Who Actually Viewed My Semi-Viral LinkedIn Post

I do not think of myself as good at sales.

People keep assuming I work for Clay, the tool sales teams use to build a list of companies and people and then fill in the details about them. I do not work for Clay. I just talk about it a lot.

Anyway. A post I wrote about cold email did better than anything else I have written. It got me about 1,000 profile views on LinkedIn in a few days.

That felt great for about a day.

Then the real question showed up. Who were they?

A thousand profile views is not a list until you know who was in it

The counter does nothing on its own. Nobody has ever paid me because a number went up.

What I wanted was the small group inside that 1,000 who I could actually help, plus the peers I would enjoy geeking out with.

LinkedIn shows you who viewed you. It will not hand you that list as a file. There is no export button, and Clay does not scrape LinkedIn, so the path I would normally take was closed.

So I took the dumb path.

I scrolled the viewers list all the way to the bottom, so every row loaded. Then right click, Inspect, and copy the whole element out of the developer panel built into Chrome. Save it.

That gave me 5.8 MB of LinkedIn's own HTML. Thousands of lines of markup with names buried in it.

The page was already the export. I just took it.

Claude Code refused 2.7 million tokens

Then I did the obvious thing and pasted 5.8 MB of HTML into Claude Code.

2.7 million tokens. It took one look and said no.

A token is roughly a chunk of a word. A model can only hold so many at once, and 2.7 million is far past what any of them hold today.

Fair enough. Obvious in hindsight. I did it anyway.

The session was cooked, so I reset it and started over.

Giving Claude the file instead of the text is what worked

Second attempt, I pasted nothing.

I had Claude Code make a new file, put the HTML in that file, and then told it one thing. Process this file.

And it did something I did not ask for and would not have thought of.

It did not try to read 5.8 MB into its head. It wrote Python scripts to chew through the file for it. Find the pattern, pull the names out, write the results somewhere, check them, fix the script, run it again.

The model never held the whole file. It wrote a small program that did, then read what the program found.

That is so smart, and it is the part of this I keep thinking about. The limit was real. It just went around it.

What came back was a clean CSV. Name, title, company.

Then I asked for a second pass. Score every person against my ICP, meaning the kind of company I can genuinely help. That part was easy for a nice reason. My ICP is already written down in my repo, so I pointed at the file instead of explaining it again.

Every row came back with a score and a reason. Some potential clients. Some peers. Some possible partners. Plenty who are not in it at all, which is fine, and is most of the point.

What do you do with a scored list of profile viewers?

Go talk to the interesting ones.

Not a pitch. More like, hey, what's up, let's geek out for a minute.

That is the only part of this that turns 1,000 views into anything. It is also the part I actually like.

Who viewed your profile is yours. Who liked a post belongs to everyone.

Here is the distinction I did not see until I had the list.

Anybody can see who liked a post. Yours, mine, a competitor's. I have written about reading their likes instead of their posts and about how the quiet ones are not inactive, and I still think that is underused. But it is not mine. It is sitting in public for anyone who bothers.

The people who viewed your profile are different. That is first-party data, meaning only you can see it. A competitor cannot buy that list. Nobody can pull it but you.

I might be wrong about how far this goes. It is one post, one afternoon, one method that worked once. I have not tried it on anything bigger.

But it has me looking around for the other first-party data I already own and have not touched. There is probably more of it than I think.

If you have done a smarter version of this, tell me. I would like to steal it.