Every Hot or Not is built on three perspectives — the buyer, the hiring manager, and the interrupt-test (a stranger scrolling past at 6 PM on a Wednesday). The Oracle returns a 1–10 score, a verdict — HOT, WARMING UP, LUKEWARM, COLD — and five prioritized moves.

This one came in last week. Senior fractional COO. Healthcare ops background. Six years freelance, no website, runs her whole pipeline through LinkedIn DMs. Calls herself "Maya" for this walkthrough. Real verdict, real prompt, with permission to publish.

What her profile looked like

Headline: "Strategic Operations Leader | 20+ Years | MBA | Healthcare | Operations Excellence"

About section opener: "Results-driven operations leader with a proven track record of driving operational excellence across multiple verticals."

Banner image: A stock photo of hands on a laptop keyboard. Cool blue.

Featured section: Empty.

If you've been on LinkedIn for more than a week, you can already feel what's wrong. But "feels wrong" isn't actionable. The Oracle's job is to turn feels-wrong into a sentence she can do something with by Tuesday.

What the Oracle said

The verdict
SCORE: 4 / 10 VERDICT: LUKEWARM The buyer cannot tell what you sell, who you sell it to, or what changes after they hire you. The headline is a list of credentials — credentials are table stakes, not a hook. The about section opens with a sentence the reader has read on six other profiles in the last hour. The banner says nothing the headline doesn't already say. The featured section is empty, which means the part of the page designed to do work is doing none. Your buyer is not "everyone with operations problems." Your buyer is a Series B-to-C healthcare founder who just lost a head of operations and is two weeks from running their next board meeting without one. Speak to her.

The verdict has a tone, on purpose. Not mean. Not cheerleading. Direct. The voice formula in the Brand Bible: a brilliant friend who happens to have 20 years of marketing experience and a subscription to every AI tool. She tells you the truth because she respects you.

The five prioritized moves

The Oracle returns moves in priority order. The first one is always the one that moves the score the most. Here's what she got:

  1. Rewrite the headline as a promise, not a résumé. Try: "Fractional COO for Series B–C healthcare founders. I run your operations until your full-time hire shows up." One sentence. Names the buyer. Names the job. Names the boundary.
  2. Replace the about-section opener. First two lines are what shows above the "see more" fold. Stop with "results-driven." Start with the moment her buyer is having: "Your VP of Ops just left and you have a board meeting in three weeks. I do this for a living."
  3. Add three Featured items. One case study, one teardown of a healthcare ops decision, one testimonial. Featured is the highest-leverage real estate on the profile and hers was empty.
  4. Banner with one line of text. Black background, cream type, the same promise as the headline. The banner stops being decoration and starts doing work.
  5. Post once a week for four weeks. Not "thought leadership." A specific operational decision she made for a client and what happened. Receipts compound. After four weeks she will have a portfolio without ever saying the word "portfolio."

What she did with it

Maya rewrote the headline that night. She rewrote the about section the next morning. She added two Featured items by Friday — she didn't have a third yet, so she left the slot. She made a banner in Canva at 11 PM Sunday using the colors from her old website and IBM Plex Mono. She posted on Tuesday: a 280-word breakdown of how she helped a Series C client decide to delay a hire by ninety days, what they did instead, and what it saved.

That post got 14,000 impressions, three DMs, and one founder who said the words you want to hear: "I think I need exactly this."

What was actually broken

The interesting thing is what wasn't on the list. Maya assumed the problem was that she didn't have enough credentials on the page, or that the wrong people were viewing it, or that the LinkedIn algorithm had quietly demoted her. None of those were true.

The problem was that her page had no audience. The audience was "everyone." Which is the same as nobody.

The Oracle is a mirror. It does not generate strategy from nothing. It looks at the page, identifies the moments where the buyer's brain has nowhere to land, and sends back the smallest set of moves that fix that. Five moves is the cap. Four is better. Three is best.

The prompt structure, briefly

For the practitioners reading: the Oracle is one Claude Sonnet 4.5 call with a JSON contract. The system prompt sets the three perspectives, the tone, and the move-prioritization logic. The user message contains the scraped public profile fields. The output is a typed object — score, verdict, summary, five moves with rationale and example copy — which the front end renders into the verdict card you see at the end of the Hot or Not page.

What makes the Oracle work is not the model. It's the system prompt. We rewrote it eleven times before it stopped being polite and started being useful. "Useful" beats "polite" every time someone is paying for a verdict instead of a hug.

Want one of your own?

Free Hot or Not is at /hot-or-not. The $25 Hot Sheet — the deeper, manually-run version with custom prompts and a 48-hour PDF — is at /hot-sheet. Both are real. Both come with receipts.

— L.

← Back to The Lab Next: Why "operator" is the wrong word →