I've started to notice something about mature products.

The size of a feature rarely determines how much work it takes.

Instead, the effort depends on how many past decisions that feature touches.

A simple "Add Google Login" suddenly affects authentication, onboarding, permissions, analytics, documentation, QA, and maybe even your pricing model. What looked like a two-hour task turns into a week of discussions because you're no longer building a feature; you're negotiating with decisions your product has accumulated over months or years.

Hi Product Hunt! I recently launched Danmu Bilingual Keyboard on the China App Store.

The idea is simple: when typing Chinese with Pinyin, English glosses appear alongside the Chinese candidates, so you can pick up vocabulary without switching to a dictionary.

The keyboard works offline for Pinyin input and glosses, and keeps its learning data on-device. It doesn t require Full Access. The companion app includes a vocabulary notebook with manual editing and JSON import/export.

A little about the build: I brought a platform-independent Rust input core into an iOS Keyboard Extension, then built the surrounding app in Swift/UIKit. The extension has strict system boundaries, so keeping input and learning local became an important part of the design.

My team has delved into the world of CX for over a decade and what truly makes it complete is actually the human agents behind the scenes, teams engaging to answer customers (the heartbeat of any business) all from one site instead of swinging one tab to another!
The launch is within a week! And your voice, support, presence and even a try on for feedback means a ton!!!

Hey everyone!

I built JSONexLab 7 B2B APIs for trade, finance, compliance, all on the same stack (Cloudflare Workers + D1 + KV + R2).

The problem: small trade/fintech teams spend $500-2000/mo on separate vendors for HS codes, sanctions screening, document parsing, finance validators. I bundled them into one platform with free tiers.

What's live:

Introducing ReversePoll a better way to run community Q&As.
Ever wondered how creators can make sure the most important questions don't get buried in a flood of chat messages?
ReversePoll lets communities prioritize questions through voting and contributions, helping creators focus on what their audiences actually want answered.
Built for streamers, creators, and online communities.
Check it out: https://reversepoll.com

demo video:

Every architecture diagram I've drawn by hand was out of date within a week.

So I built TF Diagrams, and today I launched it on Product Hunt

It reads your Terraform and draws the AWS / GCP / Azure diagram for you, in your browser. It's free, there's no sign-up and nothing gets uploaded.

Hi Product Hunt community!

I m building MindBridge, a privacy-first AI assistant for macOS and Windows, designed to bring AI closer to your desktop while keeping core inference local through Ollama.

Recently, I ve introduced several new features:

Hey everyone!

If you re building a SaaS or a side-project, setting up payments usually means hours of dealing with webhook routing, customer sync logic, and getting locked into one provider.

I built Paycan to fix this. It s an open-source, self-hosted checkout layer.

• No Vendor Lock-in: Switch between Stripe and PayPal without changing your database or application logic (more local payment providers soon).

• Drop-in Simple: Around 10 lines of code to integrate.

• Self-Hosted & AGPLv3: You run it on your own server, keeping full control over your checkout.

Hi everyone, I'm Jim, an independent iPhone developer. Time Debt launches on Product Hunt tomorrow, Oct 17.

Borrow ten extra minutes today, and tomorrow's allowance starts ten minutes shorter.

- An app blocker built on Apple's Screen Time system: set a daily allowance for chosen apps, categories and websites.

- Out of time? Borrow 5, 10 or 15 minutes and see today, tomorrow's normal allowance and tomorrow's adjusted allowance before you confirm.

So here s what happened. I spent over a decade in IT project management, and I kept seeing the same thing: a project reports green on schedule and budget, leadership feels good, and months later it misses the outcome it was approved for.

I wanted to know if it was just me. So I interviewed 20 senior leaders in 8 countries. The research was published in PM World Journal. 15 of the 20 described the same pattern without being prompted.

The status wasn t wrong. It was answering the wrong question.

Leaders were spending 20 to 60 hours every reporting cycle assembling updates. One told me that even with all that reporting in place, he still re-checks the numbers himself because he can t trust what reaches him. Others described teams quietly holding back red statuses until they couldn t anymore.

Most project tools track whether work is on schedule. They don t ask whether the project will still deliver what it was approved for.

So we built something different. DaytaBrain watches the assumptions behind each project, things like staffing will hold or adoption will hit 60%, and ties each one to the numbers that would show it breaking. When capacity drops 18% and velocity drops 15%, it doesn t just show two red numbers. It tells you the staffing assumption is no longer holding, what that puts at risk, how many weeks you have to act, and what to do next.

No prompts, no chat. You open a project and know in about 5 seconds whether it s really OK.

It s called DaytaBrain. I m looking for 50 people to try it free for 30 days on 2 real projects, every feature included, and tell me honestly what s missing.

Would love to hear if you ve seen a green project fail. What did the status miss?