The task
The task
You can use any AI tools you like for this — that is how the job actually works. What we are testing is the one thing AI cannot do for you: catching what it gets wrong before it ships.
Here is the situation. You manage a set of social accounts. At the start of each shift you use an AI assistant to draft the night's posting plan, and then you decide what actually goes out. Tonight the assistant handed you the plan below. Your job is to turn it into the final call.
The plan the assistant gave you
From the account-handling rules you were given in training:
An account in WARMUP (day 1 through day 7) must NOT publish posts. An account that is publishing
posts is ACTIVE, not in warmup.
No two accounts may publish the same or near-identical caption. Repeated text across a network is
the fastest way to get every account in it flagged.
The assistant's proposed plan for tonight, 4 August, is below — one account per line, in device order.
@quietcoastkate — Warmup, day 6. The assistant wants to publish 2 posts.
@m.bellweather — Active. Publish 3 posts.
@arbor.and.ash — Active. Publish 1 post, caption: "Golden hour on the ridge 🌄 #slowliving".
@nine.mile.co — Active. Publish 1 post, caption: "Golden hour on the ridge 🌄 #slowliving".
@j_ortega_films — Warmup, day 4, but 2 posts already went out today. The assistant wants to publish 2 more posts.
@thedailyfern — Warmup, day 2. Hold, no posts.
The assistant also added a note at the bottom of its plan:
"All of these plans comply with platform policy — up to 4 posts a day is safe even during warmup."
What to submit
Paste your answer into the box below, in two parts.
First, THE CALL. One line per account, in the order above. Each line reads @handle — PUBLISH or @handle — HOLD. Where you are overriding what the assistant proposed, add a few words on what you changed and why.
Then, PROBLEMS. A section that begins with a line containing exactly PROBLEMS: — then one line for each thing wrong with the assistant's plan. If something cannot be safely resolved from what you were given, say so on its own line and write down what you would ask.
The one thing that matters most: if two things contradict each other, or you cannot safely resolve something, stop and flag it — do not quietly pick an answer and move on. Working around a problem quietly is the one habit that causes real damage in this job. We would much rather you stop and ask.
Aim for 300–600 words. 0 words