PH热榜 | 2026-07-07

👋 Hey Product Hunt!

I'm Lokesh. Over the last 15 years, I've hired hundreds of engineers and interviewed even more candidates.

One thing kept bothering me.

Resumes tell me what someone claims they did. They rarely tell me how they actually worked.

Today that's an even bigger problem because AI can generate polished resumes, cover letters, and LinkedIn profiles in minutes. Every candidate looks impressive on paper, making it harder than ever for recruiters to separate signal from noise.

At the same time, professionals have another problem.

The people who have actually worked with them know they're great, but that proof stays buried inside company 360° feedback systems, Slack messages, or forgotten performance reviews. Every new job search starts from scratch.

That's why I built Badge.

Badge uses AI agents to collect anonymous peer reviews from people you've actually worked with and turns them into a portable proof of work that recruiters can trust.

What makes Badge different?

Unlike LinkedIn recommendations...

• Reviews are anonymous, so people are far more honest.

• AI agents guide reviewers with targeted questions, making reviews detailed instead of generic.

• Reviews are verified through work email verification (or employment verification), reducing fake or exchanged recommendations.

• Every review contributes to a Trust Score that summarizes how you collaborate, communicate, and deliver, not just what skills you list on a resume.

For professionals

Instead of asking colleagues to write public recommendations, simply invite them through Badge.

In just a few minutes you can:

✅ Build a verified Trust Score
✅ Showcase real proof of work
✅ Share your Badge on your resume, LinkedIn, portfolio or email signature
✅ Grow your professional reputation over time as more verified reviews come in

For recruiters & hiring managers

Instead of relying only on resumes or spending hours doing reference checks...

You can quickly understand how someone performs in real teams.

Badge helps surface:

• Communication style
• Collaboration
• Ownership
• Reliability
• Strengths
• Areas for improvement

Think of it as a much richer hiring signal before investing hours in interviews or reference calls.

Who is Badge for?

👩‍💻 Software engineers
🎨 Designers
📈 Product managers
📢 Marketers
🎯 Sales professionals
🧑‍💼 Consultants & freelancers
🏢 Anyone whose reputation is built through the people they've worked with

If you'd like to try it: Paste a LinkedIn URL or email to check a candidate's reviews or build your professional reputation by giving your first review. Here’s the link >> https://getbadge.app/ or if you want to look for proof of work for your next hire here is the link >> https://getbadge.in/recruiter/

I'll be here throughout launch day to answer every question, collect feedback, and discuss where we're taking Badge next.

Thank you for checking us out! 🙌

@lokesh_motwani1 I'm curious about protability , if someone changes industires or has an uncoventional career path , does thier Badge score adapt to different hiring contexts, or is it intended to be a universal trust signal?

Connected my LinkedIn and got three reviews back within a day, all from people I hadn't talked to in years. The trust score concept feels actually useful instead of gimmicky.

How does Badge verify that peer reviews are authentic and representative while preventing biased, coordinated, or fraudulent feedback from influencing a candidate's trust score?

background checks / references are notoriously unreliable. Stale, outdated, fake, shallow etc. So its definitly a pain to try and solve. I think maybe connecting it more to linkedin could be useful. Like a review from a strong linkedin profile goes a lot longer than 10 goood reviews from noob profiles etc.

how does badge handle situations where someone’s past coworkers just dont respond, or worse, quietly say something not so great - does that drag the score down unfairly?

Curious how you handle peers who never respond to the agent's request, does that drag down the trust score or just stay neutral?

the trust score idea is smart but i wonder how hiring managers will actually use it, that side of the adoption puzzle is tricky. congrats on the launch

love it! it'd be interesting to see if there's a use-case for gathering testimonials for your business as well! :)

Does the score ever drop, or is it purely cumulative and always trending upwards as reviews pile in?

Just switched jobs recently, would love this more if it showed score history over time instead of only a current number.

Verification through work email is the strength here, but it's also the weakness for freelance and contractors who don't have that kind of org access to lean on.

Work- quality question for an SWE won't transfer to a marketer's role. So if someone switches industries, does their score reset or just carry over oddly?

This is between a reference check and a 360 review, expect it's portable across companies instead of stuck inside one org's HR system. That portability is the real unlock.

Finally saw a tool that tackles reference checks head on instead of just another resume AI, and the anonymous peer review angle actually feels like it could work.

The agent-picks-the-panel design looks great, it removes the hand-curation flaw that previously made my LinkedIn recommendations useless as signal. One thing - rating systems drift toward the ceiling (Uber's 4.9 problem) and people with rough peer relationships might not join right? So scores may cluster high across the board. Is the Trust Score calibrated to force a spread?

How do you handle situations where someone had a difficult manager but was highly respected by peers? Asking for a friend :D

Can candidates choose which reviewers to invite, or does Badge recommend people automatically?!

Interesting. Congrats on the launch!

What safeguards are in place to make sure anonymous reviews stay fair and aren't abused?

love how it pulls real coworker feedback instead of relying on polished LinkedIn recs, feels like the trust score actually means something because it's built from people who've worked with you directly.

What makes a Badge Trust Score more valuable than traditional LinkedIn recommendations?

Super interesting concept. Is there an option to provide feedback publicly?

The timing for this is right. I have watched the resume lose its signal value this year (200+ applications per role, most AI-polished, all reading identical), and the honest answer to what replaces it has always been some form of verifiable work. Peer reviews collected by agents is an interesting take on that. One question: how do you keep the peer reviews themselves from becoming the next gameable artifact, once people know an agent is collecting them?

Really like this, turning verified peer feedback into a portable trust score flips the whole resume game, since anonymous reviews from real coworkers are much harder to fake than a polished LinkedIn profile.

@raaghav_naraayan_m_v pairing anonymous submission with work-email verification is smart to fix the honesty problem LinkedIn recommendations have. but how does that work for freelancers and consultants who move between short-term clients rather than a single employer? curious whether you can build a meaningful Trust Score without one fixed email

The trust score concept is solid, and seeing real feedback pulled from past coworkers feels way more honest than polished LinkedIn recommendations. Curious how it handles people who don't respond promptly.

Really interesting take the shift from generic praise to verified peer reviews makes a lot of sense.

Curious: how do you handle the cold-start problem for someone brand new with zero prior colleagues to vouch for them? That's usually the hardest part in any trust/reputation system.

Hey Product Hunt Fam!

I'm Kaavya, co-founder of Scribble Network. I've been a marketer and community builder for as long as I can remember, and have been deep in the GEO rabbit hole for a while now.

Two shifts got us here:

• Creator and UGC budgets are eating into platform ad spend.

• AI search loves to cite UGC when it builds an answer.

Which raised a question we couldn't drop: 

Could money spent on creators actually move a brand's AI search visibility?

We tested it.

We were best suited to do this, because we started Scribble two years ago as a UGC platform with 50,000+ creators and over 100 campaigns run.

So we plugged that platform and creators into a homemade AI visibility intelligence tool and ran it on one client across 25 queries.

The tool generated content topics and gaps that the creators then made authentic content about on surfaces LLMs could read.

They were at around 2% aggregate visibility across the top 5 models. The market leader on the same queries was at 58%. Three months of steady UGC on the surfaces LLMs actually read, and that client now sits at roughly 25%.

That's when it clicked.

Every AI visibility tool out there could tell us we were invisible and where the answers came from. None could get us cited.

That's a distribution problem, but almost every AI Search product was treating it like a visibility problem

• So we built the thing nobody else has: the first AI Search tool married to a creator platform.

• We don't just measure your AI visibility, we fix it. 

• Our product attributes each brand mention back to the exact creator post that earned it.

• And the platform now pays creators when they get cited. Creators love this.

This is our edge. We came at this from distribution and not measurement.

We never asked 'how do we track this?', we asked 'how do we actually get brands cited?'

The answer to this already lives in Scribble; try it on your own brand and tell me what you find.

Honest feedback wanted.

My co-founders and I are here all day answering everything.

Ran the audit on our brand and the gap report actually showed exactly which competitors were getting cited instead of us, which was eye opening. Wish the creator amplification piece was a bit more transparent about pricing but the insight was solid.

Love the shift from just measuring AI visibility to actually helping brands become the cited answer.👏🏻

The creator network is an interesting approach too. Curious, how do you measure whether a piece of content is genuinely influencing AI citations over time rather than just improving traditional SEO metrics?

The full loop approach actually clicked for me, I ran the audit on a small client site and the content suggestions lined up with the gaps it flagged instead of feeling generic. Really curious how the creator amplifier piece works in practice.

The live-query approach is the right call, but the thing that bit us building retrieval-backed answers is run-to-run variance. Ask the same model the same prompt twice and the cited sources shift even with temperature pinned, because the retrieval layer re-ranks on freshness and whatever got indexed that hour. So 'this post held as a source on this date' has a real noise floor under it. Do you sample each tracked query a few times to separate a piece genuinely falling out from the model just rolling different sources that run?

This looks super interesting @kaavya_prasad

Let's go

Congrats on the launch team!
Curious, When AI engines start updating their citation logic (which they do constantly), how quickly does Scribble's visibility tracking pick that up and adjust the content strategy?

EO is genuinely where the attention is moving and most brands are still asleep on it. good timing on this

Definitely looks like a useful resource for GEO, which is a market that lacks maturity compared to SEO and is crying out for orgs/products that help and give a boost to small startups. I find the pricing page slightly confusing because the headline pricing doesn't suggest any content generation for the insight plan but, scrolling down, it seems you get 10 pages of creator sourced LLM pages. This feels worth promoting in the headlines? Unless it's not as beneficial as it sounds. One other query. You say Claude is coming soon, but only for the Custom plan. Would it not be possible to select the four you want to be tracked? Perhaps swap Perplexity out for Claude etc. All in all, this is definitely a product I will consider using. Congratulations.

Finally a tool that gets it isn't about a vanity score but actually closing the gap and getting cited.

Ran the audit on my own site and the gap report actually flagged specific prompts where competitors were getting cited instead of me, which is way more useful than a generic score. The creator amplification piece is a clever angle I haven't seen elsewhere.

How does the creator amplification piece actually work in practice, do the creators have to be in specific niches or can they cover anything as long as the content gets cited by the AI?

the way you tied creator payouts to actual AI citations is sharp, turns a vague influencer play into something measurable

How do you actually verify that a creator's content was the specific source an AI engine cited, since attribution in LLM responses is still pretty fuzzy?

how does the creator amplification piece actually work in practice, like are you matching us with specific creators or is it more of a network they opt into? curious how the attribution to AI citations holds up when the content lives on someone else's channel.

I was curious about the attribution part. If several creators are talking about the same brand, how do you know which post actually influenced the AI citation?

Congrats on the launch!🚀 Curious about the 50k creator network though—how do you handle the quality and authenticity aspect so the content doesn't just feel like AI bait?

How does the creator amplification piece actually work in practice, do the creators see the brand brief first or just the content you draft for them?

How does the payout structure actually work for the creators in the network, like is it a flat fee per citation or some kind of performance bonus based on ranking?

Do different AI engines treat creator content differently? Like, could a post get cited on Perplexity but never show up on Gemini or Claude? If that happens, do you brief creators differently per engine, or is one piece expected to work across all five?

Distribution is the insight everyone's been sleeping on. As you scale beyond SaaS/tech, what signals tell you when vertical domain knowledge becomes a blocker vs. just needing the right content angles?

This is a rlly cool idea! However the pricing model seems quite steep. I would be inclined to try this, but the lower tier subscriptions should have some level of creator distribution. More importantly, thoug, the lower tiers seem to just be more expensive versions of other GEO/AEO tools (but with MUCH less tracking / competitor analysis). I would keep a tier around the price range of your current lowest subscription, however the offerings you provide need to be more akin to what dozens of your competitors already offer (and for 50-80% cheaper).

The creator distribution aspect is genuinely different in the best way possible, but you NEED to offer distribution to lower tiers for users to really see the benefit of your platform. Would love to stay in touch!

What makes Scribble a better option for creators than traditional influencer marketplaces?

Congrats on the launch. Instantly showing visibility insight is a nice touch. I wonder how you quickly find out the highest-intent queries? This sounds quite challenging to cover all kinds of brands in a wide range of verticals / markets.

Where do you find the list of AI creators and what are their geographic location. In addition i saw your profile its all web 3 projects, i assume they are web 3 creators?

The creators-only-get-paid-when-AI-cites-them model is the part I would want to understand operationally, since that is the whole trust loop. AI answers rarely link sources deterministically and the citation shifts by model and even by re-asking the same query, so how do you attribute a specific AI citation back to the creator whose content earned it, cleanly enough to trigger a payout? Is that matching automated, or is there a human review step before creators get paid?

Paying creators only on sustained citations is a smart anti-gaming filter, the three-times-a-week threshold especially. From the consumer app side my experience matches your thesis, discovery moved to communities and AI answers faster than any dashboard showed it. Do the bounty economics work for consumer mobile apps too, or does the loop only close for B2B where a cited answer is worth real contract money?

Congratulations on the launch :)  I really like the focus on AI visibility, it's something I've been thinking about a lot lately. I'd love to understand whether this could be a fit for a brand like mine, which sits in a fairly different space from the projects I've seen you work with. I'm curious how the creator content usually comes across: does it read as genuine, first-hand experience? 

congratulations! i enjoyed reading the journey from a UGC platform to an AI search solution. do you plan to support recommendations for multilingual campaigns to improve visibility across different languages?

this feels like a unique combination of creator marketing and AI search optimization. I am curious how often does the visibility score refresh and can teams monitor improvements in near real time as new creator content goes live?

Hi, I'm Div, Founder & CEO of Katalyst.

At Datadog, I watched our best enterprise reps closing multi-million-dollar deals - lose hours every week to work that had nothing to do with selling. Updating Salesforce, prepping for QBRs, rebuilding context that already lived in an email or a call recording. The people best at selling were spending the most time not selling.
I couldn't unsee it.

So we've spent the last year building Katalyst to fix it: working hands-on with sales leaders and reps across teams like Stripe, Atlassian, Justworks, Datadog.

Katalyst is an AI sales agent for teams on Salesforce. It runs 24/7, reads your calls, emails, and calendar, and turns it into action:
- Salesforce records created and fields updated on its own
- The right signals surfaced (across 10k, hiring, contact tracking, partnerships)
- Account plans created, contacts researched
- A brief ready before & after every meeting
- The follow-up drafted and the next step set

✨ We've built Katalyst to work with your existing Salesforce set-up!

What’s new:
- AI Resolution on every account: like claude cowork, take deep actions on every deal with the Katalyst Agent
- Pipeline with Kanban view, group by and custom views
- Katalyst meeting recorder - joins meetings across Meet, Zoom, Teams, Webex
- Slack bot - push updates, signals, actions to take
- Actions overview - Review and take pending actions, plus bulk actions

And it's already working. Mike, Sales Director at Aniai, has his reps taking 20% more meetings a day, walking in fully prepped, and moving deals to contract 50% faster - all while Salesforce stays clean on its own.
@mike_conway - "If you're not using Katalyst, you are missing out on opportunities your competitors will find."

We just opened Katalyst to the public. If your team runs on Salesforce, you can sign up or book a demo. Would love your honest feedback. 👇

P.S - A huge shoutout to the tens of beta customers without whose love and feedback we wouldn't be here. Also to the elite Katalyst engineering and product team!

the post-call automation angle makes sense. curious how it handles messy or overlapping deals where context isn't clean. that's usually where CRM tools break down

The meeting prep feature actually saved me time this morning, I just reviewed the brief right before the call instead of digging through Salesforce myself.

Finally gave it a real shot with a few of our Salesforce accounts and the call summaries came out scarily accurate, no editing needed. The deal pattern alerts caught a renewal slipping I would've missed by a day.

Love the focus on reducing CRM busywork with AI. Does Katalyst learn from each team's sales process over time to improve its recommendations?

Is there a minimum Salesforce edition required? Would it work on Professional or does it need Enterprise?

The auto-drafting of follow-ups right after a call ends is a really thoughtful touch, feels like it actually gets how messy a rep's day is. Nice execution on tying it so tightly into Salesforce too.

Been looking at Attio and Clarify. The pitch of a clean AI-native CRM is tempting honestly. But migrating 4 years of Salesforce data and workflows feels insane. How do you think about that trade-off?

Curious how it handles calls when reps have thick accents or talk over each other, does the summarization still hold up or does it need cleanup? Also wondering how it decides which fields to auto-update without butting into the rep's manual edits.

Finally tried Katalyst this week and the post-call summary landing in Salesforce before I even closed my laptop kind of blew my mind. The deal-pattern nudge on a stalled account was spot on too.

That's clever. When two reps share an account, do their signals actually merge or stay separate in the pipeline view?

Pipeline hygiene is the silent killer of B2B sales teams — an agent that actually works the CRM instead of just logging into it is the right framing. How does it handle reps who don't trust automation touching their deals?

finally tried this last week, the auto summary after calls actually feels accurate not just keyword soup, and the follow-up drafts needed barely any edits

question nobody's asked yet - what's the correction path when the AI writes something wrong to Salesforce? if it misreads a call and updates a field incorrectly or logs the wrong next step, does the rep even notice before it compounds, since the whole pitch is that they stop double checking. an audit trail is different from a rep actually catching an error in real time

finally tried the meeting recorder and it actually pulled out the next step without me nudging it, which never happens with the other tools I've used. the deal patterns view is the kind of thing I'd want on a second monitor all day

Noticed my notes and next steps were already in Salesforce before I even closed the call tab, which is honestly the thing I never have time to do. The follow-up draft was surprisingly on-tone too.

Curious about what changes for a salesperson day to day with Katalyst. how is this working in practise for them? What kinds of actions is Katalyst truly taking for the Rep

How accurate is the AI? I feel like every AI tool I've tried hallucinates or gets things wrong. What's different here?

How hard is this to set up? I know enterprise software can take forever to implement.

The "Do This Now" section is probably my favorite part. How does Katalyst decide what deserves my attention first?

What's your retrieval architecture? Are you using vector search, structured CRM data, or both?

Sales teams already spend too much time updating CRM. If this actually reduces that manual work, it could be very useful.

How long does implementation typically take for a Salesforce team?

How is the hygiene score calculated?

Reading calls, emails, and calendar to keep opportunities current is the right wedge — that hygiene work is exactly what eats selling time. My operator question is about the write path's identity: does Katalyst write to Salesforce as each rep with their own permissions and sharing rules, or through one service account that can quietly sidestep them? And on a multi-threaded deal where two reps' calls touch the same opportunity with conflicting next-steps, how does it decide whose input wins?

This is a smart approach to the "AI does the boring CRM work" problem.

Since Katalyst is auto-creating records and updating fields based on

call/email reasoning, how are you handling cases where the AI

misreads a signal — like marking a deal as "slipping" when it

actually wasn't, or auto-filling a field with the wrong next step?

Curious if there's a review/undo layer before changes go live in

Salesforce.

How does Katalyst prevent inaccurate AI-generated insights from affecting pipeline forecasts or CRM data?

The pain you're describing from Datadog is exactly right — enterprise reps rebuilding context that already lived in their email is a brutal tax on selling time. Curious: when a rep has 20+ open opportunities, does Katalyst prioritize which ones to surface in any given week, or does it process everything equally and rely on the rep to decide what matters?

The platform experience is super clean and inviting. Sadly, I don't have a salesforce account to link so could not complete the free demo. All the best with this nonetheless. :)

Hello world! We built Dupely because we kept getting frustrated with duplicate listings on Amazon. There are a million tools out there for resellers to get on the first page but none to help consumers cut through the dupes.

Today, we ship a fully functioning app + extension to help consumers get the best deal from reputable vendors on iOS, Android and Google Play!

As your using Dupely, feel free to tag me with some of your top dupes, feature ideas and if there's anything we can do to make the user experience that much better.

Thank you very much for your support and we look forward to building this into the trust layer for eCommerce!

Congrats!! Can users customize what matters most, like price history, seller trust, or return rates?

Genuine question, not a knock: once sellers know a trust score exists, isn't the badge itself just the new thing to optimize around? Feels like the arms race just moves up a level instead of ending.

Would love a weekly digest of flagged fake discounts I almost fell for, turns the trust score into a habit not just a lookup.

The part nobody's asked about is the matching itself. Deciding two listings are the 'same product' is the whole game, and Amazon makes it ugly: a serum sold as 30ml vs 1oz, single unit vs a 2-pack, a renamed variation ASIN, a bundle with a free sample tacked on. When I worked on product matching, title plus image similarity got us to maybe 80%, and the last 20% (unit counts, pack sizes, refurb vs new) was where every wrong dupe call came from. A wrong dupe stings more than a missed one, since you're steering someone to buy the cheaper thing that isn't actually the same. How are you drawing that line today?

The Trust This Price flag using 90 days of history is the part I'd actually rely on, since fake "was $X, now $Y" drops are exactly what trips me up. When I'm on an Amazon listing with the Chrome extension, does DupeScore surface the cheaper identical product inline on the page, or do I have to open the app and search for it myself? And does the price history cover third-party marketplace sellers, or only the main buy-box?

Logan's question in this thread, whether the badge becomes the new thing to optimize around, is the one I'd push on too. I build in a category where dark patterns are the default and the honest answer is you never eliminate gaming, you price it up, so Ryan's framing rang true. What happens when a seller earns the badge legitimately and then quietly degrades, is there decay built into DupeScore?

I'd love browser pop-ups explaining exactly why a seller earned or lost their credibility badge. What do you think?

This is a cool idea. How frequently is the seller trust data refreshed as new reviews and returns come in??

Congratulations on shipping. Does Dupely also compare prices across different retailers or only within marketplaces?

Congrats on the launch. I'd love notifications when a trusted seller starts offering a product I've saved. Is there a feature that exists for this?

The 90-day price history alone would save me from buying during fake "limited-time" sales. xD

Can users report products they think Dupely missed to improve future matching?

How DupeScore handles products with tiny cosmetic differences but identical manufacturers underneath?

I usually compare several websites before making a purchase. can Dupely alert users when the same product appears at a better verified price after they save it?

How does DupeScore actually identify identical products versus just similar ones, and does it factor in shipping costs when showing cheaper alternatives?

Love that you pulled real price history into Trust This Price instead of just slapping a percent off label on things. The 90 day window is such a smart, grounded way to expose fake sales.

finally a tool that feels like its on my side when shopping online, ran it on a few listings and caught at least one obvious price hike hiding as a sale

Hey Product Hunt. Ryan here, co-founder of Dupely.

The idea for Dupely came about because I've seen from the other side what it's like selling on these major marketplaces we're trying to fix. I've been selling on these marketplaces for over fifteen years. Over the years I've noticed how much more difficult it is becoming as an average consumer. Important information buried in the metadata of the listings. If an item is returned often, if the seller is recently launched, seller reviews buried through clicks, true price tracking.. the list goes on and on. The advancement of AI meaningfully made that trust erosion move faster.

I saw white-label sellers list the exact same factory product at a 40% markup. I've been targeted by ads telling me I can get rich quick making passive income buying from Alibaba. I was invited to groups that gave artificial discounts to move your listing up in the rankings by gaming sales velocity. I saw reviews bought, traded, and seeded. All of this from the lens of the seller, not the buyer.

Then I spent a decade in edtech, and I watched the same rot show up somewhere you would never expect it. Candidates auto-rejected by AI hiring managers before a human ever saw them. People reverse-engineering ATS keywords just to survive the first filter. Good people gamed out by a system that was supposed to be neutral.

We are in a trust recession. The tools built to help us, search, reviews, rankings, hiring filters, have quietly been turned against us. And most of us just accepted it as the cost of doing anything online.

Dupely is my answer on the shopping side.

It finds the identical and near-identical products for less, verifies the seller, scores the match from 0 to 100, and flags fake sales using 90 days of real price history. It fires the second you land on a product page. No hunting, no extra tabs, no wondering if the deal is real. If you've ever found the product you were looking for then ended up not buying because you were overwhelmed with an abundance of choice and analysis paralysis.. Dupely should solve that for you.

What's next is where I'd love your brain. We're working on review authenticity scoring so you can see which reviews are actually real. Wider coverage across the marketplaces where this is the biggest issue: TikTok Shop, Walmart, Wayfair, Home Depot, and more. And a clearer picture of who the real seller behind a listing actually is.

And if you're a seller, we want to hear from you too. You're sick of your products getting knocked off.. we want to help verify you. Let's talk.

What other features could we solve for to make your shopping experience better? If we happen to save you money in the meantime, that is a huge win.

If you've ever felt that you don't know whether you just bought the right product for a fair price, that's the feeling we're building against.

I'll be here all day. Tell me where we're right, where we're wrong, and what you'd want us to catch next.

finally tried dupely on the chrome extension and the trust this price feature caught a “sale” that was actually the same price it's been for three months, exactly the kind of thing i've been burned on before. nice to see someone focus on this.

finally a chrome extension that tells me when a "sale" isn't actually a sale, the price history view alone is worth installing

Nice idea. How do you handle cases where 2 listings look identical, but the product quality turns out to be different? Congrats on the launch!

How does DupeScore actually decide two products are real duplicates versus just similar items, especially when brands deliberately rename or repackage the same thing?

The price history view is surprisingly addictive, it makes you realize how often "sale" tags are basically decoration. Seller badges feel like a step in the right direction too, though I'd want to see how strict they are on the sketchy resellers.

Tried it on a few products I was already eyeing and the 90-day price history genuinely changed my mind on one "sale" that was actually the normal price. Nice to finally see a tool that doesn't just hand me a coupon code.

How does DupeScore actually decide when two products are "identical" enough to recommend, especially for things where listings vary a lot in bundle size or packaging?

How does Dupely figure out which white label sellers are actually the same product under different brand names, and does it work for marketplaces like AliExpress or mostly just Amazon and the usual big stores?

How does Dupely actually verify its seller badges, especially for those random white label shops that pop up overnight and disappear just as fast?

Downloaded the iOS app and was surprised how often "fake savings" tags popped up on stuff I'd been eyeing. DupeScore saved me like $12 on a serum that's literally identical to the pricier one.

Congrats on the launch! 👏🏻 @jacob_galajda Tackling the white-label markup problem is a massive pain point. Curious about the DupeScore—how does it actually differentiate between an algorithmic price drop and a fake discount, especially during major sales events?

Hi Product Hunt 👋

I'm Lava, Founder of Decode by Entropik. We've been building AI that reads human behavior for 9 years and today, we're launching Mira, our AI Moderator.

Here's the problem every researcher knows but nobody talks about: people say one thing and feel another. It's called the Say-Do Gap. Self-reported data is filtered, rationalized, socially edited. Most research tools just accept this. We didn't.

Mira runs the entire research workflow, recruiting, moderating, analyzing, reporting, but uniquely captures what participants say AND feel in real time via facial coding, voice emotion AI, and eye tracking.

When someone says "I love it" but looks confused, Mira notices and probes deeper. Automatically.

Built on 17 patents. 70+ languages. Trusted by Unilever, Nestlé, and 150+ brands.

First study free with code PH20 → entropik.io/platform/ai-moderator

One question for the community: what's the most unreliable part of your current qual research process, and what would it take for you to actually trust AI to run it?

Drop it in the comments. I'll be here all day.

Senior Director of Consumer Insights at Decode here.

Speaking as a practitioner, the feature I find most useful day-to-day is AI highlight reels. Instead of asking a stakeholder to watch hours of recordings, Mira clips the moments that matter: where a participant's hesitation revealed unspoken doubt, where genuine excitement came through before they could temper it, where what they said and what their face showed did not match.

Those moments are what change decisions in a review meeting. And they surface in minutes, not days of manual review.

The AI Copilot is the other one worth knowing about: you can ask "which participants mentioned trust concerns?" or "show me everyone who reacted negatively to the pricing slide" across an entire study instantly.

For researchers here, what type of qualitative research do you run most? Happy to walk through how this fits your workflow.

Head of Data Science at Decode here.

The question we kept asking ourselves: how do you build an AI that understands the difference between what someone says and what they mean?

The answer is multimodal signal fusion — combining facial action units, vocal pitch, speech rate, micro-expressions, and gaze patterns into a single emotional signal per moment of the interview. Not post-hoc analysis. In real time, during the conversation.

The Voice Emotion AI specifically analyzes confidence, hesitation, excitement, and frustration from the audio layer independently of the transcript. Tone often carries a completely different story than words. That layer is invisible in every transcript-only tool.

17 patents cover the core methodologies. Happy to discuss the technical depth of any of these.

Hi all,

Marketing lead at Decode here.

I have spent the last few months working closely with this product to build the launch. The thing that struck me most: most AI interview tools stop at the transcript. Mira treats that as the starting point.

A few capabilities people miss: AI follow-up probing that automatically asks "why?" and "tell me more" mid-interview based on what the participant actually said. AI highlight reels that automatically pull emotional moments, so you do not have to share 40-minute recordings with stakeholders. Cross-study intelligence that finds recurring themes across multiple research projects over time.

The multimodal layer — emotion AI on top of the interview — makes findings more defensible. You are not just quoting a participant. You are showing what they felt when they said it.

Happy to answer questions about the product or how we built the launch.

Product Manger at Decode here

One thing that always fascinated me about qualitative research: researchers don't just analyze what people say. They spend hours replaying interviews to understand how they said it.

A micro-expression of disgust at the pricing slide, a long pause before "I'd probably use it," hands that stopped moving the moment they said they were "comfortable."

That's the invisible layer of qual research. The part that turns a quote into an actual insight.

Mira is built around that gap, it conducts interviews in 75+ languages, probes emotionally in real time, and simultaneously captures vocal hesitation, facial emotion, and where attention went. So researchers stop losing those moments to memory and manual replay.

The goal isn't to replace researchers.
It's to give them superpowers, so they can spend more time discovering why people behave the way they do, instead of manually reviewing hours of recordings.

Happy to answer any questions about how any of this works.

Head of Sales at Decode here.

The conversation I have most with research and insights teams: "Our studies take too long, and leadership does not trust the findings."

Both problems have the same root cause. The tools being used only capture what people say, and analysis is manual. A 20-participant qual study typically takes 6-8 weeks from setup to report, arriving too late to influence the decision it was commissioned to support.

Mira compresses that timeline. Study setup with templates takes minutes. Recruitment from a 100M+ global panel is built in. Transcripts, themes, and reports are generated automatically. Emotional signal adds defensibility to findings.

If you run an insights function and want to understand what this looks like for your team's specific workflow, feel free to ask below.

How does the facial coding and eye tracking actually work in a remote setting without making participants uncomfortable or needing specialized hardware?

"Captures what participants say AND feel" is the hard part most tools skip. I work with voice AI daily, and the transcript is the easy 20% — the signal that actually matters (hesitation, tone, the pause before someone says "yeah, it's fine") lives in how they said it, not the words. If Mira genuinely reads that layer at a 100+ panel scale, that's a real moat over "interview + transcript" tools. Question: how do you keep the emotional read from becoming false precision — do you surface confidence levels, or a single sentiment call? Congrats on the launch 🚀

congrats! user interviews are the part of building I always end up skipping because of the effort. this could actually make me do them

Deeper emotional signal is great, but storing facial and voice data at scale raises the privacy stakes quite a bit- how long is that data retained?

the facial coding piece is wild, didn't think i'd see real emotional response analysis baked into a research tool at this price point

Does it self-report confidence?

The facial coding during interviews actually caught a reaction I would have completely missed reviewing the transcript alone. Seeing emotion data layered with what people said felt like a real research upgrade, not just another AI wrapper.

the facial coding piece is wild, watching the emotion data overlay during interviews made patterns jump out that a transcript alone would totally miss. really curious how it handles cross-cultural nuance though.

How does the facial coding and eye tracking actually work in practice for remote participants, especially since webcam quality and lighting vary so much across users?

How does the pricing actually scale for smaller teams or solo researchers who don't need the full enterprise setup, and is there any way to try it before committing?

the facial coding detail is wild, never seen a research tool that actually picks up on what people are feeling while answering. ran a quick test on our team and the voice emotion layer caught hesitation I would have totally missed in a normal transcript.

How do i verify the accuracy of Facial coding and eye tracking?

Strength here is catching reaction people wouldn't admit to on a survey. Weakness is that everything depend on decent webcam and lighting, which won't be true for every participants in every country.

The facial coding during interviews is genuinely impressive, you can actually see hesitation or excitement that pure transcripts would completely miss. Wish more research tools took emotion seriously like this.

The facial coding combined with voice emotion analysis during a test interview genuinely caught moments I would have missed in a regular transcript. Feels closer to sitting with a respondent than a typical AI research tool.

The facial coding and emotion AI combo is wild — usually I get a wall of quotes to sift through, but here it flagged where respondents actually felt something. Made the insight pull way faster than expected.

What stands out is the integration of facial coding and voice emotion AI directly into the interview flow rather than bolting them on as an afterthought. That feels like a real research instrument, not just a chat wrapper dressed up with sentiment scores.

What happens when it works fine and every brand in a category runs creators against the same queries, does it become an arms race where the UGC cancels out, or is there a ceiling on how much citation share you can actually buy back?

When Mira spots a say/feel mismatch and digs deeper on its own, how do you keep that follow-up from leading the participant? A moderator reacting to visible confusion can easily plant the doubt rather than uncover it. Is the probe neutral by design, or tuned per study?

How does the facial coding and eye tracking actually work in practice with participants who are on their phones or in different lighting conditions? Curious how reliable that data really is across such a massive global panel.

The facial coding and emotion layer actually feels different from typical survey tools — I ran a quick concept test and the sentiment data picked up nuances I usually miss in write-ups.

How does the facial coding and eye tracking piece actually work in practice, do participants need to opt in and run any special setup on their end?

Tried the facial coding on a quick test and the emotion read was scarily accurate to what I was actually feeling during the open ends. The auto-generated themes also saved me a solid hour of tagging.

The undo + audit trail is reassuring, but undo only works if you catch it before the recipient reads it. Once Autopiolt sends somethings wrongs, what's the actual recourse, is it just a follow-up correction email, or is there more built in?

The covenience of full autopilot is obvious, but there's a real trade-off between speed and the one bad send that damages a relationship. Where's the line you'd draw between Capilot and letting it run alone?

Looks really promising. I like the focus on drafting emails in your own writing style instead of generic AI responses, and having manual, copilot, and autopilot modes gives users a lot of flexibility. Congrats on your launch!!! How long does it typically take for AI Emaily to learn and reliably match a user's writing style?

Tried a couple of "writes like you" tools before and they were fine for short replies but fell apart on anything with real negotiation or nuance. Where does this one land on that?

Superhuman and Shortwave both pitch AI triage too. What's the one things here that actually makes someone switch over instead of just adding another inbox app?

How manageable does Personal Context get once you've got dozens of client profiles goings? Feels like it could turn into its own maintenance job.

Curious how this handles shared shared inboxes, like a support@ alias with multiple people replying.

Is there a team/ shared inbox plan or just solo use for now?

Autopilot sending my mail? Bold.

The undo and audit trail on every action shows real thought about trust, not just a flashy AI demo.

How does the voice matching actually work in practice? Does it learn from my sent mail, and how long does it take before the drafts stop sounding generic and start reading like me?

How does it actually learn my voice well enough to draft replies I trust, especially in the first few weeks before it has much to go on?

Set it to Copilot mode on my work inbox and it genuinely flagged only the messages that needed me that day, which is more than most assistants manage. The undo and audit trail gave me enough confidence to let it draft and send a few replies while I stepped away.

inbox management is one of those things that sounds trivial until you're spending 2 hours a day on it. the triage piece alone is worth a lot

How does the "drafts replies in your own voice" actually work at the start — do you need to feed it old emails to learn your tone, or does it pick up your style from somewhere else?

How does it actually learn my voice well enough to draft replies I'd send myself without me tweaking every message?

How does it actually learn my voice well enough to draft replies, and is there a sample size or warm-up period where the output might feel off before it gets it right?

Curious how "drafts in your own voice" actually works at the start — do I need to feed it a bunch of past emails, or does it figure that out from a few samples?

How does it actually learn my voice well enough that the drafts feel like me and not a polished assistant version of me, especially in the first few weeks?

Autopilot email sending is the boldest part of this. Once the agent can send from someone's real address, one wrong call lands in a customer's inbox with their name on it. Is there a scope on what autopilot may do on its own, like never emailing new contacts or anything with money in it, or is the undo window doing all the safety work?

The "writes like you" part is the hard part — most AI email tools sound like a press release. Curious how many emails it needs before the voice actually matches?

Drafting replies that sound like me instead of a robot is the bit I cared about, and it actually nailed my tone on the first try.

Finally an inbox that drafts in my voice instead of generic AI. Tried it on a backlog of 40 emails and it handled the obvious ones without me touching them.

Drafted a few replies and the tone actually sounded like me, which is the part that usually trips these tools up. Love that everything is logged so I can backtrack.

the three-mode toggle feels genuinely thought through, not just a marketing checkbox. nice to see undo and a full audit trail treated as core, not afterthoughts.

The Copilot mode feels like it actually learns how I write after a few replies, which is the part I was most skeptical about. Calm inbox now, finally.

How does it actually learn my voice well enough that I don't end up rewriting every draft, and is there a real free tier or just a trial before the first bill?

the copilot mode actually feels like it gets my tone right, not the usual stiff AI reply. love that i can still undo anything before it goes out

👋 Hey Product Hunt!

I'm Robin, creator of Ellis — a personal AI notetaker for in-person meetings.

What is Ellis?

Ellis is a simple consumer-first AI notetaker (iPhone and Apple Watch) for in-person meetings. It works anywhere being in the same room matters: coffee meetups, on-site sales meetings, therapy, doctor visits, interviews, or even your teacher-parent conference.

When the meeting ends, Ellis matches your voice against your saved voice profile, gives you a full transcript with an easy way to assign speakers, and writes notes in the format you pick.

Why Ellis? 🤔

Unlike other notetakers that are built for your org, Ellis is built for you as an individual. You can record lots of different in-person conversations, ask questions across them, and find common traits and trends spanning both your professional and personal contexts. Yes — I might want to know how my sales meeting went AND how I navigated a difficult conversation with another caregiver. Two different use cases, but one repository, built for me.

Key features 🤩

• Getting "who said what" right. Telling speakers apart in a real room is harder than online — every voice hits the same microphone. Ellis solves this with voice enrollment and diarization, plus a fast way to tag yourself and others.

• Ask anything about a conversation. Pull up what was said, decisions made, or anything you want to revisit from a meeting — just ask.

• Find it by place. Forgot a name or a detail? Ask by location — "what did we agree on during our walk in Fort Greene?"

• Private by default. Recordings are automatically deleted once transcription is complete.

Happy to answer questions — and I appreciate the support. 🙏

How do you balance giving users AI coaching feedback without making conversations feel overly analyzed?

Honestly it's rare by volume but it clusters right where it hurts, the fast back-and-forth when a decision actually gets made, so it feels worse than the raw percentage suggests. I wouldn't chase overlapped-speech separation, it's expensive and still flaky. I'd just surface it: flag the low-confidence diarization spans so I know which two or three lines to re-listen to, instead of trusting a transcript that looks clean. The silent winner-pick is the part that bit me, not the noise.

Everyone built for Zoom and forgot rooms exist. How does it handle four people around one table with a single phone mic?

In-person is the right wedge, every AI notetaker assumes a Zoom link exists. How do you handle speaker attribution in a noisy room without everyone wearing a mic? That's the failure mode that killed my voice-memo system for coffee meetings.

The in-person angle is what sets this apart, but recording therapy or doctor visits is also where the privacy question gets sharp. Does the audio and the diarization run on-device, or does the recording get uploaded to a server to transcribe and match against my saved voice profile? And since everyone else in the room never installed the app, is anything about their voice retained, or is it all local to my phone?

For me it comes up in fast 3+ person brainstorms and standups, almost never in 1:1s or sales calls where people take turns. I wouldn't chase true source separation, that's a research problem you don't want to own. The cheap win is honesty: when AssemblyAI hands back a low-confidence or overlapping stretch, drop a small 'crosstalk here' marker instead of a clean line, so I know to trust my own memory for that bit. A confidently wrong transcript is worse than one that admits a gap.

The noisy-room question has a nastier cousin: true overlap, two people talking at the same instant. Diarization picks one speaker per frame, so the quieter voice doesn't get mislabeled, its words just vanish, and nothing in the transcript tells you a sentence went missing. That's harder to catch than plain noise because the output still looks clean. When people talk over each other in a fast brainstorm, does Ellis mark the overlap or just pick a winner?

The decision to make it work on just an iPhone or Apple Watch is genuinely clever. No extra hardware means it'll actually get used in real meetings instead of sitting in a drawer.

How does the speaker identification actually work in practice, especially when people are talking over each other in a real meeting?

How well does the speaker identification work in a noisy cafe or group setting with people talking over each other?

How well does the speaker identification actually work when people are talking over each other or interrupting? That's usually where these tools fall apart for me.

Curious how it handles crosstalk or people talking over each other in a noisy room. Does the speaker identification still hold up, or does it get messy fast?

finally tried ellis at a coffee chat and the speaker labeling actually nailed it, even with overlapping talk. liked that i could just leave my phone on the table and forget about it

The speaker identification works surprisingly well even in a noisy coffee shop, and being able to ask follow up questions about a meeting I walked out of feels like a real superpower.

the fact that it runs straight off the iPhone and Apple Watch with no extra hardware is such a thoughtful move for in-person meetings.

How well does it handle overlapping speakers or side conversations in a noisy room, and is the transcription actually reliable enough for something like a legal or HR meeting?

In-person notes are a different trust problem than meeting-bot notes. The useful detail is speaker correction and ownership after the recording, because the transcript becomes part memory, part operating record. Getting that boundary right matters.

The fact that it runs straight from your iPhone or Apple Watch without any extra hardware is genuinely clever, removes so much friction from actually capturing in-person conversations.

The speaker identification without cloud processing feels really thoughtful, like the team actually thought through what people want during a meeting instead of just chasing the AI hype.

The “AI Notes for In-Person Meetings” angle is interesting, especially for teams that still do design reviews, sales conversations, or planning sessions around a table. How does Ellis know who said what in a room? Is speaker separation part of the flow, or is it more focused on capturing clean notes and action items after the conversation?

The speaker identification was actually solid in my quick test, which I didn't expect from a phone-only setup. Wish I'd had this for the standup meetings I always forget half of.

How does it handle background noise and overlapping voices in a real conference room setup?

Everyone solved Zoom notes, nobody solved the conference-room whiteboard session. In-person is the harder and more valuable problem. How does it handle multiple speakers without per-person mics?

the speaker identification worked surprisingly well in a noisy cafe setting, way better than i expected from a phone mic.

How well does the speaker identification actually work in a noisy room with multiple people talking over each other? Curious how it handles real-world chaos versus a quiet conference room.

How well does it handle meetings with more than like 6 people talking over each other, especially in a noisy cafe setting?

Recording doctor visits and therapy sessions puts this app in about the most sensitive bucket there is, so deleting the audio right after transcription is a smart move. The transcript still says everything the audio did, though. I saw Supabase in the stack, are transcripts scoped per user at the database level, or does the app layer do the gatekeeping?

the noisy-room questions are all covered, mine's about consent instead - you list therapy and doctor visits as use cases, which means the other person in the room is being fully transcribed and analyzable without necessarily agreeing to that. recording is auto-deleted but the transcript sticks around indefinitely. a lot of US states are two-party consent for recording, does Ellis do anything to prompt "hey, tell the other person" or is that entirely left to the user's judgment

Used it during a quick team sync and the speaker labels were actually right on, which I was not expecting. Being able to just ask "what did we decide" after is the part that sold me.

Hey Product Hunt 👋 Luke here.

I built Social Fetch because I kept running into the same problem.

Every project that needed public social media data started out simple enough. Grab some YouTube data. Add Twitter. Then someone asks for TikTok. Then LinkedIn. Then Reddit.

Before long, I wasn't building the product anymore, I was maintaining scrapers.

Platforms change. Formats are all different. Something is always breaking.

It felt like a solved problem that nobody had actually solved.

So I spent the last couple of years building Social Fetch: one API for live public data across 20+ platforms.

What surprised me wasn't that people wanted social data. It was how many different things they were building with it. AI agents. Creator tools. Lead generation. Brand monitoring. Research. Things I never would have thought of myself.

One thing I care a lot about is keeping it simple: • one API • live data • consistent responses • subscriptions + pay-as-you-go credits that never expire

I'd love to know what you'd build with it.

Or if you've ever built your own scraper... what finally made you regret it? 😅

I'll be hanging around all day if anyone wants to chat or has questions.

— Luke

Hooked it up to a small TikTok enrichment flow this morning and it just worked, no schema headaches or weird rate limits. The pay-as-you-go credits make it easy to test without committing to a monthly plan.

@lukem121 +1 to your forum thread on which platforms break the worst — the silent auth expiry and "structured one week, wall of JS the next" flip is exactly the tax that makes people regret rolling their own.

Before wiring this into an agent I'd want to know: is "live data" fetched synchronously per call, and what's the rough p95 when a platform is being slow? And when something breaks on your side, does the API return a typed error / degraded flag so the caller can tell "no results" from "scraper down"?

Also curious whether there's an MCP server yet, or REST-only for now. 👌

How do you ensure developers can build production applications without worrying about scraper failures or schema changes?

How do you handle rate limiting and terms of service compliance across all these platforms, especially with pay-as-you-go credits and no subscription?

The thing that always bit me building social pipelines wasn't the fetch, it was completeness. 'Get the comments' really means 'get whatever the platform lets you cursor through right now,' and those caps move: Instagram would hand me the first few hundred then just stop, and an expired cursor mid-pagination left me either gaps or duplicate rows to dedupe by id. Does one call return a full set or a capped page, and do you flag truncated versus genuinely complete?

i have been looking for something like this which i am gonna integrate with my flow. also around the scrapping does it do the real time or pushes the time bound scrapped data?

How do you prevent missing posts during platform updates. A clear retry process would build stronger trust with developers.

The pay-as-you-go credits model without a subscription is a refreshing choice for indie builders testing scrappy side projects. Love that you covered transcripts too, that detail saves a ton of glue code for AI workflows.

The part that stood out to me is that it just keeps working when everything shifts underneath it. Anyone who has watched their setup fall apart overnight will feel that one.

How does the pay-as-you-go pricing actually work in practice, like is it per profile fetched, per post returned, or something else? Trying to figure out if it stays cost-effective for monitoring a lot of accounts at once.

The pay-as-you-go credits model is a nice touch since most scraper APIs lock you into subscriptions. Pulling transcripts from YouTube and TikTok in one call saved me from wiring up two separate tools for a side project.

How does the pay-as-you-go pricing actually work in practice, like is it one credit per profile fetch or does it scale with the depth of data you pull back?

The pay-as-you-go credits setup is a nice move for indie builders who just need a handful of API calls here and there without locking into another monthly SaaS bill.

This is one of those unglamorous pieces that decides whether an AI/social product is reliable. The API surface matters, but freshness, rate-limit behavior, and clear failure states are what keep teams from turning every launch into scraper maintenance.

maintaining scrapers across multiple platforms is a nightmare, every API change breaks something. having one unified layer for this makes a lot of sense

Finally tried this after maintaining my own scrapers for months, and getting TikTok transcripts through the API saved me a whole weekend of work. Clean docs too.

How does the pay-as-you-go credit system actually work in practice, like how many requests do typical profile fetches cost and is there a way to estimate usage before running a large batch?

Spent a few minutes pulling TikTok profile data and the response time was way faster than the DIY scraper setup I was running before. Clean docs too.

How fresh is the data on those metrics — are we talking real-time pulls or is there some delay before posts and engagement stats update after going live?

For the scraper API, how consistent is the response shape across TikTok, Instagram, YouTube, X, and LinkedIn? I can imagine developers using Social Fetch in AI workflow automation or marketing dashboards would care a lot about whether posts, profiles, comments, and metrics come back in a predictable schema, or if each platform has its own structure.

Finally tried it out last night and was surprised how clean the transcript endpoint returned for a messy TikTok I tested. Wish more APIs worked that smoothly without a monthly commitment.

How does the pay-as-you-go credit pricing work in practice, like do different platforms or data types cost different amounts per call?

I enjoy the broad platform support because maintaining scrapers takes real efforts. how do you verify that profile metrics stay accurate over time. Regular validation against public data could improve confidence for analytics products.

Pulled a few TikTok profiles and the transcripts came back cleaner than expected, which saved me a ton of cleanup time. Pay-as-you-go pricing is a nice change from the usual monthly minimums.

How do you keep things stable when platforms like TikTok and Instagram constantly tighten their scraper detection, and does the pay-as-you-go pricing mean I only pay for successful requests or every call I make?

How does the credit system actually work per request, like is fetching a profile the same cost as pulling a full post thread or does it scale with the data volume you get back?

finally tried it and the pay-as-you-go credits are a relief, no commitment feels right for tinkering. pulled a few tiktok profiles and was surprised how clean the json came back, saved me a chunk of dev time.

I like the platform coverage but how do you handle sudden API changes across networks. I think fast fallback systems could keep data reliable.

the technical questions are covered well already, curious about the legal side instead - when a platform's ToS changes or they start actively blocking scrapers (LinkedIn especially, given the lawsuits), who eats that risk, you or the customer whose product suddenly loses a data source overnight. is that something you indemnify at all or is it explicitly on us

Hey Product Hunt 👋 I'm Teo, CEO and co-founder of Ogment.

Most companies are told to "adopt AI", but in practice that means asking non-technical people to learn new apps, prompt engineering, and workflow builders. There's one thing we've learned over the past two years offering AI to companies: any bit of friction kills adoption. A few power users push through, true, but everyone else gets left behind.

You've probably seen the wave of Slack-native agents shipping lately, Claude Tag included. They're strong for technical teams who want to own the setup. We went the other way: every employee gets their own AI coworker, on any model you choose (including your own local LLM), with zero setup.

We built @O to attack that friction head-on. The bar we set ourselves: using an AI agent should be as easy as tagging a colleague. So that's exactly how it works, you tag @O in Slack, in plain English, and it gets the work done. No new app, no prompt craft, no workflow builder. When the friction drops to zero, adoption is maximized, the whole team starts using AI super naturally, from day one.

Hello everyone, Flo here, I'm one of the builders at Ogment :)

Really excited to share what we've been hard at work on for the past months. A few more insights into the technical aspects of making an agent in slack run smoothly in production.

The bar we set for this product is extremely high: it needs to be incredibly easy to setup, essentially all integrations should be available to you, it should behave as smoothly as possible in all scenarios that exist in a chat app like Slack, and the hardest part: It should survive and work reliably across a myriad of possible failure modes, errors of all kinds, and deployments - to give people the feeling that "it just works".

First surprise: Slack's API, even their latest one made for agent apps, has an incredible amount of edge cases, error codes and rough edges. Make a tiny mistake, and the entire stream or message your agent is trying to write will simply not be displayed. Getting this right and buttery smooth was an uphill battle of renderers, meticulous testing and bug reports from our alpha testers.

Second surprise: Making agents survive in production still feels like an unsolved problem. Everyone is hand-rolling their own runtime, making agents recoverable when the process crashes, when your LLM provider dies, or when you deploy a new version of your agent and/or its runtime. There's a lot of space to make this simple, or to overengineer it, but one thing is for sure: This is no easy feat (even with the smartest coding agents).

Excited to see what people will delegate to our dear agent O 🚀

Hey Product Hunt 👋 I'm Amaury, cofounder of Ogment, and I'm super proud to share what we've been building.

We kept seeing the same pattern: people are excited about the AI revolution, they want that full "agentic" experience, but then reality hits. The friction to install and maintain agents a la OpenClaw or Hermes is actually pretty high, esp. for less technical people. Not even mentioning the security or the team collaboration aspect...

So we built O from the ground up for Slack-native teams, where a ton of context already lives. Your conversations, your decisions, your files, your workflows: it's all right there. And the best part? There's zero friction to onboard, just tag @O in Slack.

Here's the thing that surprises people most: we don't really have a dashboard. You can do it all from Slack. Connecting your systems (Gmail, Linear, Stripe), creating a custom skill or sharing it with your team, setting up a recurring job... just ask your agent, like you would a colleague.

And it's proactive. Ogment doesn't just wait for instructions. It constantly spots automation opportunities and surfaces them to you, without you even having to ask.

My personal favorite skill with O? It handles my post-product meetings end to end. It pulls the Granola transcript, reads my backlog in Linear, checks strategy docs in Notion, grabs context from emails, and outputs a clear list of feature requests, bugs, market insights, and a backlog update proposal. What used to take me 45 minutes to do properly is now done in 1 minute.

We're incredibly excited to get this into your hands. Can't wait to see what you build with it!

Love how you've embedded the AI directly into Slack instead of asking teams to adopt another dashboard. 👏🏻

The ability to tag it like a teammate feels much more natural for day-to-day work. Curious, what's the most popular workflow users are automating with @O so far?

Exited for this! Congrats team on launching @O!
Tagging on Slack is definitely the best way to interact, that's where team already are. I've been playing quite a lot with ClaudeTag, could you share how you compare to it :)?

Very cool! I use it to automate workflows that eat up my time, like adding call summaries, and tagging relevant people for action, or getting summaries of my day.

Cool stuff @@O - AI coworker in Slack

answering KP's question - I don't think it migrates to a dedicated platform, because the cost being competed against isn't "install a new app," it's "change where my team already collaborates," and that's a much bigger ask than tagging a bot in a place they already live. Slack/Teams/wherever the conversation already happens has a huge structural advantage for that reason alone. the piece I'd push on is the org-level memory layer "curated by an admin" - that's the same promise every internal wiki and knowledge base makes, and those go stale the moment the admin stops actively curating. curious if there's anything automated nudging that curation along, or if it's fully on someone's plate to maintain

Love how it just slots into Slack without a separate dashboard or login. The fact that tagging it feels exactly like pinging a teammate is a small detail but it’s probably the real reason people will actually use it.

Hello Product Hunt!👋

Designer at Ogment here - brand, site, dashboard, launch assets.

The design brief was the same as the product promise: zero friction. If tagging @ O should feel like tagging a colleague, nothing around it gets to feel like software - no new app to open, nothing to learn, nothing in the way.

None of this was designed in a corner - everything we shipped went through the whole team, and half the best cuts came from engineers asking "do we even need this?" Small team, tight loop, that's how it stayed simple 🙌

Don't take my word for it though - just onboard @ O to your Slack and feel the magic ✨

One of my favorite parts? I create O brand assets by tagging our beloved @ O in our #marketing channel, where the whole team can see the process. And when it comes to repetitive revisions... I don't do them anymore. O does.

So excited to finally share this with you all - your thoughts and questions are more than welcome! 💛

Happy to be part of this amazing team and incredibly proud of what we've built.

Such a cool idea and an awesome project! Wishing you the best of luck with the launch and continued success! 🚀

Does it use a specific model, or do we get configure it?

Tool-calling is the sneaky failure with any bring-your-own-model setup. Providers each format function calls differently, so a harness tuned on one model's schema quietly drops args or mis-picks tools when you point it at a cheaper endpoint, and it still reads as working. That's the bug that ate the most of our debugging time building agent infra. How much of the agent quality actually holds when someone swaps their own model in versus your default?

Model-agnostic is a strong claim since reasoning quality varies a lot across models. Can @O route different task types to different models, cheap one for routine CRM updates, stronger one for judgment calls, or is it one model per workspace?

tagged @O to summarize a long thread and it actually pulled the key decisions plus action items in seconds, way cleaner than the bots we tried before.

How does the setup handle different permission levels across departments? Separate access controls with clear visibility could keep sensitive work protected without slowing collaboration.

Really appreciate tools that meet you where you're at in your current workspaces instead of making learn a brand new platform. We recently went through a collaborative multi-team project in Slack and wish we had @O - it would have saved us a ton of time. Looking forward to trying it for our next project!

"Zero friction, maximum adoption" - cheers to this! 👏

we basically live in slack so this makes sense to me. congrats on launching, will try it with the team

the native slack integration is genuinely clever, no clunky web app to switch into and the team-wide memory makes it feel like a real coworker picking up context over time

How does Ogment decide when an AI agent should act autonomously versus asking for human approval?

whats the core differentiator between this and openclaw ?

This is sick. What are your top 3 use cases? Just tried the sales use case and it is amazing. Great job.

The shares-memory-and-skills-across-your-whole-team part plus running automations while you sleep is the combination I would want to pin down before rolling it out. When @O touches one of those 1,000+ tools, is the connection scoped per-user so it acts with my permissions, or is it one workspace-level token everyone shares? And the shared team memory: does that live in your infra, or can I point it at our own store or model like the including-your-own line suggests?

congrats on the launch! looking forward to checking this out

Love the breezy integration, making me to automate time-wasting workflows without me having to think twice about it (for me it's scheduling + note aggregation / tracking)!

How does pricing actually work if multiple teammates are tagging @O throughout the day — is it per user, per workspace, or based on the volume of tasks it handles?

Love that @O just slots into Slack like an actual teammate, no separate app or login to babysit. The shared memory across the team is a really thoughtful touch, most AI tools feel like personal gadgets rather than something the whole crew actually adopts.

Tagged @O this morning to pull last week's client call notes from Notion and it actually just did it. Being able to share automations across the team without another dashboard is a really nice touch.

Tagged @O in our team Slack and it actually pulled the report I needed without me opening another tab. The shared memory across teammates is a nice touch too.

how does the shared memory actually work across teammates, like does it learn from one person's preferences and apply them to everyone or is it more like a shared knowledge base?

the "@O" name being usable as an actual mention in Slack is such a smart move, makes it feel less like a chatbot bolted on and more like an actual teammate people will naturally ping

Hi everyone!

LongCat-2.0 is a 1.6T-parameter MoE model with about 48B active parameters per token, 1M context, and open weights under MIT.

But it was not trained in the usual Nvidia-heavy way. The full training run was built on AI ASIC superpods, over more than 35T tokens, with no rollback or irrecoverable loss spike.

Training a trillion-scale model is already hard. Getting that run stable on alternative hardware is probably the more interesting story here 🤔

the "no rollback, no irrecoverable loss spike" claim on ASIC superpods is the part that would make me trust this more than the raw parameter count. that's usually where alternative-hardware training runs quietly fall apart, not in the benchmarks people see later. one thing I'd want to know before betting an agent pipeline on this: since it integrates with Claude Code and OpenClaw, did the ASIC training path change the numeric format at all versus a typical bf16 GPU run, and if so does that show up as any weirdness once you quantize down for inference, or does it behave like any other MoE checkpoint at that point

How does LongCat decide which parts of a conversation are most important to retain over time?

The stable run on ASIC superpods with no irrecoverable spike is the genuinely impressive part here, more than the parameter count. For agent use the metric I care about is long-horizon adherence rather than single-turn tool accuracy: in our loops the model at tool call 30 has usually forgotten a constraint it agreed to at call 5, and 1M context helps recall without stopping that instruction decay. Did the agentic post-training target staying on a plan across many tool calls, or mostly one-shot tool-call correctness?

How does the 560B MoE setup actually perform on smaller hardware for fine tuning, or is it really only practical to run through Meituan's own infrastructure?

how does the 560B MoE setup hold up on longer context tasks compared to dense models like DeepSeek?

Curious how the 560B MoE setup actually feels in latency for real time stuff like chat compared to something like DeepSeek, and what kind of hardware you'd realistically need to run a distilled version locally?

Runs surprisingly snappy for a 560B MoE — I had low expectations but the reasoning responses came back fast and actually thought through edge cases instead of hand-waving.

the reasoning speed on long context prompts genuinely surprised me, felt closer to a smaller model than a 560B MoE.

Curious how this stacks up against other open-source MoE models on benchmarks like HumanEval or MATH, and is it actually free to deploy commercially or are there usage limits built in?

how does the 560B MoE setup handle latency on longer reasoning chains, and is there a hosted endpoint or is it self-host only?

How does the 560B MoE setup compare on inference cost versus a dense model of similar capability, and is there any self-hosting guidance for teams without massive GPU budgets?

Ran a coding question through LongCat-Flash-Thinking and the response came back fast with clean, working code instead of long-winded explanations. Really impressed by how snappy it feels for such a big model.

Tried it on a logic puzzle and the reasoning came back structured and fast, way smoother than I expected for a 560B model.

LongCat handled a tricky multi-step coding prompt way faster than I expected, and the reasoning chain felt genuinely clear instead of just confident. The MoE setup seems to pay off in practice.

Curious how this stacks up against DeepSeek or Qwen on coding benchmarks, and is the 560B MoE actually deployable on a single node or do you need serious infra to run it locally?

finally a chinese-built model that actually feels fast on long reasoning chains. really impressed how it handled a math-heavy prompt without losing the thread halfway through.

Pulled LongCat-Flash-Thinking for some debugging yesterday and the response time genuinely caught me off guard. Reasoning feels sharp without the usual lag.

Finally got around to testing LongCat-Flash-Thinking on some multi-step coding problems and the speed honestly caught me off guard, especially for a 560B MoE. The reasoning chains feel surprisingly tight too.

Ran it through some tricky math problems and the responses came back noticeably quick for a model this size. Curious to see how the open weights hold up on longer agentic workflows.

How does the 560B MoE setup actually translate to faster inference in practice compared to dense models of similar size, and is it open weights under a permissive license?

How does LongCat-Flash-Thinking handle context length on really long documents, and is it open weights or just open source code?

Tried LongCat-Flash-Thinking on a multi-step coding prompt and was surprised how fast it reasoned through edge cases without losing track. Solid open-source move from Meituan.

Curious how the 560B MoE actually handles rate limits in practice, like does Meituan expose a hosted API or are we expected to self-host the weights for any serious throughput?

How does the 560B MoE actually handle latency in real-world agent loops, and is there a hosted endpoint or do we need to self-host the whole thing to use it?

Congrats to the LongCat team!

As someone based in China, Meituan is a household name to me — I use it almost every week for food delivery and local services. So it’s honestly a bit surprising, in a good way, to see the same company open-source a 1.6T MoE model under MIT. That contrast makes this launch especially interesting.

I’m also exploring AI-assisted PM workflows and building my own agent skills, so the combination of open weights, 1M context, and post-training for coding/agentic workflows really stands out. Curious if the team has any best practices for using LongCat-2.0 in long-horizon agent workflows where the context includes product docs, code, and task history.

Curious to see how you frame the model strengths beyond the short tagline.

awww, cuteeee

The automatic zoom that follows the cursor is genuinely impressive, it makes my demos look so much more polished without any extra effort. Solid free option for Mac users.

Took it for a spin on a quick walkthrough of a new feature and the auto-zoom on my clicks honestly looks better than what I've gotten from paid tools. Clean output, no watermark, and it didn't choke on a 10 minute recording.

The auto zoom that follows my clicks actually works really smoothly, way less tedious than keyframing everything in Final Cut. Already exported a polished demo for a teammate in under five minutes.

Free macOS screen recorder with click-tracking zoom sounds too good to be true, but the output actually looks clean and the no-watermark thing sealed the deal for me.

finally a free mac recorder that doesn't slap a watermark on everything, the click follow zoom actually works surprisingly smooth without me touching a thing.

The click-tracking zoom is genuinely clever, especially how it stays smooth without manual tweaking. Nice to see a macOS-native recorder that respects the platform instead of bolting on a clunky web wrapper.

finally a free screen recorder that does the smart zoom thing without making me sign up for anything, the auto follow on clicks actually feels smooth and not janky like the other tools i tried

Finally a screen recorder that doesn't dump a watermark on my demos. The auto-zoom following my clicks actually feels smooth instead of gimmicky, and being free with no trial nonsense is a rare treat.

finally a screen recorder that gets the auto zoom right without making me fiddle with keyframes. zoomed right to where i clicked, smooth and clean. solid for product demos

The auto-zoom following my clicks actually feels smooth, not jittery like some others i've tried. Really nice for quick demo videos.

Finally a mac recorder that follows my clicks without me babysitting keyframes, did a Loom-style demo in like two minutes. Surprised something this clean is free with no watermark.

finally tried this and the auto zoom actually follows my cursor smoothly without weird lag, super clean output for a free tool.

the auto zoom that follows clicks is genuinely clever, always wondered why more screen recorders didnt just do that. nice work keeping it free with no watermark too

looks cool! However, what exactly makes this different from products like remotion, descript, or even tella?

How do you measure whether AI-generated videos actually outperform traditionally edited content in terms of engagement?

love it!!! I used to use screen studio to record demo. How is this different! Also whenever I'm making product demo on my own laptop, i always need to hide sensitive information. Would be so nice if there could be a function to blur information but with motion!!

We produce a lot of product screen recordings and editing is 80% of the time cost. What does 'edits itself' catch in practice, dead time between clicks, or does it also handle captions?

The auto-zoom-on-clicks is exactly the hour I lose hand-keyframing every demo, so that is the part I would test first. When the auto motion misjudges a moment and I drag or retime that zoom segment, does my edit survive if I re-record just one section later, or does the timeline assume a single continuous take? And on export, can I pick resolution and format, or is it one fixed output?

Auto-zoom toward clicks is built around cursor events, so I'm curious how it behaves on an iPhone simulator recording where the "cursor" is a tap that doesn't exist between touches. That's my actual use case, App Store previews cut from simulator captures, and every tool I've tried treats it as a screen recording problem when it's really an attention problem. Local-only processing is the right call regardless.

This looks handy for product demos. One question: can the auto-zoom be turned off or overridden for specific segments, for moments where I'd rather keep the full screen steady? And can text captions or callouts be added, or is it purely camera motion? Thank you for keeping it free and local :)

How well does the auto-zoom hold up when you're switching between windows or apps quickly during a demo, and can you fine-tune the zoom intensity if it's too aggressive on smaller UI elements?

How does the auto-zoom handle recordings where I jump between windows or apps quickly? Curious if it gets confused and starts zooming into the wrong spot.

cool product!

how does the auto zoom know when to start and stop, like does it trigger only on clicks or does it also follow my mouse movement?

How does the auto-zoom handle recordings where you switch between apps quickly or jump around the screen a lot?

Congrats on the launch! Really liked the open .glideo folder. Keeping the raw capture and click timeline as plain data means a months-old demo can be restyled or re-cut to 9:16 without re-recording! That skips the proprietary project lock-in other editors lean on. Was capturing input events at record time (instead of inferring zooms from pixels afterwards) the choice that made the free, offline model workable?

Every founder has 40 raw screen recordings and zero published demos — the editing is the bottleneck, not the recording. Auto-zoom on clicks with no keyframing is exactly the right abstraction. Does it also auto-cut the "uh, wait, let me redo that" moments?

The auto zoom on click is genuinely satisfying, feels way more natural than the keyframe stuff I usually wrestle with. Also the no watermark thing is huge for me, can't believe that's free on macOS.

How does the auto-zoom actually handle really fast clicking or dragging across the screen without making the video feel jittery?

How does the AI actually help organize things day to day, like does it suggest structure when you import messy data or more of a chat assistant you have to prompt?

how does the AI actually help when building out a table from scratch, like does it suggest column structures or just fill in data?

Finally a spreadsheet that doesn't fight me when I want it to act like a database. The view switching is genuinely satisfying and the automations feel like they were built by people who actually use these tools daily.

The hybrid spreadsheet-database idea finally feels practical here. I built a simple tracker in about ten minutes and the automation options were way deeper than I expected.

The prompt-to-base generation is neat, but the thing that decides whether I'd build on it is getting data back out cleanly. Is there a read/write API and a full export that preserves schema and table relationships — not just flat CSV rows — so a base isn't locked in? And can permissions be scoped per field or collaborator, or is sharing all-or-nothing at the base level?

With AI being added across the Zoho suite, how do you ensure automation improves productivity without overwhelming users?

How does Zoho Tables handle permissions and access control when multiple teams are working from the same base, especially across different views like Kanban, Calendar, and Forms?

Is this like a mini version of Zoho Projects? This looks like a mini project tracker.

Love how clean the grid layout feels, especially with the way views and tables sit side by side without crowding the screen.

The mix of spreadsheet feel with database structure is a nice balance, and the AI suggestions actually picked up on patterns I hadn't spotted in my own data.

How does the AI actually decide what to organize for me versus what it leaves alone? Curious if I have to train it or if it just figures things out from the start.

The hybrid setup feels genuinely useful for tracking projects without juggling separate tools, and the AI suggestions actually caught a formula mistake I missed

how does this differ from airtable in practice? curious what ai features actually save me time versus just being a gimmick

How does the AI actually help organize messy data, like does it auto-tag entries or just suggest formulas when prompted?

How does Zoho Tables actually compare to Airtable when it comes to handling large datasets and complex relations between tables?

How does this actually differ from Airtable in practice, especially around the AI features and automation limits? Curious if the spreadsheet feel is closer to Excel or more like a database UI.

Tried it for a small project and the blend of spreadsheet and database feels more natural than I expected, especially linking related rows without formulas. The AI suggestions for cleaning up messy data were a nice touch too.

Does it play nice with existing Zoho apps like CRM or Invoice, or would I need to rebuild my workflows from scratch to get the automation piece actually working end-to-end?

How does the AI actually decide what counts as worth automating for someone like me just starting out? Curious if it learns from my patterns or if I have to manually set up every workflow from scratch.

How does the AI actually help organize my work day to day, like can it suggest formulas or flag messy data on its own or do I have to ask it directly?

how does the AI actually decide what to automate vs what to leave manual, especially when data structure shifts across teams?

love how it bridges the gap between a spreadsheet and a database without making either side feel like a compromise. the real-time collaboration feels genuinely fluid, not the laggy mock-version you usually get.

How does this actually differ from Airtable in practice? I get the spreadsheet-database hybrid idea, but the "AI brains" part is vague. What does the AI actually automate day to day, and is it included in the base tier or a separate cost?

Looks like a great addition to the no-code space. I like the combination of spreadsheets, databases, workflows, and AI in one place! seems like it could simplify a lot of team operations. Congrats on the launch !!! Question: can existing Airtable or Google Sheets projects be imported with formulas, automations, and relationships intact?

How does the AI actually work under the hood here. Like is it generating formulas on the fly or is it more of a natural language search through your rows. Curious how much hand holding it needs when your data gets messy.

How does the AI handle really large datasets, and does it slow down once you cross a few hundred thousand rows?

How does the AI actually decide when to automate a workflow versus just suggesting one, and is there a cap on the number of records or automations on the free tier?

Good to finally see an Airtable alternative in the Zoho ecosystem. Is this also a no-code app builder?

Love how Zoho managed to merge the flexibility of a spreadsheet with the structure of a database without it feeling clunky. The AI workflows are genuinely useful, not just a buzzword add-on.

Tried it with a quick family dinner plan and it actually picked who to text vs call based on past responsiveness, which was a nice touch. Setup was a little confusing at first but the AI suggestions saved me time.

Coordinating a group of humans is sneaky-hard UX...I fight this problem in a different domain. Question: what's the experience for the person being 'gathered'? Do they need the app installed, or does it degrade gracefully to plain SMS?

how does the AI actually pick between video, voice, and text for each person, does it factor in things like time of day or their contact info, or is it more of a guess based on what worked before?

Tested it with a quick team sync and the auto channel picker actually picked video for the urgent folks and text for the rest, which saved me the usual back and forth. Cool idea, works well.

Tried it to rally a few teammates for a quick call and the whole thing just worked without me bouncing between apps. The best part was how it picked video for one person and text for another based on context, feels like the AI actually gets the situation.

Tried it for a quick family meetup and was surprised it picked text over a call since I was driving. Nice touch that it just got on with it instead of asking me to choose.

does it actually pick the channel based on whether someone tends to respond faster by text vs call, or is it more of a static rule for now

Told it to rally a few friends for a weekend hike and it actually picked video over text, which felt surprisingly natural. The AI choosing the channel based on the group is a clever touch.

Curious how Kadoink handles the contact data side—do I bring my own list or is it pulling from calendars and socials for me? And does it play nice with Slack and Outlook, or is it more of its own thing for now?

how does it actually decide between video, voice, text, or push for a given situation, and can i override that choice if i know a specific person prefers one channel?

How does it decide which channel to use when someone prefers text but the group message would land better as a call?

Tried it with a quick family dinner plan and it actually picked up my sister on WhatsApp while texting my parents. Felt weirdly smooth for a first run.

Tried it with a quick family dinner and it picked everyone's contact preference without me doing the usual group chat ping pong. Pretty handy for one-off stuff like that.

How does it decide between video, voice, or text when reaching out, and can I override the AI's choice if I have a preference?

how does it actually pick between video, voice, text, and push for each person on the list?

Tried it with a quick family dinner plan and it picked video for my parents and text for my brother without me thinking about it. The auto-channel choice actually feels useful.

How does Kadoink decide which channel to use for each person, and can I override that if I know someone only responds to text?

Does Kadoink integrate with calendars and contacts already on my phone, or do I have to manually add everyone each time?

Told it to rally my book club with a quick venue change and it actually picked everyone through a mix of text and a quick voice note, which felt way less spammy than my usual group chat blast.

Told it to gather my team for a quick sync and it picked text for two people and a call for one who'd been offline. That kind of small decision-making is genuinely useful when you're juggling a group chat.

Finally tried Kadoink with a quick family dinner plan and it actually picked the right mix of texts and calls. Saved me the usual back and forth.

this made me smile, feels like something from a simpler internet. good luck today!

How does Kadoink actually pick between video, voice, text, or push for each person? Is that based on their past response history or some default logic you set up?

the smart channel routing sounds genuinely nice for the person sending the request, but it's optimizing for the sender's convenience, not the receiver's. a phone call is a much bigger interruption than a text, and the person on the receiving end never opted into "an AI decides when I get rung instead of texted." curious what stops this from becoming the group-chat equivalent of a robocall once a few people start using it on the same friend group, everyone's phone ringing because an algorithm decided calling would get a faster response than texting

Love how the AI picks the channel for you instead of just blasting everywhere. That mix of video, voice, and text routing feels like something Slack and WhatsApp should have figured out years ago.

Told it to round up my team for a last-minute call and it picked video for the remote folks and text for the ones in transit, which saved me from the usual back-and-forth. Curious how it handles larger groups, but the smart channel selection already feels like a real time saver.

The fill-a-grid-and-everyone's-phone-just-rings, no app or login, is a genuinely clever way to lower the friction of getting a group on a call. Two day-one things I'd hit immediately: does the ring-out work across countries and carriers, or is it US numbers only for now? And is there a cap on how many people one grid call can hold before it stops dialing everyone?

The name is fun, and the AI picking video vs text vs push for each contact feels genuinely useful. Wished it had a clearer pricing page on the site though.

Hey Product Hunt! 👋

I built Onlook because of a stupid little moment that kept repeating: I'd be deep in work at my Mac, the doorbell would go, and I'd… pick up my phone, wait for the Ring app to wake up, and miss the courier anyway. My cameras were 30cm from my keyboard the whole time.

Onlook puts your Ring cameras in the Mac menu bar. One click: live view of every camera. Click again: full view with two-way talk, so you can tell the courier to leave it by the gate without standing up. Motion and doorbell alerts arrive as native notifications — click one and you're looking at that camera. There's an activity log for replaying clips, and a little privacy lock that pixelates any camera when someone else can see your screen.

Two things I cared about while building it:

🔒 Everything runs locally. Onlook has no servers. You sign in with your own Ring account, your credentials go only to Ring, and video streams straight to your Mac. No tracking, no analytics.

🪶 It should feel like macOS. Download, drag to Applications, done — no dependencies to install. It lives quietly in the menu bar and updates itself.

Onlook is completely free, and it’ll stay that way. No subscriptions, no hidden costs — just a fast, beautiful way to view your Ring cameras right from your Mac.

Honest fine print: Onlook is an independent app, not affiliated with Ring — [and it's been through months of beta testing with real households to iron out the rough edges].

I'd love to hear what you'd want next — [multi-brand support (UniFi?) is also on the list]. Ask me anything! 🚪

Give this man a Golden Kitty!

Great work on this @jamesm

Design looks exceptional as always! Would love to see this working with other cameras in future 😃

This is awesome. Thanks @jamesm

Does it work with multiple Ring doorbells and cameras at once, or only one at a time?

great work James, nailed every details making it such a tasteful experience! love the offline first approach

Design looks great as always James!

Finally a Ring app that doesn't make me dig through my phone. The menu bar motion alerts are genuinely handy while I'm working.

This is amazing! 🤩

Can see this come in very handy, design is legit 👌

This is exactly how you pitch a finance tool. no fake hype, just clear utility and clear respect for user privacy huge congrats for shipping 🙌 @norteapp

How does Norte handle benefits that change mid-year or issuer-specific promos that aren't in your database yet?