The comment section is the last unautomated growth channel — and that’s exactly why it’s worth your attention
Most social media operators I know have their publishing stack solved. Buffer or Later handles the calendar, CapCut handles the edits, a spreadsheet handles the repurposing matrix. What almost nobody has solved is the inbox: the 400 unread comments on a Reel that popped, the DMs from people asking for the link, the returning fan who commented on your last three posts and deserves a real answer. That’s the gap MePersona is aiming at — a workspace that reads your comments and DMs across platforms, drafts replies in your voice, and remembers individual fans. Whether that’s a good idea is a different question from whether it’s a good product, and I want to pull those apart.
What MePersona actually is, stripped of launch-day gloss
The founder is Steven Hacker, who says he spent 25 years in film and TV before building this. His framing on the launch thread is that artists hit a wall: they burn out on replies, go silent, or hand the account to a social media manager who doesn’t sound like them, and “true fans can tell.” That’s a real observation. It’s also the exact pitch every AI ghostwriting tool has made for two years.
What’s actually shipped, per the maker’s own posts: a unified inbox for comments and DMs across nine platforms, AI-drafted replies in a learned voice, an agency workspace where managers run multiple creators with separate personas and a shared calendar, a “surge mode” for viral posts, and a rebuilt “fan memory” system. The team is also adding WordPress (blog writing and scheduling), Mastodon, and Pinterest. Current language support is English and Spanish, with more promised. Pricing is not disclosed in the source beyond a 30-day Premium trial via a founding-member link.
The interesting part isn’t the drafting. It’s the memory layer and the surge logic. Those are the two places where this either becomes genuinely useful or collapses into another ChatGPT wrapper.
How the voice training actually works (and why that matters)
I’ve tested enough “learn your tone” tools to be cynical about them. The usual pattern is: paste 500 words of your writing, get a system prompt, watch it produce something that sounds like a LinkedIn thought leader with your name on it. MePersona’s approach, as described by the maker in the thread, is more structured: you record yourself talking so it picks up how you actually speak, you play a swipe game picking replies that sound like you, and you can connect Instagram or X so it learns from posts you’ve already written.
There’s also a “voice match score” on drafts, and a tuner that looks for patterns across your edits — the maker’s examples are things like “always cuts the exclamation points” or “never calls fans babe” — and suggests voice updates for your approval. Nothing changes without approval. That last detail is the one I’d actually test. In my experience, the failure mode of these systems isn’t that they can’t learn; it’s that they over-learn from a single sarcastic reply and then every future draft sounds sarcastic.
The fan memory feature is the real product, and the real risk
Here’s the mechanic, per the maker’s detailed answer in the thread: after every exchange, the system extracts facts about that fan — interests, location, plans they mentioned, things they don’t want — and keeps up to 50 of them per fan. Pinned or manually added notes never get pushed out. When drafting, the persona reads those facts plus the last 10 messages in the conversation. Moods fade after two weeks, plans fade once they’ve passed, so it won’t ask how your trip went three months later. On public comments it only uses what the fan shared publicly; nothing from DMs ever surfaces in a public reply.
That’s a thoughtful design. It also solves a problem most creators don’t realize they have until they’re big enough to have it: at 50k followers, you cannot remember that @sarahbuilds mentioned she was moving to Lisbon, and she will notice when you don’t. The maker’s own line in the thread — “fans can tell when they’re getting a canned reply, and they light up when someone remembers what they said last time” — is the whole thesis, and I think it’s right.
Where the math breaks
Fifty facts per fan sounds generous until you do the arithmetic. If you have 20,000 engaged followers and even 5% of them DM you regularly, that’s 1,000 fans × 50 facts = 50,000 facts the system is maintaining, refreshing, and reconciling across platforms. The maker openly acknowledges that “reconciling handles and names across social media platforms is a challenge.” I’d bet that reconciliation is where the product’s real engineering cost lives, and where the cracks show first — the same person commenting as @mike on Instagram and @mike_runs on Threads is two fans until proven otherwise.
There’s also a privacy question the thread doesn’t fully resolve. The maker says fan data “stays with the creator until the creator says to delete,” which is the right default, but it’s a policy statement, not an architecture description. If you’re an agency running ten clients through one workspace, whose data is whose, and what happens when a client leaves?
Surge mode is the smartest thing here, and the most dangerous
The feature I’d actually pay for: when a post gets 30 comments in an hour or less, MePersona waits a few minutes, ranks the wave, and drafts replies only for the comments worth answering — questions, price/link requests, returning fans. Spam, trolls, and three-word praise get skipped, and every skip is labeled with a reason so the creator can override.
This is a genuinely good product decision, because it maps to how virality actually works. The first hour of a post’s life is when the algorithm is deciding whether to push it further, and comment velocity is one of the signals it reads. But the comments that arrive during a surge are also the ones most likely to be low-value noise. A human can’t triage 300 comments in 20 minutes. A ranked queue can.
The maker also flags two guardrails: a surge can never consume all of a creator’s message/token budget, and the system sends replies one at a time rather than in a burst, explicitly to avoid platform spam flags and to protect API access. One commenter, KurachiWeb, called this out approvingly, and the maker confirmed the reasoning. That’s the kind of detail that tells you someone has actually operated these accounts rather than just built the demo.
Why TikTok creators should care more than LinkedIn ones
If you’re posting on LinkedIn, your comment volume is manageable and your audience expects a slower cadence. A drafted reply that sits for two hours before you approve it is fine. On TikTok, a video can go from 2k to 2M views in 90 minutes, and the comment section becomes a live performance. The creators who win there are the ones who reply fast, reply funny, and reply in a way that generates more comments. Surge mode is built for that reality. The LinkedIn use case is basically “inbox zero with better tone.” The TikTok use case is “don’t lose the wave.”
How this compares to what you’re probably already using
Let me be concrete about the competitive set, because “AI replies to comments” is a crowded shelf.
Buffer and Hootsuite both have unified inboxes and have been bolting AI reply suggestions onto them. Later leans harder into visual planning and link-in-bio. Metricool gives you analytics and scheduling at a price point that’s hard to beat. Sprout Social goes upmarket with social listening and team workflows. None of them, as far as I know, are trying to build a persistent per-fan memory layer or a voice model trained on your own recordings. That’s MePersona’s differentiation, and it’s a real one.
The closer comparison might be the wave of AI ghostwriting tools — Taplio for LinkedIn, Typefully for X — except those draft posts, not replies. Drafting posts is a solved-ish problem. Drafting replies at scale, with memory, is not.
Where MePersona looks weaker against incumbents: analytics. There’s no mention of reporting, UTM tracking, or performance dashboards in the source. If you’re a social media manager who has to justify spend to a client, an inbox tool without analytics is half a tool. I’d want to know whether they plan to integrate with Google Analytics or build native reporting before I’d move my whole workflow over.
The limitations, stated plainly
The disclosure question is unresolved, and the maker knows it. When Ninuna Ungiadze asked directly whether people writing in know it’s an AI, the answer was: by default nothing sends without creator approval, auto-send is opt-in, and “whether they tell their fans they use an AI assistant is up to them.” That’s a defensible position — plenty of creators have human assistants — but it’s also a position that will get tested publicly the first time a large account gets caught. My take: the creators who disclose will be fine, and the ones who don’t will have a bad week eventually.
“Nine platforms” is not disclosed as a list. The source mentions WordPress, Mastodon, and Pinterest as recent additions, and links to Facebook, Instagram, LinkedIn, Threads, and X profiles. The full roster isn’t spelled out. If you’re running a YouTube channel and comments there are your biggest backlog, verify coverage before you commit.
It’s a solo build. The maker says “it’s just me building this one.” That’s impressive and also a risk. Solo-built tools have great responsiveness and terrible bus factor. If you’re an agency putting ten clients on this, ask what happens if the founder gets hit by a bus.
Not for you if: you post less than daily, your comment volume is under ~50/week, or you actively enjoy replying to fans and consider it part of your creative practice. The whole value proposition is volume triage. Below a certain threshold, you’re paying for a solution to a problem you don’t have.
What I’d watch / test next
If you want to evaluate this properly this week, here’s what I’d do. First, sign up for the 30-day free trial — no card required, so the cost of finding out is an hour. Second, connect only one platform, ideally the one where your comment backlog is worst, and let it draft for a week without auto-send. Third, and this is the important one: keep a running tally of how often you hit “regenerate” versus “send as-is.” That ratio is the only honest measure of whether the voice model works for you. If you’re regenerating more than half the time, the persona isn’t trained yet, and you should spend ten minutes in the tuner before you judge the product.
The thing I’ll be watching from the outside: whether MePersona ships analytics, whether the platform roster gets published, and whether anyone builds the same memory layer into an existing suite. Because the memory idea is good enough that Buffer or Sprout will eventually copy it, and when they do, the question becomes whether a solo founder can out-execute a company with a sales team. My bet is that the niche — creators who genuinely care about fan relationships and have too many of them — is real and underserved. Whether this is the tool that serves it is a question the next six months will answer, not the launch thread.
