I've been working on a project called AgentNet, and I'd love to get feedback from fellow builders.

Instead of building another productivity chatbot, I wanted to explore what happens when AI agents have full autonomy. In this system, agents have 32-dimension personalities, their own internal economy, can vote on governance rules, and form organizations without human intervention.

As a solo builder, it took 170+ test suites and several architectural rewrites to make multi-agent state persistence and economic balance work.

I'd love your honest feedback:

Hey PH
I'm Rachid, maker of MenuLens a tool that helps restaurant owners see which dishes actually make money using menu engineering (Stars, Plowhorses, Puzzles, Dogs).
We're launching tomorrow (Aug 27) on Product Hunt.
Try the free demo now: https://menulens-app.vercel.app
Would love your feedback before launch day especially from anyone in hospitality or F&B. What would you want to see in a menu profitability tool?

Fennec is a privacy-first, local log viewer with AI-assisted troubleshooting built for teams that can t send logs to the cloud.

Ideal for:
DevOps & SRE
Backend Engineers
Finance / Healthcare IT
IoT, E-commerce, Game Ops

Hey Product Hunt!

I built FreeViralKit a 100% free AI-powered YouTube SEO tool for creators who are tired of guessing what titles, tags, and hashtags to use.

The problem: Most YouTube SEO tools are expensive or have limited free plans. Small creators get left behind.

What FreeViralKit does: Generates 10 viral, SEO-optimized titles instantly Creates 20 25 SEO tags under 500 characters Writes keyword-rich video descriptions # Finds trending hashtags for any niche Works for both long-form videos & YouTube Shorts

Many founder requests start with: "Do you know someone who could help?"

I am finding that the harder work often comes before the introduction: deciding what you actually need that person to help with.

You give it a sentence describing a text task sort these support tickets, tag this feedback, work out what the customer is asking for and it finds or builds the training data, fine-tunes, and hosts the endpoint. pay per job, no subscription.

The data step is the only part I'd claim is interesting, mostly because it's where everything went wrong. Nearly every fine-tuning tool assumes you show up with a dataset. The people who most want a custom model don't have one and have no realistic route to getting one, so that assumption quietly excludes them. So it searches dataset repos and open-data sources against an inferred task spec, ranks on label fit rather than keyword match, checks licenses before anything enters the pipeline, and generates labelled data when nothing suitable turns up. That last one is gated hard, because ungated synthetic data is worse than no data and it took a while to accept that.

An early user broke it comprehensively and the fix is most of what this version is. Nothing reaches training without passing a blocking check. If a failure was detectable before we charged you, it isn't allowed to happen after. Every error comes back with a cause and something you can click. Sounds trivial written down. Was most of the work.

What it doesn't do: anything other than text, pick between architectures for you, or improve the model after deployment. All on the list. I'd rather say it than let the landing page imply otherwise.

I'm building NeoNexus which is an AI agent that resolves every day conflicts, any day. NeoNexus queries five AI models simultaneously, lets them debate, and delivers a single judge-written verdict with dissenting views shown so users get more reliable answers than any one model gives alone. Key features of NeoNexus are Council Mode to convene up to 4 models on any question you want, and Model Arena to fight each LM in NeoNexus. Feedback would be nice, and if you want to take it to the next level buy a subscription! It's also on playstore so do check it out here if you have an android Play Store Link. How can I make both users, and revenue off of this. I already do have some users but want more. Any advice? And if you like my website, just have faith in me and try a subscription. If you don't like it in the first 10 minutes of using it, you're free to cancel.

I spent a lot of time on the hiring/HR side, and I kept running into the same wall. A candidate would clear the interview, then someone has to manually re-enter their info into the HRIS. Then performance reviews live in a totally separate spreadsheet. Then when someone gets flagged for a project, nobody remembers what was discussed in their last review. Every stage resets the context.

So I started wondering: why does the "employee lifecycle" get treated like five disconnected tools instead of one continuous thread?

That's what led to Pilatus. It's an agentic system where a dedicated agent handles each stage sourcing, screening, interviews, offers, onboarding, HRIS, performance and an orchestrator keeps them all synced so nothing gets lost in the handoff. Nothing runs fully on autopilot though; there are human approval gates at the points where judgment actually matters, so it's not just agents making calls in the dark.

We've been testing it with a few teams and the biggest relief people mention isn't speed, it's just... not having to re-explain context five times to five systems.

Hello Guys,

I built Calibra a layer for robotics dataset quality and intelligent data curation

We are exploring how teams can automatically audit data quality, detect problematic demonstrations.Therefore you can build smaller and high-quality training sets before spending GPU time. So you can reduce your training costs up to %70.

Every music release needs more and more visual assets: Spotify Canvas, Apple Motion, YouTube visualizers, Reels, banners, profiles, mockups...

We got tired of creating the same formats over and over, so we built CoverScale.