretching what is left.

My Reddit list has subreddits that are just people posting videos. Suggest replacements where people actually ask questions, and ask me before changing it.

Regenerate today's ideas and show me the board.

Rebuild the scoring. Replace the three scores with five pillars, each 1 to 10, with a one-line reason for every number.

Stop: would a stranger hold for the first three seconds.

Useful: does the viewer leave with something they did not have. 9 or 10, they can do it themselves tomorrow. 7 or 8, they now know something concrete: what is possible, what it costs, where it breaks, an honest verdict. 4 or below, spectacle only. The topic does not decide this, the video does. A result on its own scores low. The result plus the how and the verdict scores high, even if most viewers will never copy it.

Spread: would someone share it, save it, or argue about it in the comments.

Proof: has the market already shown demand. Not an opinion. Compute it from the sources the idea cites: their multipliers, how many mined comments asked for it, whether a watched creator broke out with it. Originals get 0 and that is fine.

Fit: can I credibly make it, and does it pull people toward my offer.

Rules. Fit under 6 means the idea is out. All five add up to the total, out of 50. Too close to its source costs two points on Spread. The judge must see the source breakdowns, not just the idea.

One more thing. Any idea with Stop 8 or higher and Useful 4 or lower gets rewritten before scoring: keep the hook, replace the payoff with something the viewer actually takes away, and name it in the title. Drop the original. Score the rewrite like any other idea. On the board it is just an idea.

Originals cite nothing, so give them Proof another way: match each original to the sources about the same topic and score it from those, two points lower than a direct source. Hold three seats on the board for originals scoring at least 30, and ask the writer for at least five originals out of twenty.

Show the five pillars and the total on every card, with the reasons when I expand it. Regenerate today's board.

Add a My Content screen. The data is already here: when we built My DNA the app saved my last 30 posts from each account, with views, likes, comments, shares, dates and thumbnails. Show them.

At the top, one box per platform with my median views. Under that, my posts as cards: thumbnail, title, platform, and a badge with how many times my own median it did, like 20x your normal. Stats: views, views per day, likes, likes per view, how long ago. Buttons: Open post, and Read my audience, which mines that post's comments the same way we mine other people's.

Group the same video across platforms into one card. Let me hide a post. Add a refresh that re-pulls my accounts, once a week unless I force it.

Run the same breakdown we built for other people's posts on mine, so I can see why my winners won.

Connect it to the ideas. Give the writer my top five posts of all time as evidence. An idea can be a follow-up to one of my posts or a remix of it, never a remake. When an idea builds on one of my posts, that post is a cited source and its multiple of my median feeds the Proof score, the same as a stranger's post would. Comments mined from my own posts count as demand, and stronger than strangers' comments, because these are my people asking.

Now that the app has my posts, give the judge anchors: my three best and three weakest, with their numbers, as what an 8 and a 4 look like for me. Re-score today's board so I can see what moved.

Right now ideas disappear after three days unless I script them. Add an Ideas Bank.

Put a Keep button on every idea card, on the board and on the bench. Kept ideas go to a new Bank tab and stay there forever. Each one has a status I can change: Saved, Scripted, Filmed, Posted, or Dropped. Scripting an idea moves it to Scripted automatically. Keep the original score, the sources, the "what I borrowed" lines and the date it was written.

When I mark an idea as Posted, let me pick which of my posts it became from My Content. From then on the idea shows that post's real views and its multiple of my median right on the card, next to the score it was given. That is how we will find out whether the scoring is any good.

The bank talks back to the writer. Kept ideas are never re-suggested. Ideas I have marked Dropped join the dismissed list. And once an idea is Posted, the writer treats it like one of my own posts: follow-ups yes, remakes no.

Bank screen: filters by status and by type, newest first, and a small summary at the top: how many saved, scripted, filmed, posted.