Zyaku
Zyaku is building an AI Operating System
We're building something at Zyaku that we don't want to announce as finished until it actually deserves the word.
Today I want to share where we really are with Zyaku Mesh - not the marketing version, the engineering one.
Mesh is our long-term attempt to build a communication layer where devices can participate in a more resilient, distributed network instead of assuming every interaction needs a continuously available central connection. That's not a matter of putting "P2P" on a roadmap - there's a much harder foundation underneath it. So we started with the parts that can't stay vague: identity, authorization, acceptance, risk, policy, decision, persistence, recovery, lifecycle, deduplication, validation. Those aren't slides in a deck. They're executable parts of the current Mesh kernel, with 94 of 94 registered test cases passing within their current bounded scope.
Then came authenticated transport. We've spent real engineering effort characterizing EDHOC 2.1.1, CBOR behavior, and AES-CCM reference behavior, instead of assuming a protocol that works in one environment automatically works everywhere. That work changed the actual question. We're no longer asking whether the cryptographic protocol can work. We're asking how to make this known-good protocol stack behave correctly inside the real runtime Zyaku needs.
That's where it gets interesting. Our Metro/RN experiments have already surfaced a real crypto dependency blocker. The code can be correct, the protocol can be correct, the reference implementation can be correct, and the target runtime can still say no. That's exactly the kind of problem we want to find now, not after calling something production-ready.
Here's the part I want to be especially clear about: Zyaku Mesh is not live yet. The peer-to-peer data path still has major engineering work ahead - signaling, WebRTC, data channels, synchronization, recovery, conflict handling, cross-peer behavior. Those components currently have architectural definitions, but not yet the behavioral runtime engine a production Mesh actually needs.
So I'm not posting that Mesh is almost done. I'm not posting that decentralized communication is solved. And I'm not manufacturing a percentage to make a chart look impressive.
The honest picture is more interesting than that. The foundation is real. The authenticated transport research is substantially characterized. Mobile runtime integration is the immediate battle. And the distributed data path is still ahead of us.
Our current planning range for a production-ready Mesh is roughly 11 to 23-plus weeks, assuming a focused engineering stream and no major architectural reset. Could it move faster? Possibly. Could something force us to redesign part of it? Absolutely. That's why it's a planning range, not a promise.
I don't think building in public means announcing every milestone as a victory. Sometimes it means showing people how far you still have to go.
We're not claiming Zyaku Mesh has solved resilient peer-to-peer communication. We're building toward it - one verified layer at a time, one failed assumption at a time, one runtime problem at a time.
The goal isn't to make Mesh sound revolutionary. The goal is to make Mesh actually work.
That's the work.
If anyone here has shipped a P2P/WebRTC stack to production, especially on React Native - I'd genuinely like to hear what broke for you and how long the Metro/dependency side actually took versus your original estimate. That's the part of this I trust the least right now, precisely because we haven't lived through it yet.
In March we launched Zyaku on Product Hunt as an all-in-one creator platform. It ranked #187 for the day. Not a disaster, not a win. The honest read was that "all-in-one creator platform" told nobody why they should care.
Today we're back on Product Hunt as Zyaku 2.0, and the product is different: offline-first social infrastructure. Creators and journalists publish, earn and communicate — and it keeps working when the internet doesn't.
Why we changed direction
The signal came from where our users actually are. We have people in 100+ countries, and a lot of them are on connections that drop for minutes or hours at a time. The complaint was never "I want another feature." It was "I wrote the post and it disappeared." That is a reliability problem, not a feature problem, and it pointed to one thing everyone else treats as an edge case: the network going away.
What we built since March
Offline publishing: posts and messages are composed and queued on the device with zero connectivity, then published exactly once when the network returns. No lost drafts, no duplicates. This was the hardest thing we have shipped — exactly-once on reconnect is where most of the engineering time went.
Offline Emergency Access: PIN/biometric-gated emergency tools (broadcast, safety check-in, cached contacts, crisis info) that work with no network at all.
Disaster Alert broadcast, live in production since 1 September.
Clips with an in-app editor, Motion Comics, AI video.
Creator payouts in 100+ countries — Stripe Connect where it exists, manual payouts where it doesn't. 80% revenue share, no follower minimum.
Where we are
Two people, fully bootstrapped, no outside money. Live revenue, still under $100K. Users in 100+ countries. Everything runs on a Node/Postgres backend on Railway that auto-deploys from main, because with two people the ops budget is zero.
What I'd tell March-me
Lead with the thing only you do. "All-in-one" is a list; "works when the internet doesn't" is a reason. Six months of building went into earning the right to say the second one.
The demo was recorded on a real phone in airplane mode, start to finish: https://www.youtube.com/watch?v=FL2SZ2bZbi4
Product Hunt launch, live today: https://www.producthunt.com/posts/zyaku-2-0
Happy to go deep on the offline queue, how exactly-once works on reconnect, or payouts in countries Stripe doesn't cover. Ask anything.
I’m starting to think the most dangerous phase for an early startup is not when nobody uses it. It’s when a few people use everything. They give feedback on multiple features. They explore the product deeply. They find things you didn’t expect them to find and suddenly, you have a problem, you can’t tell whether you’re seeing product-market fit or simply the behavior of highly curious early adopters.
That distinction matters.
If one person uses five different parts of your product, should you:
Double down on the feature they use most?
Build around the workflow connecting all five?
Or ignore the data until hundreds of less-motivated users show the same pattern?
Right now, I’m learning that usage is not the same thing as validation.
A click can mean curiosity.
A session can mean exploration.
Even repeated use can sometimes mean someone is trying to understand what your product is.
The signal I’m starting to trust most is much simpler:
What do people come back to when they have no reason to explore anymore?
That’s probably where the real product begins.
Curious how other founders think about this:
At what point do you trust early user behavior enough to make a major product decision around it?
What Zyaku actually is now
I didn’t set out to build an “AI Operating System.” I set out to build a fairer creator platform - 80% revenue share, zero follower minimum to earn, real ownership over what you create. But somewhere along the way, the problem got bigger than “creator economy.” We ended up building out identity, commerce, publishing, AI tooling, and resilient infrastructure - all connected, not bolted together.
One thing I’m genuinely proud of: we just shipped Disaster Alert - an offline-first emergency communication system. It works even when the internet doesn’t. Local broadcast queuing, biometric auth, cached safety info - built so people can still reach each other when networks fail. It’s part of a broader “Wellbeing as Infrastructure” idea I’ve been pushing internally - that platforms should protect people, not just engage them.
The honest bootstrapped-founder stuff
• No funding means every infra decision is a real tradeoff - we migrated off Replit/Neon to Railway + Cloudflare partly for cost, partly for control
• Building solo (then with 2 people) across 10 product domains sounds insane in hindsight, and some days it still does
• ~2,400 registered users so far, live creator payouts running, and real coverage from international outlets including AP News - earned the hard way, not bought
Why I’m posting this here
I’ve mostly been heads-down building. I want more honest feedback from people who’ve actually shipped things solo/bootstrapped - what worked when you tried to do “too much” at once, and how you knew when to cut scope vs push through.
Happy to answer anything about the stack (React Native/Expo, Node/Express, Postgres/Redis on Railway), the offline-first architecture, or just the mess of building solo without funding.
zyaku.com if you want to poke around.
-
12,400 users across 10 domains is a lot of surface area. What’s the one part users keep coming back to most?
-
1
That’s exactly what we’re watching closely. Right now, the strongest signal is around the core workflows where users can actually get something done, rather than just explore features. We’re still early, so I’d rather let the repeat behavior tell us what deserves to become the center of Zyaku
-
1That’s the right thing to watch. I’d be curious which workflow is actually showing repeat behavior already, even if it’s still too early to call it the center of Zyaku.
-
1
I think the strongest signal may be broader than one workflow. The interesting behavior is when people start using Zyaku for things they’d normally need separate apps for creating, publishing, managing identity and earning, but also the safety side: Safe Space and our offline-first emergency and disaster features.
The goal isn’t to make users switch between more apps. It’s to see whether one connected system can become useful enough that they keep coming back for different parts of their digital life.
-
1That’s an interesting distinction — you’re really testing whether the value comes from any single workflow or from the system becoming useful across someone’s digital life. I’d be interested in digging into that a bit more privately. What’s the best email to reach you on?
-
-
-
-
Hey Indie Hackers! 👋
I'm Rohit, founder of Zyaku — and I wanted to share what we're building and why.
---
THE PROBLEM
Social media promised creators the world. Instead it delivered:
❌ Shadowbans with no explanation
❌ Algorithms that bury real voices
❌ 0% revenue share for creators
❌ Bots inflating fake metrics
❌ Opaque moderation with no appeal
72% of social media users want alternatives. The creator economy is growing 35% YoY. But no platform has truly solved this.
---
OUR SOLUTION: ZYAKU
Zyaku is the All-in-One Creator Ecosystem built around 3 words:
Create. Grow. Earn.
Here's what makes us different:
🤖 DUAL AI ENGINE
First platform to combine OpenAI GPT + Anthropic Claude with real-time model switching. Smart replies, content rewriting, thread generation in 15+ languages.
💰 CREATOR MONETIZATION
Creators keep 80% of revenue. Built-in subscriptions, tipping, digital marketplace, and instant payouts — no third-party tools needed.
✅ HUMAN VERIFICATION
Multi-tier identity verification eliminates bots completely. Your follower count actually means something on Zyaku.
🎨 COMICS ECOSYSTEM
Static comics, motion comics with parallax animations, short animated content, and storyboard tools — all built in.
🔒 TRANSPARENT MODERATION
No shadow bans. No hidden rules. Users always know why action is taken and have the right to appeal.
💬 REAL-TIME MESSAGING
Socket.IO powered chat with typing indicators, read receipts, and end-to-end encryption.
---
THE NUMBERS
📊 2,400+ creators already on waitlist
📊 5.2B global social media users
📊 $250B social media ad market
📊 35% YoY creator economy growth
---
WHERE WE ARE
We're currently in pre-launch with 2,400+ creators on our waitlist. Building in public and would love feedback from this community!
🔗 zyaku.com — join the waitlist
📝 Full story on Medium: https://medium.com/@rohit.kashyap_53123/zyaku-the-all-in-one-platform-where-creators-build-empires-e60a291e8f75
🚀 Product Hunt launch: tomorrow!
---
QUESTIONS FOR THE COMMUNITY:
1. As a creator, what's your biggest pain point with current social platforms?
2. Would you pay for a platform that gives you 80% revenue share?
3. Any feedback on our positioning?
Would love to hear your thoughts! 🙌
-
1
Hey Rohit — came across Zyaku and the 2,400+ people already on the waitlist caught my attention. It seems like you’re building something pretty ambitious, especially with the whole Create. Grow. Earn. idea.
I’ve been curious about how founders actually decide what to work on next when there are a million things competing for attention. Do you usually go with gut instinct, user feedback, data, or some combination? Genuinely curious how you approach it.
-
1
Appreciate that and honestly, I don’t think the answer is just gut instinct, user feedback, or data on their own. At Zyaku, we try to start with a simpler question what is actually limiting the product right now? Sometimes the data tells us. Sometimes users make it obvious and sometimes neither gives you the answer yet, because you’re building toward something people can’t fully articulate before it exists. That’s where founder judgment matters, but I’ve also learned that not everything competing for attention deserves to be worked on. A good idea can still be the wrong priority if it doesn’t strengthen the core product or move the larger system forward. So it’s really a combination user feedback and data help us understand reality, while instinct and first-principles thinking help us decide where Zyaku should go next. The hard part isn’t finding things to build. It’s being disciplined enough to decide what not to build yet.
-
Comment