Connect with the Product Hunt community

@ArkTabs V4 is Coming!

@ArkTabs is spotlight for used by 7500+ users on Chrome, Edge, Brave and other chromium based browsers. In new update there are massive upgrades:

I'm three weeks from launch and I have a list of fourteen things I could still build. I know I can't build all of them. What I don't have is a rule for deciding.

The rule I've been using "cut anything that isn't core" is useless, because everything feels core when you're the one who built it.

What I'm trying to get to is a single non-negotiable: the one thing that, if it's missing, the product isn't the product. Everything else is a candidate for cutting.

So for the solo makers here: what was your one non-negotiable thing at launch, and what did you cut that you thought you couldn't? And looking back, did cutting it hurt, or did nobody notice?

Hey Product Hunt community, today we released a major Guitar Wiz update, available day one for iOS 27.

We focused on everyday practice tasks, especially identifying chords and matching a song's tempo.

Last month I tried 5 AI based tools for different tasks. I kept 2 and dropped 3. The 2 I kept had the same pattern: they gave me suggestions but I made the final call. The AI didn't try to do everything, it did the boring part and I did the deciding part. The 3 I dropped either did too much automatically and got it wrong with no way to correct or did too little to be worth opening.

My read is that "human in the loop" isn't a feature it's a requirement. People say they want full automation but in practice they want control.

So for those who use AI tools daily: do you actually prefer full automation or do you want the tool to stop at "suggestion" and leave the decision to you? And the harder question: if a tool automated 100% of the task correctly every time, would you trust it or would you still check?

The SODAX oracle now has a public candle endpoint. GET https://api.sodax.com/v1/be/orac... returns USD-quoted OHLC data for 55+ symbols across four intervals (1m, 5m, 1h, 1d). No auth, no API key, no SDK dependency.

GET /v1/be/oracle/candles?symbol=BTC&interval=1h&from=1785952800&to=1786039200

These are the same prices solvers quote against and the money market uses for loan pricing. If you're building a chart and pulling from CoinGecko or DefiLlama, your chart will show one number and the contract will settle at another. That's not a rounding issue, that's your UI lying to the user. This endpoint gives you the actual oracle price.

Bound Exchange needed chart data for their UI and we had a choice between building a separate chart API against a market feed or exposing the oracle we already use internally. The market feed option had exactly that problem, so we went with the oracle.

From September 2026 you can't submit an update without saying whether your app has social media capabilities. Say yes and Apple applies a minimum 13+ rating and drops you into a Time Allowance parents can cap. The definition is broader than having a feed, and your secondary category is load-bearing too.

If you've shipped an update in the last few days you've already met this question. If you haven't, you will on your next one, and I'd rather you met it here than at 11pm with a release branch open.

The growth example hits home because AI is terrifyingly good at agreeing with whatever hypothesis you feed it.

If your starting prompt is biased, the tool will just find data to make you feel smart about a wrong decision.

True collaboration only happens when you force the model to look for counter-arguments or blind spots you missed. Have you found a reliable way to get AI to challenge your strategy instead of just echoing your thoughts back to you?

Dictantor can now become a part of your AI workflow.

With the latest update, you can give your AI agents access to a folder containing all your Dictantor transcripts. So recordings from meetings, voice notes, and random thoughts can become context your agents can actually work with.

Transcripts also got a small structural upgrade to make them easier for both humans and agents to read.

Next, I m working on custom folders inside Dictantor. The idea is to keep recordings organized by project or topic, while also giving agents a cleaner way to understand and use the right context.

Been working on this update for a while, and I wanted to share what s new in 3.0.

Hey everyone

Netfox 0.22.0 is out, and it is about one thing: the window finally behaves when you tile it.

It used to refuse to go narrower than 960 points. So if you put it beside the thing you were actually working on, its right edge went off the screen. It was an app you opened, glanced at, and closed

You rename packages/cli to apps/cli. The next scan tells you a dependency started running code at install, and another one stopped. Nothing started you moved a folder.

0.34.0 is four changes that are the same sentence from different angles: a security signal is only worth reading if it fires on what happened, not on what you did to your own repo. An axis that cries wolf is one you learn to skip, and then it's worth nothing on the day it's right.

You ask for a quick script addition, and suddenly background logic, helper functions and folder structures are completely rearranged without permission.

It feels less like a coding assistant and more like a coworker who won't stop uninvited cleanups.

How do you lock down your agent workflows so they only touch what you explicitly point to?

Vibe coding a personal project that works on your laptop is one thing.
Building a tool your team opens every morning to manage projects, serve customers, and run the business is another.
Suddenly there s shared data.
Different people need different levels of access.
Your teammates need to understand how the app works so they can change it too.
And the app needs to keep evolving without turning into a black box nobody wants to touch.
At that point, would you be comfortable letting AI change anything it wants?
We ve been thinking about this a lot while building WeWeb MCP (launching on PH tomorrow!)
Agents like Claude, ChatGPT, Cursor, and Gemini can now build apps directly inside WeWeb.
Across the frontend & backend from: UI, data, workflows, integrations, permissions, and more.
But the important part for us was making sure the builder stays in control of both the app and the AI building it.
That s why we built guardrails into the way AI works inside WeWeb:

• Before AI builds: choose what AI can and can t touch: pages, components, workflows, tables, APIs, and more.

• After AI builds: see exactly what changed in the visual editor. Tweak it yourself, keep iterating with AI, and publish when you re ready.

So AI gets room to move fast, without getting free rein over the whole app.
I m curious how you think about this: Would you give an AI agent access to everything?
Or are there parts of your app you d always want to lock down?

As Termexo s maintainer, I am focusing on a practical problem: several agents can be working while the developer is still finding terminals, checking who needs approval and reconstructing task context.

For individual developers, Termexo groups real terminals by project, surfaces agent status, organizes tasks and helps resume supported native sessions. Handoff packages can carry a reviewed summary of progress and next steps.

Hey everyone

Netfox 0.21.0 is out, and it is mostly one feature plus a confession.

@Youkti helps you track exactly where your deals stand.

From early conversations to active opportunities, get the context you need to know what s happening, what s changed, and what to do next.

Stalled deals can come back to life. @Youkti analyzes the context around each deal and recommends the next best steps to help you re-engage and move it forward.

Everyone talks about how fast AI can write code now but nobody talks about the mental exhaustion of reviewing three different feature branches at the same time.

I realized last week that if I run more than two agents simultaneously I stop catching small architectural mistakes and just start blindly approving diffs.

Now I cap myself strictly based on how many code reviews I can realistically handle before my focus drops. Do you use any specific workflow or rule to keep yourself from drowning in unreviewed AI-generated code?

I built something no customer requested. No data pointed at it. No one had mentioned it in a support ticket.

I built it because it annoyed me that it didn't exist.

I've tried Reddit posts, a few Discord communities, and cold DMs. Most responses were polite silence. The few people who did respond gave feedback that was too vague to act on. "Looks cool." "Nice idea." Nothing that changed what I built.

The problem isn't finding people it's finding people who care enough to tell you something uncomfortable.

So for those who got past this stage: where did your first 10 users who actually gave actionable feedback come from? Not "post on Twitter," a specific channel, a specific approach, a specific thing that worked. And the harder follow up: what did you do with the vague feedback before you found the useful feedback?

AppGacha 0.1.1 is now available for Windows and Apple Silicon Macs.

This is a small release focused on making generated apps feel more independent and improving the update experience:

• Opening an app from its desktop shortcut no longer brings up the main AppGacha window.

• Generated app windows now display their own icons, with a fallback when no icon is available.

• Settings now includes version history and localized release notes.

• Update prompts explain what changed, and installation waits for the current build to finish.

• Added minimal, content-free usage statistics to help identify failed builds and repeat usage. Collection will not begin before September 26, can be disabled at any time, and never includes prompts, generated code, API keys, files, app names, account IDs, or persistent device identifiers.

Thank you to everyone who tried the first release. If something still feels confusing or gets in the way of using a generated app like a normal desktop application, I d love to hear about it.

We re working on making Ski more comfortable to use for people who don t speak English all the time.

The latest update adds more control over both voice input and voice output, with platform-specific improvements:

• macOS: More languages available for voice input in High and Ultra modes

• Windows, Linux: Ultra (default) mode can automatically detect the language being spoken

• Voice output: We re expanding the number of supported languages, and currently supports few languages in Kokoro. Each language has its own voice list.

• Audio devices: You can now choose your input microphone and output speaker/headphones directly

• Easy updates: Get the latest version through automatic updates or download it directly at heyski.io

We re still expanding the language list, especially for voice output.

Something I keep running into with vibe coding: the agent makes a failing test pass by loosening it instead of fixing the code.

One example: it replaced a strict deep equality check with expect.anything() on an error payload. The test still exists, still runs, still passes. It just stopped asserting anything. The build stayed green for days until users hit silent null-pointer crashes.

The same can happen with CI workflows and migrations. If the agent can edit them, it can make the gate disappear.

One cheap guardrail I ve seen: a CODEOWNERS rule requiring human review on tests/, migrations/ and .github/workflows/, so any diff there needs approval no matter how green the build is.
I d love to hear how you handle this:

Voice agents should become more than transcription tools. Instead of simply turning speech into text, they should understand what users are trying to achieve. The same spoken idea may need to become a short Slack message, a detailed email, a technical issue, or a personal note. A useful voice agent should adapt automatically.
This is the direction Loqua is exploring: voice as an intelligent layer across the apps people already use. It can understand the current application, nearby content, personal vocabulary, and writing style, then shape spoken thoughts into something ready to use. Users should be able to speak naturally without organizing every sentence beforehand.

The next step is real-time duplex conversation. A good agent should know when someone has finished speaking, when they are only pausing, and when they want to interrupt. It should also recognize signals beyond words, such as emotion, hesitation, emphasis, and tone. Its own voice should feel equally natural, expressive, and appropriate to the situation.

Behind this conversation, the agent needs deeper capabilities. It should remember relevant context, think through difficult requests, use tools, and continue working after the conversation moves on. The voice experience remains fast and fluid, while more complex reasoning happens quietly in the background.
We all know that long-horizon coding tasks are difficult, but understanding long, multi-turn conversations is even harder. The agent must remain consistent, respond with the right emotion, use tools correctly, and remember what matters.

Hi everyone, Matheus here, Toone's maker. We're preparing our next Product Hunt launch for September 18

Toone is a macOS app for crafting complex AI workflow automations and routines with natural language. I'd like to hear from people who want to try one small routine and share candid feedback.

If you're interested, share the task you repeat, the inputs you'd start with, and what a useful result would look like. Pick something you can check yourself so we can see where the routine helps and where it falls short.

I can help arrange access individually, though I can't promise immediate availability.
Thank you and see you by then PH