← All posts

Forward an X post to Marvin and get the actual content back

Forward an X post to Marvin and get the actual content back

You've done this before. You're scrolling X, you see a thread that looks important, you don't have time to read it carefully. You copy the URL and paste it into ChatGPT: "summarize this thread for me."

The summary looks plausible. You move on with your day.

Then later, when you actually go back to the thread, you realize the summary was about half-imagined. The AI couldn't actually load the page. It generated a summary based on the URL slug, the username, and what a tweet on that topic might say. The plausibility was the bug.

This is a known limit, and it's not just ChatGPT. It's most AI assistants. X is hostile to automated readers. The public web page for a post requires a login to show substantial content. The official API is priced for enterprises. Scraping gets blocked. So most AI tools that "read the web" silently can't actually read the one site you most need them to read.

We fixed this for Marvin. You can forward an X URL (or paste it in an email body) and Marvin will read the actual post and tell you what's actually in it. Or, when something genuinely can't be read, Marvin tells you that, instead of hallucinating coverage.

What it looks like

Send Marvin a link with one line about what you want:

To: marvin@ccmarvin.com

Hi Marvin,

What's the actual content of this post, and any context I should know?

https://x.com/jack/status/20

A minute later:

From: Marvin marvin@ccmarvin.com

Hi Alex,

The post reads, in full: "just setting up my twttr"

Context: this is the first tweet ever posted, by Jack Dorsey (@jack), on March 21, 2006. "twttr" was the platform's original name before it became Twitter. It's the single post the URL points to: five words, no media, no thread.

Sincerely,
Marvin

That's a deliberately tiny example (a five-word post), but it makes the point: Marvin tells you the post says exactly five words, rather than padding it into a plausible-sounding paragraph. When you forward a real thread with substance, you get the substance, accurately.

A more typical investor use:

To: marvin@ccmarvin.com

Hi Marvin,

One of my portfolio founders just posted this. Read it and tell me whether it's material (funding, product, hiring, or a customer announcement) or just noise.

https://x.com/<founder-handle>/status/<id&gt;

From: Marvin marvin@ccmarvin.com

Hi Alex,

Read it. Short version: material: it's a product launch with a customer name attached.

The post announces general availability of their enterprise tier and names a Fortune-500 logo as the launch customer. That's the kind of post that tends to precede a formal investor update by a few days. No funding or hiring signal in this one. If you want, forward me the replies and I'll tell you how it's landing.

Sincerely,
Marvin

What you can do with it

Some uses we see daily:

  • Triage a thread you don't have time to read. Forward the URL with "what's the gist?" and get an accurate summary in your inbox.
  • Get the source for a Bloomberg article. Reporters often cite X posts; forward the linked post to Marvin for the original.
  • Catch up on a founder's announcement. Forward the URL and ask "is this material to my position?" and Marvin reads the post and gives you the relevant context.
  • Pull a quote you remember seeing. Forward the URL and ask "give me the exact text and timestamp," useful when you're writing something that needs to cite the post precisely.

For investors, the third use is the killer. Portfolio company founders increasingly announce material things on X before they file or formally email investors. Being able to get accurate text back on demand, without having to switch to the X app, is a real time save.

What Marvin can't do

A few honest limits, which Marvin will state clearly rather than paper over:

  • Replies and quote-tweets. A single Marvin read covers the single post the URL points to. The replies underneath aren't included; if you need them, send the URLs of the specific replies you want read.
  • Deleted or geo-restricted posts. If a post has been removed or isn't visible from the request's region, Marvin reports that and stops, rather than guessing what the post might have said.
  • Locked or private accounts. Posts from private accounts aren't readable through any of Marvin's backends.

When Marvin can't read a post, the reply is unambiguous about it:

From: Marvin marvin@ccmarvin.com

Hi Alex,

I couldn't read this one. The account is protected (posts are only visible to approved followers), so none of the ways I have to read X posts can see it. I'd rather tell you that than guess at what it says. If you can see it yourself and paste the text into a reply, I'm happy to summarize or analyze it from there.

Sincerely,
Marvin

If you'd rather have "I couldn't read this" than "here's a plausible-sounding summary that might be wrong," this is the trade Marvin makes by default. You can act on what Marvin tells you because Marvin will tell you when he can't see something.

How to try it

Find an X post you want summarized: your last interesting thread, an announcement you bookmarked, a reply you can't remember the wording of. Forward the URL to marvin@ccmarvin.com with one line asking what you want: a summary, a quote, an assessment. If you want to see it work on a known post first, the first tweet ever is a fine test:

To: marvin@ccmarvin.com

Hi Marvin,

What does this post say, and what's the story behind it?

https://x.com/jack/status/20

The reply you get back will either be the actual content, accurately summarized, or a clear "I couldn't see this for [specific reason]." Either answer is more useful than the imagined one most tools give you.