I didn't set out to build a social media tool. I set out to post more OCaml memes.
The Setup
As a developer advocate for the OCaml programming language, I started and managed the language's social media presence across four platforms: X, LinkedIn, Bluesky, and Mastodon. We had good content sources - a community RSS aggregator, the OCaml Planet with blog posts from companies and developers, release notes on the OCaml Changelog, behind-the-scenes and experimental updates on Backstage OCaml, and the discussion forum discuss.ocaml.org.
For the OCaml.org RSS feeds, I set up the automation tool dlvr.it to automatically post articles from the feeds to all four accounts.
It worked. The OCaml community is small but invested. The OCaml social media accounts grew slowly but steadily with almost no manual effort - by December 2025, they were at around 2k followers on X, 1.6k on LinkedIn, 1.2k on Bluesky, a couple hundred on Mastodon. For automated link posts of a programming language that's not (yet) widely known, that's decent growth.
There's a Better Way Than Just Posting Links
But I also knew we could handle this much better.
Automated link posts are fine. They get the content out there. But posts with your own commentary - a summary, a quote, a reason why this matters - are much more interesting. Your own posts signal that someone's there who wants to hear what the community has to say.
I told myself I'd do better. I'd read the posts, write thoughtful summaries, craft platform-specific content.
I never did, because something else was always urgent and genuinely more important (e.g. writing the newsletter, answering comments, improving the website).
The mental overhead was too high. Read the article. Understand it well enough to summarize. Write something in my own words. Adapt it for each platform's style and character limits. Do this for every post, multiple times a week.
It felt like starting from scratch every time. And I already struggle with building habits and sticking to them.
OCaml Memes for the World
Fun fact: I like making OCaml memes. I'm probably the person on this planet who has created and posted more OCaml memes than anyone else. I'm not sure if that's something to be proud of, but there it is. When I start something, I do it thoroughly.
Here's how it goes for me: when I create memes and I'm in the mood, I have ideas for 3-5 memes per day, for a few days. I'd post them all at once on my personal Twitter account, then also on Bluesky and finally on LinkedIn - basically spamming my followers. Bluesky expects alt text, so I started creating alt text with Claude. Then I naturally used those on X and LinkedIn too. Then nothing for weeks or months, and I'd wonder if that was too much, or went too far.
Actually posting memes to three different platforms with proper alt text for accessibility was annoying, but as long as I was motivated, doable.
Building FeedMansion
When I saw the opportunity to build something of my own, I took it.
I knew the workflow I wanted:
- Monitor RSS feeds automatically
- Generate draft posts with summaries and commentary
- Let me review and edit before anything goes out
- Post to multiple platforms at once
- Schedule posts so I could batch my meme creation without flooding everyone's feeds
So I built it. FeedMansion.
What Changed
I haven't switched the official OCaml accounts to FeedMansion yet, but I'm already using FeedMansion to share hand-picked posts from the OCaml RSS feeds when I like the draft the AI wrote. I'm still not sure which prompts deliver good results, but the great thing about my tool is that I can keep adjusting the prompts until almost every round of drafts has a hit.
Unlike the pure link/article posts from the automated OCaml accounts, my LinkedIn posts get more comments and engagement, which I can then respond to. Besides LinkedIn, X, and Bluesky, I now automatically post to Mastodon too (my Mastodon account has existed for a long time, but due to lack of followers and engagement I never managed to post there consistently).
Now I upload a batch of OCaml memes to FeedMansion. The AI generates alt text descriptions for each one - usually quite usable, and if not I edit them. I schedule the memes spread out over the next few weeks. One click: X, LinkedIn, Bluesky and Mastodon done simultaneously.
Observation: People notice when you use alt text on memes. I've gotten positive comments about it a few times. Most meme accounts don't bother. And as a bonus, proper alt text helps images rank in search.
I post more now, without it feeling like a burden. I'm still not consistent by nature - maybe I never will be. But I'm doing better than I used to. Just having scheduling solved half the problem. When I create 3-5 memes in a burst, I can spread them out over weeks instead of posting them all at once or letting them collect dust.
Philosophical Question: Is Using AI in Social Media Still Authentic?
Once I understood what's possible today with AI, and especially with FeedMansion, I started thinking about the role of AI in social media.
I believe AI can help people express themselves, especially those who struggle with finding the words. Writer's block is real. Having a draft is always easier than starting from nothing. Even if the draft is garbage - because then suddenly you realize what you actually want to say.
But I also think it's tempting - and wrong - to believe you can fully delegate self-expression and personal brand building to AI. If it's not my voice, what's the point?
And what about comments? I see more and more social media bots leaving completely inappropriate and obviously AI-generated comments. In that case you face a dilemma: the platform algorithms usually reward responding to comments. But every human user who sees that has to endure this content diarrhea in their feed just because I commented. So I avoid responding to obviously AI-generated comments.
FeedMansion is a ghostwriter, not an autopilot. Nothing gets posted without your approval. You read every draft. You edit, skip, or approve for publishing. The AI helps you get past the blank page. The voice is still yours.
For company accounts, I see no concerns about using AI for social media content creation. Many companies already have social media teams speaking on behalf of the brand. In contrast: for personal accounts, where you want to build real relationships - AI automation is simply problematic.
What's Next
FeedMansion solves my posting problem. But there's another side to social media: the consumption side.
I still spend too much time doomscrolling. I miss posts from people I actually care about. I discover interesting conversations too late to join in.
That's why I'm building Nitfish - a semantic, AI-powered filter for social media. Nitfish finds posts worth engaging with. It monitors social networks and forums. With AI support, it finds what's actually relevant to you. The inverse of brand monitoring: not "what are people saying about me?" but "what should I engage with?"
The idea: content arrives at my doorstep. Nitfish finds it, FeedMansion helps me respond. Less scrolling, more interacting.
For developing both tools I use OCaml - a programming language I know and trust, and one that's actually very well suited for writing maintainable, correct code that can still be easily adapted to new requirements.
I'm not sure exactly who the ideal user for FeedMansion is, or if there's only one. Maybe it's DevRel people like me, solo founders, content creators who also spend way too much time getting their content onto different platforms.
If you think FeedMansion or Nitfish could help you, reach out.
FeedMansion starts beta in January. Nitfish is at the beginning of development.