PH热榜 | 2026-03-03

🎉 Proud moment for @Krisp .

Accent Conversion is a thoughtful addition to the Krisp desktop app.

Feels great to hit publish on this.

i am impressed with this development specially for meeting 7 day trial is enough for me. i wish it will be part of my office in future.

What accents does it support?

Hey PH 👋

How many times have you nodded in a meeting… and guessed?

Accents aren’t the issue. Misunderstanding is. And it costs time, clarity, and sometimes credibility.

1.1B people speak English as a second language. Global teams are the norm. AI voice agents are scaling fast. Yet both humans and AI still struggle with accented English.

So we built Accent Conversion

It runs on the listener’s side and converts accented English into neutral American English in real time. The speaker doesn’t change. They don’t install anything. You just understand — the first time.

Fully on-device. Near-zero latency. No weird robotic voices.

Would you use this? Or is misunderstanding just part of global work we’ve learned to tolerate?

Let’s talk.

Asti

Congrats on the launch.
Would love to see this also as a Chrome extension for YouTube.

that's a kewl video ! congrats on this upgrade.
the accent conversion feature has been there for a year, but does this mean - that this new update willy only update the selected words? and how?

I've been a long-time @Krisp fan, going back to at least 2019, when I hunted Krisp for Windows!

Now the team's back with something new: Think noise cancellation, but for accents.

You’ve been on calls where you think you caught what someone said — but did you really? You smile, nod, and hope you got the gist. But when accents are thick and clarity counts, can you risk it?

Recently I was on a call supporting a founder with their Product Hunt launch whose first language was clearly not English (I won't say who!). Try as I might, I could only discern 75% of what they were saying. Fortunately I was also supporting @asti_pili and @lusine_mnatsakanyan4 with this Krisp Accent Conversion launch! Once enabled, my comprehension went up to 97% and we finished the call with a clear plan to get them to launch on time.

The technology behind Krisp's Accent Conversion tech is genuinely impressive: it runs locally on your device's CPU, and dampens accents in real time. You hear the speaker’s voice—but minus any heavy accent that can get in the way of comprehension.

If you've used Krisp's products before, you know they can ship real breakthroughs in audio.

This one feels like the next big one.

If you’re on global calls or work with a distributed team—this one's worth trying out.

Great work @Krisp team! This removes whole bunch of communication barriers!

Really happy to share this launch. Accent Conversion is designed for real meetings, not lab demos. Thanks for checking it out.

Proud moment for @Krisp. Accent Conversion tackles the listener-side problem in meetings. Curious what use cases pop up first.

Congrats on the launch. It's indeed a great product for global and multi-lingual teams.

Big milestone to celebrate today. 🚀

Accent Conversion makes accented English easier to follow in real time. Curious what use cases pop up first.

Excited to see this live today. Accent Conversion is designed for real meetings, not lab demos. Happy to answer questions in the comments.

Hey PH 👋

Big milestone to celebrate today. Accent Conversion is a big step for our Voice AI work.

Happy to answer questions in the comments.

Excited to see this live today. Accent Conversion brings the ‘hear it once’ dream closer to real life. Hope you try it and tell us what you think.

Congrats on the launch team.
I've definitely faces this while being on call with different stakeholders

Would love to see this on slack as well

Finally. I've been nodding along in global meetings for 3 years understanding 60% and hoping the other 40% wasn't critical. Spoiler: it was always the critical part :)

Finally shipping this feels great. Accent Conversion is built for those misheard moments in global calls. Thanks for checking it out.

Huge day for the team. Accent Conversion focuses on comprehension, not changing speakers. Happy to answer questions in the comments.

Huge congratulations to the @Krisp team on the launch! 🎉

This is truly one of the most impressive features I’ve seen from Krisp. Solving real-time accent friction in such a seamless way is a big leap forward for global communication.

Amazing to see how Krisp continues to innovate and solidify its position as one of the leading players in the Voice AI space.

This has been a long build. Accent Conversion brings the ‘hear it once’ dream closer to real life. Would love feedback from anyone in global teams.

This is super cool! Curious how well it handles overlapping speakers / fast back-and-forth conversations?

I tested it and...whoa! I love it!

It's a pity I only get 60 minutes a day, as I'm already paying for the subscription I was expecting this to be included.

How do you do that by the way, is it locally executed? As there is almost no latency, it must be.

I work with contractors in four time zones. It's the mental tax of re-parsing every other sentence while trying to stay in the conversation. Does this run locally or does audio hit your servers? I'm prepping my own launch soon and real-time audio latency is something I think about a lot.

Congrats on the launch. I think India could be a great market.

This is a lifesaver for global teams! Clear communication is so important, and being able to convert accents in real-time will definitely boost confidence for many speakers. Does it support multiple languages for conversion, or is it focused only on English right now?

Hilarious video. I actually clicked this feature on when I first signed in this morning. I've been hoping for it for months. Really hoping it can do this and help me FOCUS. :-D

Very cool, congrats! and nice ad ❤️

Is the accent conversion only available for certain languages? I'm Greek; will it be helpful for me?

Hi everyone!

The Qwen team just dropped the Qwen3.5 Small Model Series: 0.8B, 2B, 4B, and 9B, along with their Base versions.

This release fills the missing piece for on-device deployment and completes the full Qwen3.5 matrix from 0.8B all the way to 397B. Now you have clear choices:

• 0.8B/2B for embedded/IoT/Mobile

• 4B for lightweight multimodal agents

• 9B for edge servers

• Plus the bigger MoE models for heavier workloads.

The 9B is the real shocker: matching or beating GPT-OSS-120B on several key benchmarks while being 13x smaller.

Even Elon chimed in:

Edge AI is heating up fast. This opens up exciting new opportunities for AI hardware and local innovation.

Play with these models on @Ollama!

Native multimodal at 0.8B is genuinely impressive - most teams trade off size for capability, but 262K context windows + text/image/video in under 1B parameters changes the edge deployment math.

The 9B beating GPT-OSS-20B on GPQA Diamond is interesting. Curious about structured output reliability at 0.8B though - small models tend to drop JSON schema adherence under complex instructions. Is there a differentiated training approach for structured data tasks?

MTP for faster inference on constrained hardware is a smart addition. Real-world throughput numbers on consumer GPU vs Apple Silicon would help developers size their deployment targets.

The 9B punching above its weight class is the real story here. Running capable models locally without needing a data center changes what's possible for privacy-conscious apps and edge deployments. Been waiting for small open-source models to close this gap.

The 9B matching models that size is wild — smaller models getting this good makes edge + local AI way more practical than most people realize.

What kinds of tasks can be performed with these models?

@binyuan_hui @chen_cheng1 @junyang_lin Hi guys. Can non-technical developers use these models easily? What tooling or platforms do they support?

I'm writing this from Dubai.

For the past three days, Iran has been firing rockets at the city. Flights are canceled. The airport is partially closed. I'm stuck in a hotel room watching smoke rise over Jebel Ali.

We launched Viktor anyway.

Think of your best coworker. We launched someone better. Maybe even better than you.

Viktor lives in your Slack, connects to 3,000+ tools, and does the actual work: reports, code, web apps, ad campaigns.

While I've been stuck here watching the news, Viktor posted 28 real-time missile updates to our team Slack, tracked every team member's flight status, and told us when to shelter in place. Then it ran our ads, flagged a spend anomaly, and shipped a PR to our codebase. Nobody asked it to.

Over 1,000 teams use Viktor. Backed by Daniel Gross, Nat Friedman, and the founder of ElevenLabs. Salesforce lists us in their app store.

You can't make this up.

Fryderyk, co-founder.

P.S. Viktor is coming to Microsoft Teams very soon. Very.

I run growth at Wispr Flow, so I spend most of my time deep in AI tools and automation. I'd been following Fryderyk and the Zeta Labs team since Jace, their AI email assistant, and was impressed by how fast they shipped product.

When Fryderyk showed me Viktor, it clicked immediately. I'd been spending hours building automations in Claude Code - stitching together context, writing scripts, trying to make things persistent and scheduled. Viktor did all of that natively. It just lives in Slack, already has the context from your tools and conversations, and runs on its own.

But the thing that genuinely blew me away was the proactive behavior. Viktor doesn't just wait for you to ask. It observes how your team works, chimes in when it spots something relevant, and suggests automations you didn't think to set up. I've never seen an AI tool take initiative like that.

I ended up advising the team on growth strategy because I believe this is how every team will work within a few years. Not another tab. Not another tool. A coworker that lives where your team already communicates.

If you're skeptical, give it your worst task. That's what convinced me.

Can't wait to try it!

I use Viktor daily for managing support. When a ticket comes in, it gets forwarded in Slack automatically and tags Viktor - it pulls up the customer's account in Stripe, checks activity in PostHog, searches for related bugs in Linear, analyzes code base, and drafts a response. What used to be a 20-30 min investigation per ticket now takes under 2. Absolute game changer.

Launched Viktor on my GTM stack (Lemlist + GA4 + Internal Database). Viktor managed to find a few mistakes (mine, unfortunately), helped me analyze which outbound campaigns worked best for which persona, then proposed a few fixes to better attribute traffic (I hadn't added UTMs for our outbound campaigns). When I ghosted him for a day, Viktor told me we cannot wait on this, generated its own UTMs, and asked for permission to apply all of them.

Would recommend, 10 out of 10.

Congrats on the launch! Viktor’s been on our team for two weeks, and I’m already used to handing him work 🔥

I have been using Jace for a while, a tool from the same team and when Viktor came into the scene, I also ran it and I love it! I've run out of credits, but it's pretty credit efficient and useful.
Really really recommend this tool and its team!

Viktor submitted a PR while one of our engineers was on vacation. It read the codebase, found the issue, wrote the fix, opened the PR with a proper description. The engineer approved it from his phone. That was a Wednesday

I'm on the support team and one morning Viktor just dropped a list in Slack - it had gone through PostHog, found 12 accounts that hadn't logged in for two weeks with open tickets, and cross-referenced Stripe to flag which ones were paying customers. Nobody asked it to do this. It just noticed they were slipping through the cracks before I did.

good luck boys

Also what is the overall success rate of doing tasks in your internal testing? Can victor hallucinate?

We got early access to Viktor because we built their website - so we've been testing it in beta for a while now.

I'll be honest, day 1 was just okay. Cool tool, nice concept, but nothing crazy. Then around day 5-6 something clicked. Viktor had absorbed enough of our Slack context to actually understand how we work, who our clients are, what we care about. That's when it went from "another AI tool" to something I genuinely rely on.

The thing that got me first was the sales prep. Before every call, Viktor just drops a full persona report - who the person is, how they match our ICP, what we should lead with. I used to spend 15 minutes scrambling through LinkedIn before meetings. Now I just read what Viktor already prepared.

But for an agency like ours, the real game-changer is how it connects everything. We run ClickUp, Clockify, Slack, Meta Ads, CRM - the usual chaos. Viktor sits across all of them. It auto-generates Clockify reports we send to clients, flags Meta ad spend issues, pings us before deadlines slip. I stopped opening half of these tools directly.

It even generates UI wireframes based on our brandbook straight in Slack. HTML, on brand. For a design agency that's wild.

Two weeks in and Viktor honestly became one of the most productive "people" on the team. Huge congrats on the launch.

Rooting for Product of the Day 🚀🖤🤝🏻

Let’s go Viktor!

Hey, I work on growth at Zeta Labs. Here's what surprised us most.

AI tools have terrible day-1 retention. Nearly all of them. Users sign up, try it once, leave. We spent months assuming this was a product quality problem.

It wasn't. It was the blank prompt box.

Someone opens a new AI tool, sees an empty text field asking "How can I help you?" and freezes. They type something generic. Get a generic response. Decide the product isn't for them. The product was fine. The entry point was broken.

Viktor flips this. It doesn't wait for you to figure out what to ask. It joins your Slack, reads what's happening, and messages you first.

"I noticed your team's been going back and forth on X. Want me to handle it?"

Your first experience isn't inventing a prompt. It's saying yes or no.

That one design decision moved our activation numbers more than any feature we shipped.

And "chat with AI" is the wrong model for work. A good coworker doesn't sit in a corner waiting for instructions. They pay attention and come to you with context. That's what Viktor is.

Happy to go deep on anything: how we built the proactive engine, growth experiments that failed, whatever. Ask below.

Antoni

Been using this for a few weeks, insane!

Viktor looks incredible, will definitely give it a try! Congrats on the launch!

Tremendous work ethic. Hope you guys land safely. Cheers on shipping a cool product. Signing up right away :)

congrats on the launch! meaningful step beyond copilots.

ps: so if your AI tools answer questions and Viktor does the work… does that make Viktor the one actually earning the promotion? 😉

I've been testing out Viktor for the last few weeks and it's awesome. It's able to give me daily insights to what competitors are doing and who they're working with. Plus I love how proactive it is at telling us what automations should be built

Good luck to the team being stuck in Dubai! Hope you're safe 🙏

Then, congrats on the launch! I'm curious, is there any technical challenge you solved and can openly talk about? And is adding your own API key on the roadmap or against the vision?

Asking because it's not obvious to me what's best between a) closed source and better outputs but being chained to a system that burns thousands a month and b) somewhat lower quality output? (although agentic OSS is progressing insanely fast) but no added margin on token and possibility to run things on your own hardware. Or maybe I'm just not ICP 😄.

Thank you!

Any ways to prevent leak of company data or to filter sensitive information?

Very cool! Congrats on the launch team!

I worked on Viktor in the last months and there are some genuinely impressive technical breakthroughs in it.

The agent is one the first ones able to use 10k+ tools without any context regressions by doing an on file system routing and is able to compose tools via code into more complex actions.

It also is significantly more proactive than any agent you might have interacted with before.

Viktor really helped us to stay up to date in Dubai! I also use it for data analysis and debugging while walking around town all the time lol

We had an error spike. Viktor read the logs, cross-referenced with recent PRs, identified the likely cause, linked the relevant code, and opened a Linear issue with full context. By the time anyone saw the alert, half the work was already done.

Launched from a warzone is a hell of a backstory and already builds trust damn. The "runs for weeks without losing context" claim is the part I want to hear more about since that's exactly where every long-running agent stumbles. Congrats on shipping Viktor.

This is a rlly interesting... Curious what kinds of tasks teams are seeing Viktor take over first in practice?

too bad it's not available in EU due to GDPR but it looks great! I used LLMs heavily to analyse results from previous tests I took and the tips I got have been useful

Its cool. But Maybe add a interim assesment at milestones. At 15% complete there should be some insight, so that the user is encouraged to continue. # my two cents 🙏

I had the opportunity to beta test this app.

If you want (or need) to be seen, this is it.

You don't need a therapist. You can do it on your own time. You can save some money. You're also likely going to get better results.

Opinions are from my own personal experience :)

This looks fantastic. I love the level of depth that you're going into. I can imagine it will spark a lot of good insights.

Almost 20 years ago one of my execs made everyone at Justin.tv (the company that turned into Twitch) take the DISC personality assessment. I thought it was a silly idea at the time, but immediately after we got our results I learned so much about how my work personality interacted with all my cofounders and other leaders in the company. Love this idea!

THIS IS AMAZING -- super helpful. Can't wait to use for our staff :)

After seeing how accurate my Deep Personality assessment is, I asked ChatGPT to recommend 20 activities based on the profile to elevate happiness and ensure meaning as I approach retirement. I am looking forward to pursuing all of them! Well done, Andrew. I have already recommended Deep Personality to family, friends and business associates.

It is cool! I've seen how hard Andrew has worked on this, how much he's put into it, to make it legit.

This tool is such a gamechanger. I took it, then my wife and mom took it, and then we told friends about it and they've had their work partners take it and everyone has found it super enlightening despite having taken many personality tests before.

When I saw this on X I was curious.

But when I took the test I couldn't believe how much i had learnt about myself.

Things I struggled to explain had a reason for being the way they were.

And when i compared my results to my wife's results. It shone a glaringly big spotlight on why our issues were our issues.

This tool was one of the biggest changes in the last couple of months for me.

Thank-you so much for creating it Andrew.

I’ve done a bunch of personality profile assessments as well as therapy. Deep Personality aligned with all that prior effort to give me comfort that it's sound and make the new or different insights all the more interesting and useful. The HR companies aren't in trouble in theory, its here.

Congrats on the launch Andrew!

The origin story of asking ChatGPT what it would need to be the ultimate relationship coach and then just building the answer is a great way to solve a product definition problem. The custom AI prompt pre-loaded with your psychological data is also the feature that makes everything else click. No more spending six sessions explaining who you are is a real unlock.

I'm married to a psychologist, so we're both going to take this and compare notes, which should make for a very interesting evening. Curious whether the couple compatibility report holds up under professional scrutiny or surfaces things even a trained eye would find surprising.

@awilkinson One question totally unrelated to the product itself: how did you make the promo video? The motion and pacing are really smooth, and I'd love to know the workflow behind it. Congrats on the launch!

Excited to hear everyone's feedback!

👋 Hey Product Hunt, I'm Miron, founder of SuperMoney.

Nearly 9 in 10 Americans started 2026 feeling financially stressed, according to the National Endowment for Financial Education.

My wife and I were part of that statistic coming out of college, buried in debt. I became obsessed with tools like Mint, but frustrated that you couldn't actually compare financial products side by side with any real transparency.

So in 2013 I started SuperMoney as a financial product reviews marketplace. I thought it was wild that no one had done it.

Fortune 500 companies apparently agreed, sending us cease and desist letters in the early days. We kept going anyway, and it took off.

By 2017 we were already sketching out what an AI financial advisor could look like internally. The problem we kept coming back to: traditional financial advisors operate on an assets under management model, so they're built for the top 1%.

Everyone else is dealing with real financial stress and getting no guidance at all.

The prototypes never quite got there. When LLMs arrived, we knew it was time to actually build it, and give people something most of them have never had: genuine financial calm.

That's what today is. Ten years in the making.

The SuperMoney App connects your accounts and actually does something with the data. Not just pretty charts. It:

1. Builds your budget, tracks your spending, sets goals, and suggests what to do next to achieve them.

2. Serves you a feed of insights based on what's actually happening in your finances.

3. Surfaces lower rates on your existing loans and smarter debt paydown paths. Real savings, not just insights.

4. When you have a question, you can ask it directly. The AI knows your accounts, your credit, your loans, and your goals, so the answers are actually relevant to you, not generic advice you could Google.

If you earn good money but still feel financially overwhelmed, this was built for you.

Try it free for 2 weeks. If you do, I'd love to know: what would you want us to build next? 🙏

This could be translated into many languages to make more people across the globe financially aware. 🫠

How do you protect user data and what is your USP?

Hey Product Hunt! 👋

Excited to finally have this out in the world! I worked on the AI chatbot — the piece that lets you ask real questions about your money and get answers based on your actual financial picture.

If you try it, I'd love to hear what surprised you (good or bad). Here all day to answer questions!

👋 Hey Product Hunt! So excited to finally share SuperMoney App with this community.

We've been building toward this for over a decade, starting as a financial product comparison platform and now evolving into something that actually tells you what to do with your money, not just what's happening.

The AI doesn't just surface data. It thinks through your situation and gives you a clear next step. That's what we couldn't find anywhere else, so we built it.

Happy to answer any questions about the product, the journey, or the vision. Let's go! 🚀

whenever i see @benln hunting the product , its always valuable

Hi Product Hunt! So excited to be live. 🚀

Since I started out as a CFP I've wanted to make high-end financial planning accessible to everyone, not just the wealthy. Our team has been building toward this for 10 years, and I’ve loved helping bake real financial logic into Sense AI. I’ll be here all day to answer any questions you have. We’re looking forward to your honest feedback!

Hey Product Hunt!

I'm one of the product managers at SuperMoney, and today's launch is the culmination of years of work.

My focus has been on the insights experience—delivering tailored tips based on your data and spending habits, not generic advice. The app doesn't just show you data. It tells you what it means and what to do next. Simple, clear, actionable.

Happy to answer any questions about how we built this!

Big day for the team! 👋 It’s incredibly rewarding to finally introduce the SuperMoney App to the Product Hunt community.

As the SEO lead here at SuperMoney, I spend my day looking at the financial questions people are actually asking Google. The truth? Most 'how-to' articles give the same generic advice to everyone.

What’s been so exciting about building Sense AI is that it finally moves past the 'top 10 tips' and gives you a personalized roadmap based on your actual debt and rates. It’s the difference between reading a map and having a GPS.

We’ve been in the game since 2013, so we didn't just 'add AI' to a spreadsheet. We built this engine to understand the actual mechanics of the financial products you're using.

We're going to be hanging out here all day, so please drop any questions!

To the PH community: I'm curious, if you could 'Google' one thing about your personal finances and get a 100% accurate, personalized answer instead of a generic article, what would it be?

Looking forward to your feedback!

Hey product hunt, I am a software engineer over at Supermoney.

The team has been hard at work on this product and are eager to share it with you. Looking forward to your feedback and hopefully we can help improve the world a little bit!

Hey Product Hunt! 👋🏻

I’m the Lead Product Designer at SuperMoney, and over the past several years I’ve had the privilege of working closely with our incredible team to craft and refine the SuperMoney experience across every platform.

Today, we’re excited to finally share the SuperMoney app with you. SuperMoney is a powerful personal finance management tool designed to help you take control of your financial goals with clarity and confidence. At the heart of the experience is Sense AI — your intelligent companion that provides personalized insights, smart recommendations, and ongoing support tailored to your financial journey.

We’ve put a lot of thought, care, and iteration into building something that truly empowers users, and we can’t wait to hear what you think. Your feedback means everything to us — let us know how we can make SuperMoney even better for you!

How could I migrate all my financial status and data to SuperMoney? It could be quite tedious sometimes...

Hey Product Hunt, I am one of the product managers that worked on the app.

We're very happy to launch the SuperMoney app. At SuperMoney, we have more than a decade of experience providing financial tools and services to our customers, and this app is the next evolution of our mission to provide automated financing and insights to help you super power your money. Our app (available on web, iOS, and Android) is an AI-powered decisioning platform for all your financial needs that provides you with the tools and means to help you get out of debt, build your financial portfolio, and gain useful insights into your financial health.

We'd love to hear any of your questions about how SuperMoney can help you meet your financial objectives and what we have in store for you in the future

So excited to finally see this out in the world! The team has poured an incredible amount of care into this and it shows. If you've ever felt overwhelmed by your finances, give it a try — this is the tool we wish existed. Proud of everyone who made it happen. 🚀

Hi Product Hunt! 👋

I’m a software engineer at SuperMoney and I am so excited we can finally share this app with everyone! Our team has been working hard towards this goal for years and we can’t wait to hear your feedbacks.

How does product aggregation work on the platform? Do you partner with all financial institutions and lenders? Essentially, how can we, as users, be sure we're getting an unbiased view into the best possible options for our personal situation?

Hey Product Hunt!

I am VP of Operations at SuperMoney.

Today is a day I have been looking forward to for a long time.

What I want to highlight from my perspective: the SuperMoney app is built on top of relationships with the broadest set of financial service providers in the space. That means when the AI recommends something to you, it is not generic advice. It is backed by real options that actually exist for your specific financial situation.

That robustness is what allows our AI to move beyond tips and actually tell you what to do next. It is not just smart data analysis. It is smart analysis connected to real, actionable solutions.

Huge congratulations to the team. We are here all day and would love your feedback.

Hi everyone, thank you so much for all your support and helping us get to #5. Onward and upwards from here on!

Super cool! Sad to see all those predictions happen but love the creativity here.

wow this is a game changer for people like me there are many claims i am not able to crosscheck all

How cool is that, since I have watched only a few episodes and we have the internet full with such claims it would be nice abd useful to fact check them.

More accurate than reddit about who's going to win the next election. Crazy stuff. Congrats on the launch

Hi everyone 👋

I kept wondering If thousands of people claim “The Simpsons predicted this”…

Why was there no actual database?

Springfield Oracle is our attempt to turn internet chaos into something structured.

The real question is: What happens when pop culture becomes data?

Curious to hear your thoughts.

Love the name behind this idea, spot on and very pop culture. Congrats on the launch, @isha_godboley!

super super cool! got an early preview and i spent a LOT of time going through this.

this is the kind of personal projects to like to see 🫡

Honestly, in a sea of AI agent projects, it’s so refreshing to come across something like this! The branding and visual identity really stand out too, it's really nice !

Just out of curiosity, how long did it take you to build this database? And what’s your process for keeping it updated as new viral clips (or new predictions) pop up?

Great work, in any case!

The deepfake problem is what makes this actually necessary rather than just fun. When half the viral clips are fabricated, the whole "Simpsons predicted it" phenomenon becomes impossible to reason about without a sourced database. This is the infrastructure that should have existed years ago.

The scoring system is the right call too. Confirmed with receipts versus Pending versus debunked are very different things and collapsing them all into "the Simpsons predicted X" is where most of the misinformation comes from.

Curious about the methodology for borderline cases. When a prediction is genuinely ambiguous, like something that kind of happened but not quite the way the episode described, how does the scoring handle that nuance? Is there a partial match category or is it binary?

This is called gold

This is great, @vinoop_thayatt! Does it sort your email into folders as well? Congrats on the launch!

This sounds like a powerful tool. Since it supports sending up to 1,000 emails at once, does it also have safeguards in place to ensure deliverability and prevent messages from landing in spam or promotions?

Congrats on the launch! This looks like a solid upgrade for teams that need both marketing and transactional emails in one place 🚀

This looks super interesting - great job Vinoop and team. Keen to test Mailercloud out.

The move from email marketing tool to full email infrastructure is the right expansion. Teams building SaaS products don't want to stitch together separate providers for transactional and marketing emails, and the deliverability risk of doing it wrong on OTPs and password resets is too high to cut corners on.

99%+ deliverability is the number everyone claims but the real question is how it holds up on cold domains and newly authenticated sending infrastructure. How does Mailercloud handle the warm-up period for new domains, is that managed automatically or does the developer need to control it manually?

Building a SaaS platform myself right now and reliable transactional email for things like billing confirmations and auth flows is exactly the kind of infrastructure that needs to just work without babysitting. Congrats on the launch!

Hi Product Hunt! 👋 I'm Jensen Harris, one of the founders of Textio, and I've lost a lot of sleep over the last year waiting for this day 🤞

Here's the deal: I hate many parts of the standard job interview process SO MUCH.

Don't get me wrong, I love hiring people. Meeting smart, engaged humans and figuring out how they could thrive—that part is genuinely wonderful. But all the make-work and drudgery that goes into a standard interview process? It fills my heart with dread.

First the prework, where I think about what I intend to ask and then promptly forget during the interview. Then the endless notes I have to take that make the candidate think I'm multitasking. Don't forget the post-interview feedback essay I'd put off writing for days because I had an actual job to do, by which point I'd forgotten half of what we talked about. Oh, and the debrief meeting a week later where everyone shares half-remembered observations and their most visceral gut feelings.

Candidates who receive job offers are 12x more likely to be described as having "a great personality" than rejected candidates. No wonder, it's probably all most people remember. Great way to make hiring decisions! 🤮

Here's a story I've been embarrassed to tell publicly: I've never successfully been hired for a job as an adult.

I started at Microsoft through a college intern program and just never left. When I was ready for my first big career change, I was rejected dozens of times. Seven full onsites, zero offers. For one, I prepped for weeks, flew a thousand miles, bought and put on the scratchy wool suit they required (!!!). Spent two days with them, felt optimistic. And when the rejection call eventually came weeks later, they described conversations I know for a fact I never had. They'd mixed up my feedback with someone else's.

After all that effort, they'd lost track of who I even was.

That feeling—"they didn't actually hear my words!"—still sits in the pit of my stomach over a decade later.

It was in Chicago, digesting a (mediocre, honestly) deep dish pizza with the team when what became Lavalier showed up suddenly like a hot fire in my brain. What if we centered the entire interview process on the candidate's own words? Not the game of telephone from interview to notes to feedback form to verbal debrief, but what the candidate ACTUALLY SAID!

I wrote the very first lines of code at the Alaska Airlines gate in Chicago's Terminal 2 as the core parts of the platform started to form in my mind: role intake that helps recruiters define what actually matters in minutes instead of meetings. Interviewers who can actually be present in the conversation instead of focus on transcribing it. Recruiters who can ask something like "which candidates have deep experience mentoring" and instantly get back answers across 40 hours of interviews, linked directly to the candidate's own words. Hiring decisions based on evidence, available in hours—instead of on weeks of fading memory and half-remembered vibes.

Now, Lavalier doesn't make hiring decisions—we believe humans should make hiring decisions. And Lavalier supports those human decisions by giving hiring teams the data they need to make intelligent, informed decisions based on what actually happened during the interview.

This is the most exciting product I've built in my career. It's the first thing at Textio I've personally written a substantial percentage of the code for since the original product back in 2014. For our company, it feels like a rebirth—a second chance to think boldly about how to take a really hard set of hiring problems and make them radically, magically better.

We built one of the first products to bring AI to HR over a decade ago. Since then, companies like Bloomberg, Cisco, Johnson & Johnson, and Spotify have relied on our products every day, and we've picked up some cool recognition along the way (Fast Company Most Innovative Companies, Forbes AI 50 list).

We're launching today at lavalier.ai and could really use your feedback! I'd love to hear everything you have to say, and our team will be iterating fast on what we learn because we want Lavalier to be the most indispensable software in your hiring process. I'll be here in the comments throughout the day!

Oh, and if you click the Visit button on this page to sign up for Lavalier, special deal—we're unlocking 500 bonus credits for you (enough to hire for your next 5+ roles for free.)

So let's start here: what is the one thing that makes interviewing—as a candidate or an interviewer—most painful for you today?

I was in the room when this product was first imagined — what an honor. I’ve been along for the journey from the very beginning.

I watched the hours. The hard conversations. The relentless work. The heart and soul poured into this build. And what stood out to me most was the insistence that if we were going to create something new, it had to genuinely respect the humans in the process.

That’s why this launch hits differently for me.

Yes, I’m excited about the solution.
Yes, I’m proud of what the team has built.

But I’m even more excited for the recruiters and hiring leaders who will use it.

Because this won’t just help them move faster.
It will help them hire better.
But more importantly, it will create a better experience for the most important humans in the process, the candidates.

And after 30 years in recruitment, that’s the part that matters most to me.

Congratulations, Team Lavalier! It has been magic watching this come to life. I had the honor of leading Textio through the augmented writing wave, building a team that was thinking about practical outcomes in AI (and bias in AI) before foundation models were broadly available. As you've created Lavalier to solve modern recruiter problems, I've watched from afar as you systematically think through 1. what makes a great experience for candidates, hiring managers, and recruiters alike, 2. how to place fairness and auditability at the center of the interviewing process, and 3. making sure the process is AI-enabled but led by human decisions and interactions throughout. Excited for more people to benefit from what you have created!

Congrats on the launch , keep making more great products like these

Loving using Lavalier! We use it extensively at Cowboy Ventures when we are doing internal hiring, and have noticed improvements in our hiring process from higher quality deliberations to faster decision making. Can't imagine hiring without it now.

I'm Colleen, CEO of Textio (and yes, I can attest that @jensenharris is not lying about not sleeping!)

I was also in Chicago for that pizza — and for the record, it was excellent deep dish, I'll hear no slander. What Jensen described as a hot fire in his brain started in a conference room in one of the greatest cities in the world (I'm from Chicago, I can confidently say that).

I mapped out the entire landscape of how hiring actually works — every step, every handoff, every place where signal gets lost. Where our customers were telling us the most friction lives. Why all the ways people have tried to fix it keep failing. And it kept coming back to the same thing: the fixes all depend on making another human do something that's either not their top priority or something they're just not that good at — remembering everything from a conversation, assessing it against job requirements they may not have even read, and then writing it all down in a way that's clear and useful. That's a lot to ask of someone who has an actual day job on top of interviewing.

That's where Lavalier started. Not with the technology, but with that problem.

I'm using it right now to hire for a role on my own team. It's changed how I show up in interviews — I actually listen instead of frantically typing, and afterwards I can see whether I assessed what I said mattered or just followed the conversation wherever it went. That honesty is uncomfortable but it's the whole point.

Try it and tell us what you think — we're building this fast and your feedback matters.

I had the opportunity to beta test Textio's new product, Lavalier, and I like it more than I expected to. The suggested questions and follow-ups were especially helpful for less experienced interviewers, and the notes it generated were stronger than what we usually capture — the quotes and structured summaries made writing feedback easier without starting from scratch. I also appreciate that it doesn’t try to make hiring decisions for you, which feels important. That said, the first time I used it I wasn’t totally sure if I was navigating it the “right” way — a little more in-app guidance would help. It’s simple, but there are clearly deeper features that aren’t obvious right away. Overall, it feels genuinely useful, especially for teams who want better interview quality without handing decisions over to AI.

It's cool!!!

Is there any version for interviewee?

I can see the value for teams running dozens of interviews a week — consistency across that many hiring managers is a real problem. But for a team of 3-4 people who interview maybe twice a week, we already piece this together with Teams recordings and AI summaries. What would Lavalier give us that that workflow doesn't?

Congrats on the launch! The interview space definitely needs better tooling. Does Lavalier work well for technical interviews too, like system design rounds?

🚀🚀🚀 We're excited to launch Skyvern MCP — give your AI assistant superpowers to browse the web, fill out forms, extract data, and run multi-step automations. Works with OpenClaw, Claude Code, Codex, Cursor, or your custom agent.

🎯 Get started at app.skyvern.com or check out the open source repo at github.com/Skyvern-AI/skyvern.

Skyvern MCP connects your favorite AI assistant to a real cloud browser through the Model Context Protocol. Instead of writing Selenium scripts or wrestling with CSS selectors, you just tell your AI what to do in plain English — and Skyvern handles the rest.

⚡ Setup takes 30 seconds. Seriously.

No Python. No pip install. No local server. One line of config, and your AI assistant can browse the web:

claude mcp add-json skyvern '{"type":"http","url":"https://api.skyvern.com/mcp/","headers":{"x-api-key":"YOUR_API_KEY"}}'

That's it. Your AI now has access to 33 browser automation tools across 6 categories.

❓ How is this different from other browser automation tools?

Traditional browser automation (Selenium, Playwright, Puppeteer) requires you to write code, manage selectors, and handle every edge case manually. Skyvern MCP flips this on its head:

🗣️ Natural language, not code — Say "Submit this" instead of writing document.querySelector('#btn-submit-form-v2').click()

🧠 AI-powered extraction — Ask "extract all job listings with title, company, and salary" and get back clean JSON

✅ AI validation — Check conditions like "is the user logged in?" and get a true/false answer

🔄 Reusable workflows — Chain actions into multi-step automations you can run again and again

☁️ Cloud browsers — No local browser needed. Skyvern runs browsers in the cloud with geographic proxy support

🤔 What can you actually do with it?

Here are real use cases we see every day:

📊 "Go to Hacker News and get the top 10 posts with titles and scores" — Your AI opens a browser, navigates, extracts structured data, and returns it to you

📝 "Fill out this government form with my business details" — Multi-page form automation with intelligent field detection

🧾 "Log into my vendor portal and download last month's invoices" — Secure credential-based login + file download

🔍 "Search the Secretary of State website and verify this business registration" — Multi-step research across dynamic pages

💼 "Find remote Python jobs on Indeed paying over $150k" — Navigate, filter, extract, all in one conversation

🛠️ 33 Tools. 6 Categories. Infinite Possibilities.

🤩 We're open source!

Skyvern is fully open source with over ⭐️ 20K+ Stars on GitHub ⭐️. Check out the repo: github.com/Skyvern-AI/skyvern

Want to self-host? Just pip install skyvern

Want the MCP? Follow the instructions here

Want the skills? Follow the instructions here

🖥️ Works with all the tools you already use

• OpenClaw

• Claude Desktop

• Claude Code

• Cursor

• Codex

• Any MCP-compatible client

📞 Want to build complex automations?

We'd love to help! Shoot me an email at founders@skyvern.com and let's chat about what you're building.

🎁 Launch Offer

We're giving everyone a 10% discount to try Skyvern MCP. Get started here: app.skyvern.com and use the coupon code PH10 Happy automating! We'd love to see the creative ways you put your coding agent to work in the browser.

Let's go, @suchintan_singh!! I was excited to see that you and Shuchang are also launching today

Great product and a great team that really cares about seeing their customers succeed!

Big kudos, this is a genuinely great product you’ve built!

The MCP approach is smart — connecting AI agents to a real browser through a standard protocol instead of brittle Selenium scripts. How does it handle sites with complex auth flows or dynamic SPAs?

Just tried it out for our current setup with Skyvern - it's solid!

Congrats on your 4th launch! One thing I've been wrestling with on the MCP side: The protocol is stateless but browser sessions obviously aren't. When someone's agent logs into a portal, clicks through three pages, and pulls invoices, how are you keeping that session alive between tool calls? Are you pinning to a browser instance on your end?

That stateless-protocol-meets-stateful-browser tension is one of the more interesting design problems in the MCP ecosystem right now.

Skyvern can do so many incredible things - and being able to integrate it into llm workflows is such a choice upgrade. Nice work on this! Congrats to you and the team @suchintan_singh🙌

Hi all! Konstantina here — the founder of Kaedim.

We built Secret Sauce 3D because most online AI 3D apps stop where real production work begins.

Generating a mesh is easy. Turning it into an editable, usable, engine-ready asset is not.

For the past 5 years, our internal team of 100+ 3D artists has used AI tools we built in-house to ship real assets for AAA pipelines, under real constraints — poly budgets, topology standards, UV workflows, engine compatibility.

Today, we’re opening some of that tooling for all professional 3D artists out there!

Secret Sauce is an AI co-pilot designed to integrate into real workflows, targeting each step of the 3D modelling pipeline — while keeping everything editable and useful.

✨ What this means in terms of features we are releasing today:

• Generate high-poly foundations you can refine

• Optimise geometry without starting over

• Segment meshes to parts or combine parts into groups

• Create UVs you can actually work with

• Build texture foundations to use in your flow — not locked outputs

• Iterate step-by-step instead of regenerating from scratch

The philosophy is simple: Get a starting point, edit it forward.

You still need 3D skills. That’s intentional. We’re not replacing artists — we’re reducing friction so your time goes into improving your work, not recreating base geometry.

We’ll be shipping updates multiple times per week and evolving this alongside artists’ real workflows.

Hope you guys like it, we’d love your feedback!

If you give it a go, please comment on what you think would make this more valuable for your pipeline 👇

💭 We have also started a Discord server here for everyone to share ideas and experiments — feel free to join the discussion: https://discord.gg/HjEAXbxq

P.S. HUGE shoutout to the team that worked super hard on this for the past few months to get this over the line at a v high quality for a first release. WELL DONE 👏👏👏

Just dropped a full technical walkthrough for anyone who wants to see the actual workflow end-to-end 👇

🎥 https://www.youtube.com/watch?v=OHtvkIQYsB8

Let us know what you think, all feedback welcome!

A practical approach to AI in 3D, built to assist, not replace, artists. Editable geometry and UV stages make it genuinely useful in real workflows.

Is it possible to bring in an existing mesh and refine in further ?

a complete tool for me without needing me to jump here and there again. When I create a render scene and suddenly I notice there something missing in the scene I can directly create the missing asset just by a couple clicks! with optimized, uv ready and even texture ready asset! what a very time saving.

The "refine, don't rebuild" concept feels like the right angle. How much time are artists actually saving?

I like that this focuses on flexibility and editable outputs. That's honestly the kind of AI support that makes sense for real production work.

how steep is the learning curve for experienced artists?

Hey team, congrats on the launch. If I don't like the generated topology on a certain area of the mesh, can I re-run for just that section?

Usually in AAA production, predictability matters more than speed. Can predictability be something to be expected with this?

Ai in 3d has felt more like a novelty than a tool for real production so far, Not sure how this can be different in day to day use

🚀🚀🚀 So cool! Can I use this to texture models as well?

AI-assisted design sounds amazing! "Claude code/codex for designers" type of thing

Similar to vibecoding can someone have no experience and be vibe-3d-designing with this?

This looks awesome!

How much control do we have over the retopo generation?
Any way we can direct where certain loops should be or go, or any plans on implementing it?

You should keep a pricing page 😁

Looks great, does it fit well into the SoloDev workflow aswell?

Wow! this looks promising, does is create good topology for animation? , keeping loops clean?

🚀

I've seen AI optimize geometry poorly for mobile. Can this target specific platform constraints?

I'm wondering how stable the outputs are across iterations. Is there a lot of randomness?

AI in 3D is exciting, but expectations are often unrealistic. What's something this tool intentionally does not try to do?

wow, excited to use this in my projects. I especially like how you can iterate and change the input image, most online AI generators don't have this so its nice seeing everything in one package.

Can you store pipeline presets for different projects?

Heya! Aaron here from MonoDesk - we built the app we wished existed when we were freelancing.

One workspace for projects, clients, planning & a helpful AI assistant.

Really keen to hear what you think!

Hey, I’m Lou, part of the MonoDesk team and a freelancer myself.

When we say we built MonoDesk for creatives tired of juggling briefs,
feedback, tasks, and admin across a dozen tools, I’m one of them.

If you’re a creative pro, we'd love your take.
What’s missing? What should we build next?

It's free during beta. Jump in and tell us what you think.

Love it! Makes me wish I was freelancing at the moment, as this looks super helpful :) I'll be sharing with my freelance friends this week to get their take. Nice website btw!

This is an impressive app, congratulations! The quality of the AI-generated content genuinely exceeded my expectations, other apps I’ve tried haven’t come close. I’ll definitely be integrating MonoDesk into my project moving forward. I just found a bug on the calendar, it isn’t working properly, it may be due to a bug.

Love, love love this! In my previous role as Cretive Director, I had to juggle so many tools - Pixieset for client submissions, Milanote for client feedback, Box… the list goes on. For admin stuff, does this support secure client document sign offs like contracts, project completion paperwork, NDAs, etc? Congrats on this! I’m all for unification

Great product! I wish I had this when I was a freelance designer. Way better than having to pay for and use different tools. Congrats on the launch!

Context-aware tone switching per app is a genuinely hard problem to get right. Most voice tools are either totally generic or require you to babysit the profiles manually, so if Ghost Morph works well in practice that's a real differentiator. Congrats on the launch!

It's been very surprising how quickly we're able to ship code once checks are in place. For the last few years there has been (rightfully so) a dominating conversation on scaling the writing of code, but with that problem feeling largely solved, it's become, for us, a question of how to make sure that all code meets our standards. Checks in Mission Control have made this pretty easy to do, especially given we can just ask Claude to write checks for us. With checks doing the heavy lifting, most PRs probably don't take more than a few seconds of human review

let's go, amazing product in the wild

This is exactly the kind of tooling that's been missing in the AI-assisted development workflow. We use 4 different AI providers at TubeSpark (OpenAI, Anthropic, Groq, Gemini) for content generation, and the quality variance between models is real — what passes review from one provider often needs manual fixes from another.

The idea of encoding quality standards as source-controlled markdown files that run on every PR is brilliant. Right now we rely on manual code review to catch AI-generated inconsistencies, which doesn't scale.

Curious about the feedback loop — when Mission Control flags an issue, does the developer fix it manually or can it suggest/apply fixes automatically?

My favorite check on the Continue team is our "Next.js best practices" based on this skill. It runs on every PR and catches something subtle almost every time!

Even though we have this same skill for the agent to use locally, agents still make mistakes as context windows grow, so running it in CI gives us assurance that we aren't letting slop make it into production.

The markdown-file-as-check approach is smart because it keeps standards reviewable and diffs visible, same as any other code change. One thing I'd want to know: how do you handle checks that are too broad and start flagging everything? That noise problem killed a couple internal lint rules on our team before we got the scope right.

Love the idea of encoding standards as markdown checks. Curious — do teams typically start with a few core checks and expand, or do you recommend defining all standards upfront?

This solves a real problem. With so many teams pasting data into ChatGPT and other AI tools without thinking, having a safety layer makes total sense. Does it work with self-hosted LLMs too, or just cloud-based ones?

Every AI product built in the last 3 years was optimized for the same interaction: one human, one AI, one thread. But that's not how real conversations work.

Team standups. Classroom discussions. Conference Q&As. Game NPC interactions. These are multi-party, fluid, and full of interruptions.

We've been designing AI for the exception, not the rule.

DialogLab is Google Research's open-source framework to author, simulate, and test dynamic group conversations involving both humans and AI agents.

Not just "prompt a model and hope for the best." An actual design environment for multi-party dialogue.

Here's what it does differently 👇

• Visual scene builder - drag-and-drop canvas to set up participants, roles, subgroups, and shared content

• Snippet-based flow control - break conversations into phases like opening, debate, and consensus, each with its own turn-taking and interruption rules

• Human-in-the-loop simulation - an audit panel surfaces AI response suggestions during testing; you accept, edit, or dismiss in real time

• Verification dashboard - visualizes turn distribution and sentiment flow so you're not reading 200 lines of raw transcript

Tested with 14 domain experts across game design, education, and social science research. Human control mode rated consistently higher on engagement, realism, and effectiveness vs fully autonomous agents.

Who should care right now 🎯

• Game devs building NPC dialogue systems that feel alive

• Educators creating AI-simulated practice environments like mock interviews or debate prep

• Researchers studying group dynamics without coordinating 10 humans in a lab

• Product teams prototyping AI experiences beyond the single chatbot window

We're entering a world where AI agents talk to each other and to humans simultaneously.

DialogLab is an early, honest attempt to build it. It's a research prototype, not a polished SaaS tool. The GitHub is open, the paper is published. Worth an afternoon if you're building anything in the multi-agent or conversational AI space.

GitHub: https://github.com/ecruhue/DialogLab

If you could simulate any group conversation before shipping it, what would you test first? Share in the comments! :)

This is interesting, I would like to see multi party chats that include humans and their AI assistants.

Really cool research — the idea of orchestrating transitions between scripted and improvised conversations is fascinating. Can you test how different agent personas handle conflicting instructions in a group setting?

congrats @sundar_pichai for the launch

Let's gooooooo @Upsolve AI team! This is awesome. Great work

No SQL, no pivot tables — that's exactly the workflow most people actually need. How large of a CSV can it handle before performance starts to drop?

Cool! Can I use it with my already working openclaw agents?

The "mission control" framing is exactly right for this problem. Once you move past single-agent experiments into multi-agent workflows, the operational overhead becomes the real bottleneck — not the agent logic itself. Most teams are cobbling together logs and dashboards from separate tools, which defeats the purpose of automation.

>

> How does AgentCenter handle runs where an agent spawns sub-agents mid-workflow — does the visibility cascade down to the child runs automatically?

Congrats on the launch!

This resonates with me, Raymond. I've been deep in the knowledge management space and the thing that kills most of these tools is they all eventually try to become Notion. Then nobody knows what the product actually does anymore. The voice-to-docs angle is what caught my eye though. Getting a clean transcript is the easy part now. The hard part? Figuring out what from a 20-minute ramble should be a heading vs. a bullet vs. just cut entirely. That structural judgment is where the real value is.

How are you handling that? Does the AI auto-structure the doc or does the user get to shape it first?

Nice! All the models requires a knowledge base to be actually useful and I've found many founders dealing with fetch and organize all they already know about their own business. It could change that game! Congrats!!

This solves a very real problem that most team chat tools completely ignore.

WhatsApp groups with international teams are a mess, half the conversation gets lost because people default to their comfort language and others just stop engaging. Automatic per-user translation without any manual action is the right UX call.

29 languages on launch is a strong foundation. Would love to see Hinglish or regional Indian language support down the road, massive multilingual user base there.

Congrats on the launch!

Nice, do you plan to integrate this to other tools that are widely used? I understand that the focus lays on internal teams but I think this is something useful for everyone and sooner or later we will have such options on the majority of the platforms.

One more real-world story we’d love to share:

A hotel cleaning company with staff from multiple Southeast Asian countries started using Shavely for daily coordination. Before, instructions often had to be repeated in different languages, and small misunderstandings would slow things down.

After introducing automatic per-user translation, they reported fewer clarification messages and smoother task handoffs between shifts. Seeing Shavely work in a fast-paced, multilingual environment like that reinforced why we’re building this.

It’s a small example, but it shows how invisible language friction can impact real operations.

Where else do you think this kind of friction exists?

The corroborated / contested / unclear distinction is the feature that matters most here. Most news aggregators just show you more sources, which doesn't solve the problem. Knowing which claims have cross-source agreement versus which ones are disputed versus which ones nobody actually knows yet is a fundamentally different reading experience.

The closest comparison is Ground News, which does a solid job of showing left/center/right framing differences. But it still sends you to individual articles rather than synthesizing them. AllSides does the same, side by side rather than structured. The Bias is trying to solve a harder problem: not just showing you where sources differ but telling you what they actually agree and disagree on. That's a much more useful output if you pull it off consistently.

To answer your question on what I'd change first: I'd want a way to track how a story's corroboration status evolves over time. Something that looks contested on day one often gets clarified within 48 hours, and being able to see that arc would add a lot of value. Congrats on the launch! 📰

That’s awesome! How many sources are you tracking? We’re doing similar classification at MediaThrive. If you want, ping me and we can exchange ideas. 🙂

I get nervous when skills turn into a grab-bag of styles, agents start feeling random. How are you scoring Skills Refiner's benchmark, by running a small task suite or just grading the refactored text, including translate/refine cases? That's what makes a score like this trustworthy.