Connect with the Product Hunt community

There is one global CLAUDE.md that every project follows, plus one per project. Works most of the time. Not always. After a long session the agent drifts and does exactly what the file says you are not to do.

So I put anything that must never happen into a hook. There are 62 of them at the moment that run before the command and just stop it.

They include things such as DROP, rm -rf, editing files directly on a production server or clicking into a password field on the browser. The permission list contains 71 allows and 25 denies on top of that.

I m building SmartCheck a browser-based tool that counts similar objects in a photo after you tap just 3 examples.

No model training. No app install. Just open it in your browser, tap 3 objects, and let AI do the counting.

The one where a big grouped inbox stops making you wait a full minute to open an email. Turns out asking SQLite to rescan the whole folder once per message is bad for business. Who knew. (We should have.)

Faster

• Grouped inboxes are fast again. Picking the newest message in each thread used to walk the entire folder once per message. That's quadratic, and on a 15k-message inbox it took over a minute, and every other read waited in line behind it. So after launch you got a blank list and a nice long think. One new index later, the same query takes 68 s 20 ms. A test now pins the query plan so it can't quietly come back.

• Inline images no longer freeze the window while a sync is running. They're now loaded off the UI thread, where they should have been all along.

New

The new Lingo.dev is here, and it is live for every organization.

Lingo.dev is completely redesigned around the way you work, so everything now has a clear place, across five areas:

• Home  your starting point: a quick read on recent activity.

• Infrastructure  where you build your localization engines: the models, glossaries, rulesets, and brand voices you configure once and reuse everywhere.

• Studio  where localization connects to the tools you already use: the GitHub App, a Zendesk Help Center, a Payload CMS, plus the CLI and the API.

• Queue  one inbox for everything that needs a person: human review, localization engine suggestions, and triage, in a single list you filter and save as views.

• Observability  where you see whether the output is good, even in languages you cannot read: reports, AI reviewers, and a log of every review beside the translation it scored.

This is the foundation for everything we build next, and the new home for your localization work.

Netfox has always found devices by listening. This release lets it ask.

Your network has things on it that never say a word. A printer asleep. A camera that only ever talks to its own cloud. A Pi you set up in March and forgot. They are powered on and perfectly reachable, and until now Netfox could not show them to you not because it was hard, but because everything it knew came from whatever announces itself, or whatever has already spoken to your Mac.

Most of the time the people we build for are abstract. Personas, segments, users.

This week I set my mom up with something my team built and she caught domething she would have missed otherwise. Small thing. But watching someone you look up to get real value from your work hits different than a dashboard.

Hi everyone, I'm Taissa Maleh with 1752vc. We are a top GTM-focused accelerator in California writing $100k checks in startups at the inflection point of founder-led growth to building repeatable & scalable processes.

Lightning Round IX finals is on October 8th. It's our signature pitch competition where the winner gets a shot at $100k investment thru Accelerate. Applications are open until September 25th. More info at 1752.vc/lightning-round

A monthly active user metric can look completely stable while underlying product usage slowly drifts into a passive state.

Realizing that logins do not equal active value creation made us completely rethink how we track long term retention.

For founders looking past the initial growth phase: what retention metrics actually give you an honest signal about product health?

octoscope 0.35.0 is out and this one started as a small feature and ended as a bug I am slightly embarrassed about.

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?

1 days ago we launched ScreenCursor and hit #4 product of the day.

I wanted to take the time to thank everyone who upvoted and commented on the launch.

An agent can have the correct command and dependencies, then still fail because Postgres wasn t ready when an integration test started.
Teams often handle this with a README instruction: Start Postgres first. But developers, CI jobs, and agents can easily miss it.
With Ota, the dependency lives in the execution contract:

requires_services: - postgres

Ota can start the declared service and check its readiness before running the task. This creates one consistent, reviewable execution path for humans, CI, and AI agents.
Want to see it in practice? Request a free demo at demo@ota.run.

Hey everyone

Small release, but it's the kind I find interesting: nobody broke this one. It broke by itself, the day macOS 27 shipped.

Yesterday we launched Hello Inbox and hit #3 product of the day.

I wanted to take the time to thank everyone who upvoted and commented on the launch.

Read original article on - https://forg.to/articles/forget-...

A quick note from the maker: ScrollMark is $5 USD for Safari on Mac.

Keep losing a useful answer in a long chat or article? Press Option + 1 to save your spot, then 1 to jump back. You can save up to 10 spots per page. No account needed.

Get it here: https://apps.apple.com/us/app/sc...

Which of these is you?

1. Pre screen-share / demo panic: thirty windows, notifications armed, everyone waiting

2. End of the day: shut everything so tomorrow starts clean

3. Before gaming or a recording: free the machine

4. Focus reset: close the tabs I've been "coming back to" for a week

5. None. I upvoted because it looked neat and I won't use it

And the one I really want: if you looked and did NOT install, what stopped you? Closed source, telemetry, "I can just Cmd-Q", two clicks instead of one, no Spotify exception, SmartScreen warning, something else. Say it plainly. I would rather build the version you'd actually keep than defend this one.

I m curious how you d push Dictantor beyond just recording and transcription.

The obvious things are already on the radar better organization, smarter summaries, integrations, and improvements to the recording experience.

But I m more interested in the less obvious use cases and features.

Is there something you d want Dictantor to do with your recordings or transcripts that would make it genuinely useful in your daily workflow?

We launched dif on Sep 5 and the most-asked question in the comments was some version of what does the agent part actually do? So here is a bit more context:

dif init writes a managed block into AGENTS.md, CLAUDE.md, and .cursorrules. The block tells the agent where flag files live, the frontmatter schema, the dif("flag-id") call at the render site, and to run dif validate before opening a PR. It also installs three Claude Code skills into .claude/skills/: one proposes the surfaces in your app, one drafts a flag or test, one concludes it and writes the learning back to the repo.

Every dif build refreshes dif/context.json, so the next session starts knowing what is live and what already got killed.

Pick what you want with --agents claude, --agents cursor, --agents general, or --agents none.

@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:

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.

Bloated files are the worst part of maintaining a growing project.

We tend to hold onto old code because it feels like throwing away actual work, even when nobody has used that feature in months.

Cleaning up dead code always makes the app faster and cuts down build times, but it is super hard to convince the team to delete it.

Have you ever deleted a large chunk of code only to realize later that you actually needed a piece of it back?

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.