Connect with the Product Hunt community

Thanks so much for all the support on signspell!

One thing building this taught me: almost every gesture/action recognition project ends up rebuilding the same infra from scratch capturing frames, labelling landmarks, exporting clean datasets for training.

July 2026 has been a turning point for search. Two major updates landed within weeks of each other, and they change how visibility works.

1. Google: AI visibility is now part of Search Console

In early July, Google launched generative AI performance reports inside Search Console. The data is now isolated: impressions from AI Overviews, AI Mode, and generative AI in Discover are separated from classic search. For the first time, you can see whether your content is actually being surfaced inside Google's AI answers not just rankings, not just clicks, but presence.

The message is clear: AI visibility is not a separate discipline. It's search visibility. Google put it inside the tool SEOs already use.

Personal, single use, the biggest discount Pulse will ever run.

• Open the email

• Tap claim

• Done.

Can't find it?

Reply here and I'll sort you out.

Your inbox now reads newest-first by default, subject lines are yours to control, and clearing unread mail takes one tap.

Your inbox, in the order you want

The inbox now opens as a simple newest-first timeline. Prefer the priority-grouped view that floats important mail to the top? It's still there now a choice in Settings General.

One thing I've learned while building a SaaS is that it's easy to make decisions based on what you think users need.

The difficult part is accepting that you're sometimes wrong.

We had several ideas that seemed obvious during development, but after talking to real users, the conversations went in a completely different direction.

Instead of asking for more features, many people cared more about clarity.

The one where Skim stops insisting the model is loading while it is, in fact, sitting there thinking very hard. Turns out silence has two meanings, and we only knew one of them.

Entirely the work of @gildas-refacto in #28 thank you!

Fixed

• "The model is loading " was a lie. Skim only ever watched for answer text. A reasoning model produces none while it reasons, so Skim saw silence and concluded the thing was still booting up. It now says "Thinking ", because that's what's happening.

• Not just a local-model problem. Current Anthropic models reason by default, so the wrong label showed up on the default provider too. Both streaming clients report reasoning now.

• The recap counter no longer freezes at "reading 20/20". It steps aside for "Thinking " the moment the mail is actually read, instead of staring at you until the answer lands.

• The loading hint no longer beats the thinking one by a single frame. A flicker so brief it was almost a rumour. Gone anyway.

The one where Skim stops being picky about which AI you talk to, and the whole app goes on a diet at the same time. Efficiency and choice, in one download. You're welcome.

New

• Bring your own AI endpoint (OpenAI-compatible). A third provider joins Anthropic and OpenRouter: point Skim at any OpenAI-compatible API Ollama, LM Studio, vLLM, your homelab gateway, whatever. Just a Base URL and a Model ID; the API key is optional (leave it blank for local servers that don't want one). No key = no Authorization header, so your Ollama box stops complaining. (Closes #12.)

• AI chat that can actually breathe. The command palette (Ctrl+K) can now expand to near-fullscreen for reading long answers one toggle in the corner, only in chat mode. Repeat Ctrl+K flips it open/closed instead of nuking the palette. It remembers your preference, and text stays in a comfy ~760px reading column instead of sprawling edge-to-edge.

Fixed

Happy Friday

A brick design is only useful if you can build it from the box you already have. v1.3.0 is about closing that gap tell BrickSolvr which bricks you own, call the colors what you call them, and the design comes back in your language.

AI can remember a fact.

But can it understand how that fact connects to everything else?

Just shipped a big one: Netfox is now fully localized.

New languages: Italian French German Spanish (alongside English) and it follows your Mac's system language automatically, no setting to flip.

We launched Honestly on Thursday and I wanted to again say that the support has been absolutely overwhelming. Thank you to everyone that has helped so far.
We placed #2 on Thursday and are now less than 10 upvotes away from securing the #3 spot of the week!
For anyone that wants to see David beat Goliath (aka the bootstrapped smaller startups beat the bigger companies), any support would be greatly appreciated!

since the last launch, the goal was one thing: make FetchSandbox something you can actually gate a release on. not "looks tested" but provably tested.

what changed:

Hey everyone,

If you've been following our journey, we'd be incredibly grateful if you could check out our launch and show some love:

https://www.producthunt.com/p/pr...

Your support, feedback, comment, or share can help us maintain our position and keep the momentum going.

Hi Indie hackers!

Been a very long time since we haven't spoke about, 1 year already I suppose?

While building Jesse, I talked with a ton of founders, builders and asked about their experience with the best and worst lead gen tools they have used.
I am curious to know the opinion of this group on this matter.
Some common ones that many use today are:
> Apollo
> Clay
> Lusha
> Zoominfo
Feel free to name and mention any other tools that you have come across.

AVE could already analyze footage locally and distinguish what was happening in different clips but the asset library could still be full of names like IMG_4821, MERT8581, or butterfly_2.

That felt wrong. If an AI video editor understands that one clip contains a red butterfly and another contains a blue butterfly, those distinctions should be reflected in how the footage is organized.

So I shipped Smart Rename in AVE.

It uses the project context and AVE s local analysis to suggest distinct, descriptive names across a collection. Every suggestion is reviewable and editable before applying. You can exclude clips, undo the operation, and keep the names through save, reopen, and relink.

Paybond Kit 0.12.13 0.12.14 are out TypeScript (@paybond/kit) and Python (paybond-kit).
Also shipped a new landing page on paybond.ai.

0.12.13 CLI that helps you recover

Auth and validation failures map to stable errors with clear next steps (paybond login, paybond doctor). Optional --debug when you need a stack trace.

0.12.14 hosted MCP over Streamable HTTP

Hey Product Hunt Tony and Steven here from ClawTeams.

Three days ago we launched ClawTeams, the first goal-driven, proactive AI team for e-commerce. We hoped for a warm welcome. What caught us off guard was how much the comments themselves would end up teaching us. We landed at #1 Day Rank, crossed 800+ points, and picked up over a thousand followers but honestly, the questions in the thread have shaped our roadmap more than any planning doc we wrote before launch. So I wanted to write down what we heard, what it changed, and where we're headed.

Hi everyone,

Glideo 0.3.0 is out. It auto-updates in place, or grab it fresh at glideo.app. This release turns Glideo from a single-take recorder into a more robust multi-clip editor, while keeping the auto-zoom that does the editing for you.

 Hey folks
 Instead of another feature list, let's make this useful. If you have a Chrome extension idea sitting in your notes app, one you keep meaning to build but the setup always kills the momentum, drop it in the comments. I'll reply with how PlugThis would actually approach it: the structure it would generate, the permissions it would need, and where the tricky parts are.
No idea is too small or too weird. Some things people have already been describing:

 - Alert me when I've spent too much time on a certain website

 - Score a Twitter or LinkedIn profile against my own criteria, inline

 - One-click clean up and export the current page

We're live on Product Hunt today with the Second Brain desktop app. I wanted to share what this journey has actually looked like.

Every feature in this product came from a comment. Not a roadmap. Not a strategy session. A comment from someone in this community who told us what was missing.

One of the earliest problems we ran into building v2 was this:

You decide something in January. You change your mind in March. You ask about it in July.

Skills Marketplace by Databox launched yesterday (June 30th), and the community pushed it to #2 Product of the Day. Thank you.

Most of my day is spent in Codex, Claude Code & Cursor building DunSocial.

Opening another dashboard just to save memories, write product updates, draft posts & schedule them always broke my flow.

Now I just connect the DunSocial MCP.

When I finish a feature, I ask my AI to create a PR. It saves the latest updates to memory, drafts a post, finds the next free publishing slot, schedules it & creates the PR. I never leave my editor.

I want to talk about something that doesn t get discussed enough in product development: what happens when you actually listen to your community instead of your roadmap.

When we launched v1 of Second Brain here on Product Hunt, we had a plan. Then the comments came in. People told us newer isn t always more correct. That conflicting memories needed an authority model. That compression should preserve the through-line.