Jellynet.net

Introducing JellyNet 🚀 The API marketplace where everyone wins. → 40 providers → 12 categories → One universal key (jn_xxx) → Suppliers earn 60% from idle keys → Buyers save up to 60% → AI agents pay per call via x402 Live today on @ProductHunt 👇 producthunt.com/products/jelly…

@AmitSingha89 months just to standardize terminology before AI could even touch the data — that's the part nobody talks about. the "AI replaced my workflow" crowd skips straight to the output. the unglamorous data foundation is the actual work.

@jellynet_ The hardest part is digitizing that physical reality. A model can't feel the tension of Jamdani yarn on a loom. We spent months just standardizing vernacular Bengal weaving terms before AI could even process the data.

345 automated tests. 40 LLM providers. 24-hour proxy uptime. Built by a non-developer CEO using AI tools. "You need a technical co-founder" is outdated advice. You need taste, persistence, and the right AI stack 🔥

the hook. the current first slide is too product-led — features before feeling. rewriting it around the pain first. "your API keys are wasting money right now" lands harder than "meet JellyNet" ever will. emotional framing underneath stays the same — the freedom angle is right, just buried too deep.

yeah actually sense. Ph momentum kinda acts like borrowed trust early on so when social proof is missing users scrutinize positioning way harder. what part of the relaunch assets are you changing most aggressively first slide/hook itself or the emotional framing underneath the whole sequence?

We launched on Product Hunt yesterday. #12. 20 upvotes. Not featured. Zero community prep. Cold launch. New account. Lesson: PH rewards momentum, not products. Relaunching mid-June with actual prep. Building in public means sharing the Ls too 🔥

Things that broke this month: - Data migration crashed prod (forgot IF EXISTS) - Rate limiter double-counted retries - Google banned our admin email - PH launch flopped (#12, no prep) Still shipping. 345 tests now prevent 3 of those 4 💀

The average API subscriber uses 60% of their quota. The other 40% expires monthly. Poof. Gone. On JellyNet, that 40% earns you money while you sleep. Bring your keys. Use them yourself free. Sell the surplus. jellynet.net

How JellyNet routes your API call in <200ms: 1. App sends to jn_xxx key 2. Router detects model + provider 3. Weighted-random picks from supplier pool 4. Call executes on real API key 5. Fails? Silent retry (up to 3x, <9s) You never touch a provider dashboard again.

@AmitSingha89 "AI as the translator" is exactly it. the model doesn't know what a good handloom weave looks like. you do. that gap is the actual moat and it only widens as AI gets more accessible to everyone else.

@jellynet_ I built Anuprerna's supply chain tech stack the exact same way. The real moat has shifted from writing code to domain expertise. If you deeply understand the offline artisan ecosystem, AI just becomes the translator.

@AmitSingha89 phone photos from the loom straight into a defect classifier is genuinely impressive. software regressions are deterministic — physical quality variance is a completely different class of problem. curious what your false positive rate looks like on the fabric detection.

@jellynet_ Replacing QA with AI is the logical next step. We're testing computer vision to spot fabric defects via phone photos straight from the loom. Software regressions are predictable; standardizing physical handloom quality is brutal.

physical regressions are a whole different beast. a monsoon isn't a bug you can reproduce in staging. honest answer: our AI QA handles predictable software failures well — wrong response codes, latency spikes, auth errors. unpredictable human inputs still need human eyes. the 345 tests are a floor, not a ceiling 👀

@jellynet_ Replacing QA with AI is a massive shift. In artisan supply chains, our regressions are usually physical events—like a monsoon delaying a loom—that cascade into the data. Does your AI QA layer handle unpredictable human inputs well?

@Razumasu this is the part most people miss. AI compressed the execution gap, not the taste gap. knowing you need failure assumptions, retry logic, and test coverage — that's still on you. AI just stops being the bottleneck between the idea and the code.

@jellynet_ The interesting bit is not “non-developer shipped software”. It is that the stack has tests, uptime thinking, and provider failure assumptions. AI can compress coding, but it does not remove the need to recognize what good engineering looks like.

@ApplyWiseAi brutal but accurate. the algo decides in the first 2 hours. if your day-1 supporters aren't pre-warmed, you're invisible before most people even wake up.

@jellynet_ cold launches expose this every time. without early supporters the algo buries you before day one ends

@AmitSingha89 exactly — domain knowledge is the moat, not the tech stack. anyone can learn to code, almost no one understands rural artisan supply chains. maintenance is maybe 20% now. 345+ tests catch most regressions automatically. the AI stack basically replaced the QA layer 🔥

@jellynet_ I built Anuprerna's supply chain platform with AI for this exact reason. Knowing the nuances of a rural artisan cluster is much harder to learn than Python. Curious how much time you spend on maintenance vs building new features now?

You're paying for 7 different API keys. 7 dashboards. 7 billing pages. 7 rate limit docs. One breaks at 2am and you're debugging which provider throttled you. We route across 40 LLM providers with one key. jellynet.net

@DesignByMaeL 100%. PH is social proof arbitrage — early momentum signals quality regardless of actual quality. yes on both. tightening the assets + sequencing the outreach 2 weeks out this time. first hour is everything.

@jellynet_ cold launching with zero prep exposes perception side of Ph harder than the product side itself. When momentum doesn’t form early users subconsciously assume the product matters less. are you changing the launch assets/sequencing too for the relaunch or mostly distribution?

@jtkeepsmore exactly this. we got filtered hard — new account voters just don't count. warming up the list for 2 weeks before the June relaunch. PH is a game, not a meritocracy 🎯

@jellynet_ PH is basically a hype contest dressed up as a launch platform. cold accounts get punished hard. relaunch with a warmed up list and it's a different game

Next week on @jellynet_ 📣 → Screen recording of the full integration flow → First supplier earnings report → Deep dive on x402 micropayments for AI agents → Solana Frontier judging countdown Follow so you don't miss it 🔔

I'm an indie founder building an API marketplace solo. My CTO is Claude. My test suite is my safety net. My prod env is my only env. 34 sessions · 262 tests · 40 protocols · 4 prod incidents recovered If you're building alone: tests are not optional 🛡️

Every dev is paying a hidden "key management tax" Multiple providers = multiple keys = multiple billing pages = multiple rate limit strategies The answer isn't picking the cheapest provider. It's making provider choice irrelevant 💡

W3 building JellyNet in public 🔧 ✅ Protocol detail pages (40 providers) ✅ Dashboard overhaul + key management ✅ Marketplace browsable without login ✅ 380 automated tests Next: → Email login → Stress test key rotation → More protocols Every week: what shipped + what broke.