Surface Alignment Check

Watch the two-minute walkthrough 2:23

A two-minute walkthrough: why our resume, profile and portfolio drift apart unseen, the fields, the verdict, and how to run the prompt in any model.

Play video

The decision it forces

Do our resume, our profile and our portfolio describe the same person?

They're three documents written at three different times, each for the version of us that existed when we wrote it, and the drift between them is invisible from the inside. The check lists where they disagree and rewrites nothing; choosing the one to align to is ours.

To run it, copy the prompt below into any model you already use.

The rule that makes it a drill

Lee Rodrigues ran the check on his own three, on Episode 3:

take that portfolio site, take that LinkedIn profile, take that resume and tell your Claude, hey, do some black hat thinking for me. Are these three people, this resume, this LinkedIn profile and this portfolio site, are they the same person? Because I did mine three days ago, they were not the same person. They had gone in different directions and Claude was like, are you sure this is you?

His portfolio had stopped in 2015, and his current clients were nowhere on it. Nothing was wrong with any one surface; the three of them described three people. Christine Rodrigues, the episode's guest, named what that is:

like you're talking about, I mean, signal, but you had multiple signals and the signals weren't aligned, which means they were really noise.

Lee's rule is the part people skip. Make the resume make sense first, then align the LinkedIn to that and the portfolio to that. Wiggle all three at once and we paint ourselves into a corner and ask a model for a resume that looks like everyone else's. Sam's argument on the same episode is why the check never rewrites: ask AI to make the resume and by default it makes it as average as possible. A model asked only to find the drift leaves the choice with the one person who knows which of the three is true.

The fields

claim through keep repeat once per claim, and divergences counts them. There's no worked example on this page: a filled check is one person's three documents with their contradictions listed.

The verdict rule

When they don't, the decision is which one the other two align to. When divergences is above zero and chosen_focus is empty, the finding is three people and no decision: choose the focus before touching any surface. When chosen_focus is filled, the verdict is align to it, and the divergence list, top to bottom, is the work. Zero divergences is aligned, which is rare.

A claim on one surface only is the one place alignment can do damage. Some of those are drift. Some are the corner case: the odd job, the strange client, the thing nobody else in the pile has done. Lee's hires on Episode 3 were exactly that, and Sam's argument on the same episode is that a model writes toward the middle of the distribution. Aligning by deleting whatever isn't on all three surfaces does the same thing by hand. So every single_surface claim gets a keep before any aligning starts, and only we fill it.

What not to do with the result

Christine's rule, from the same conversation, is about the move the divergence list invites. Push a button and send, and it isn't really our document. Every bullet point a model writes into a resume is a story we may have to tell in an interview, and if we weren't involved in creating it, we haven't been rehearsing.

That rule is hers and it stands on its own. The check on this page is built to survive it: the model reads three documents and reports; it writes nothing that will later have to be defended in a room.

Run it with any model

Nothing to install. Paste all three. The model compares and reports; it does not write, rank, or advise.

Run a Surface Alignment Check on the documents I am about to paste: my resume,
my LinkedIn profile as text, and my portfolio or personal site as text. If I
have more or fewer than three, use what I give you.

The question is one: do these describe the same person?

DO, IN ORDER
1. From each surface, extract every claim about me: a role, a skill, a result, a
   client type, a specialism, a period, a location. One claim per line, in the
   words the surface used. Do not merge similar claims from different surfaces;
   keep them separate so the differences show.
2. For each claim, record which surfaces it appears on and which it is absent
   from.
3. For each claim, record any claim on another surface that contradicts it: a
   different title for the same period, a different specialism, a client type
   one surface leads with and another never mentions. Record the contradicting
   claim verbatim, or null.
4. Count the divergences: claims that are not on every surface, plus claims with
   a conflict. Separately, count the claims that appear on one surface only.
5. List the one-surface claims and, for each, ask me whether to carry it onto
   the other surfaces or drop it. Record carry, drop, or null. Do not suggest
   which, and do not call any of them unusual, weak or memorable.
6. Then ask me one question and record my answer: which surface, or which single
   story, is the one I want the others aligned to? If I do not know, record
   null. Do not answer it for me.

RULES YOU DO NOT BREAK
1. Do not rewrite anything. Not a bullet, not a headline, not a summary. If I
   ask you to, tell me that is a different job and this one is not it.
2. Do not rank the surfaces or tell me which one is best.
3. Do not propose the focus, suggest a positioning, or tell me which claims to
   drop. chosen_focus is mine.
4. Do not tell me a claim is weak, strong, impressive or generic. Whether it is
   true and whether it is on every surface are the only two properties you
   report.
5. Do not add a claim that is on none of the surfaces, even if it is obviously
   implied.

OUTPUT
Valid JSON matching the schema surface-alignment/v0.3 as published at
https://sigsub.show/takeaways/surface-alignment/#the-answer-format. If you can't
read that page, use these keys, in this order: schema; surfaces, a list; claims,
a list, each with claim, appears_on (a list), absent_from (a list),
conflicts_with and keep; divergences, single_surface, chosen_focus. Then a table
with one row per claim and one column per surface marked yes or no, then the
divergence count and the one-surface count, then the questions in steps 5 and 6.

What to check in the answer

It may extract, compare, and flag what is missing from one surface and present on another. That is reading.

It may never rewrite, rank, propose the focus, or decide which one-surface claims to keep. Every one of those returns us to the average, and the average is the one place a job search cannot afford to be.

The part that's not optional

Choose the focus before touching any surface. Lee's order is resume first, made true, then the other two aligned to it, and the reason is in his own words: wiggle all three at once and the exit is a fresh generic resume from a model.

Then do the aligning by hand. The check has told us where the three documents disagree. Which one is right is the one thing on the page a model was never asked.

{
  "schema": "surface-alignment/v0.3",
  "surfaces": ["resume", "linkedin", "portfolio"],
  "claims": [
    {
      "claim": "what is being asserted",
      "appears_on": ["resume", "linkedin"],
      "absent_from": ["portfolio"],
      "conflicts_with": "the contradicting claim, or null",
      "keep": "carry | drop | null, only on a claim that appears on one surface"
    }
  ],
  "divergences": 0,
  "single_surface": 0,
  "chosen_focus": null
}