Weekly Strategist

Install it

Open Custom AI → Explore, choose Weekly Strategist, and install it. Start with Set up my Weekly Strategist. It asks for three things:

• the topics you create around and the audience you serve,
• your own social profiles,
• optional competitors, creator lists, platform constraints, and the result you care about.

It resolves exact creator accounts, creates or reuses one Weekly strategy board, and places a Weekly Strategist — Settings note on it. Later runs read that note instead of repeating the interview. You can change the setup in chat at any time.

What happens in a weekly sweep

The strategist reads your settings, your Market findings table, and the newest prior memo. The table is its memory, so it rarely needs to reach back through older memos. It then runs four connected research passes:

1. Your scoreboard. It checks your recent posts against your own baseline and grades whether you actually tested last week's recommendation.
2. Creator discovery. It finds people who substantially cover your topics, creators adjacent to your known competitors, or accounts just outside a selected creator list. A normal run uses no more than two discovery calls and selects three to five strategically distinct creators.
3. Market evidence. It searches early breakouts, proven current patterns, and older examples that explain durable mechanisms. It searches only a few distinct hypotheses and reads two to four pivotal posts in full before claiming to understand why they worked.
4. The decision. It separates new, building, saturating, and disproven signals, then joins the strongest evidence with your audience, proof, positioning, and realistic production capacity.

Your own posts inform personal fit, but they never count as proof of a wider market pattern. A confirmed finding must repeat across creators, including at least one account outside your saved radar. A normal full run targets 12 to 16 tool calls and stops at 18 rather than expanding research just to fill a report.

What lands on the board

The board stays deliberately selective. It contains your settings, one Market findings table, one dated memo for each full sweep, and native cards for the market posts actually cited in the final findings. It does not save every search result, discarded candidate, or post from your own scoreboard. If you choose a play and ask another Custom AI to plan or draft it, that finished artifact can return to the same board.

The findings table

Findings are the one thing that outlives a single week, so they live in their own Market findings table on the board rather than being copied from memo to memo. Each row is one finding with its status (watch, active, cooling, retired), the date it was first seen, the date it was last confirmed, a confidence label, what it claims, what it means for you specifically, and a link to the strongest post behind it.

Every sweep adds genuinely new findings, moves the ones that changed, and leaves the rest alone. A first-seen date is set once and never rewritten. A finding that stops being true is marked cooling and then retired rather than deleted, because a bet that decayed is worth remembering.

You can sort it, group it, and edit it yourself. Change a status by hand and next week's sweep will respect it. Anything still unresolved sits in the table as a watch row, which is how the next run picks it up.

What the memo contains

Every full run saves a Weekly Strategy — YYYY-MM-DD memo of up to roughly 1,500 words (shorter when evidence is thin) with:

• one clear bet and why it matters now,
• the research scope and any important coverage gaps,
• your own recent winners, misses, and last-week grade,
• up to three findings named with one line each on what changed this week,
• exactly three concrete things you could make,
• one small experiment with an observable success signal,
• one thing to stop or deprioritize when the evidence supports it,
• a short method note when something failed or lowered confidence.

The three plays stay inside the memo until you choose one, so a sweep does not clutter the board with speculative drafts. A memo is a snapshot of what the strategist believed that week and is never edited later to look correct in hindsight. The findings table is the opposite: it is the running record, and the only place a status changes. It also grades the previous bet honestly. If you did not publish it, the memo says so.

If a research tool fails, Weekly Strategist retries it once, records the missing coverage, and finishes a useful partial memo from verified evidence instead of looping or abandoning the run.

Smaller jobs work too

You do not need to run the full memo every time. Ask Weekly Strategist to find competitor creators, check one platform, compare two accounts, review the watch rows in your findings table, or explain what changed since the last memo. Narrow questions do not add material to the board unless you ask. It saves a dated strategy note and receipt cards only when the work includes the complete scoreboard, findings, receipts, and weekly bet.