Just shipped LinkedIn post scheduling to XreplyAI. Wanted to share what made it interesting to build.
X (Twitter) is relatively forgiving — the API is well-documented, scheduling is straightforward, and the content format is constrained (character limit does the work). LinkedIn was a different fight.
Three things that made LinkedIn harder:
The voice training piece ended up being the differentiator. XreplyAI trains on your tweet archive and uses that as the style layer for every platform, including LinkedIn.
BYOK — you bring your own AI key, no extra subscription cost.
https://xreplyai.com?utm_source=indiehackers&utm_medium=social&utm_campaign=feature-2026-05-19
The strongest part here is the shift from “reply tool” to “voice-preserving publishing layer.” LinkedIn is a much better test than X because the penalty for sounding automated is higher. People can feel templated writing faster there, and that makes the voice profile layer more important than the scheduling feature itself.
I’d make that distinction sharper in the positioning. Scheduling is useful, but it is also easy to copy. “Your own writing style carried across platforms without sounding like AI” is the more defensible wedge.
The naming is where I’d be careful. XreplyAI made sense when the product was mostly tied to X replies, but LinkedIn support already shows the product is moving beyond that original frame. If this becomes a broader AI social publishing and voice system, Beryxa .com would age better than a name locked to X/replies/AI.