Connect with the Product Hunt community

Thanks for following Comcent! We're starting a series of short tutorials, one task each.

The first one is the whole setup: sign up, create your organization, connect the Twilio number you already have, choose who its calls ring, and answer a real call in the browser.

We're taking part in the GPT-6 Astra Challenge with OpenAI right now, which got me thinking about this.

New models drop constantly. Some people test everything immediately, others stick with what works until there's a real reason to switch.

I'm not talking about the obviously ridiculous AI hallucinations. I'm talking about those answers that looked completely believable until someone actually checked the numbers.

Maybe it was a calculation, a misleading chart, or a metric that meant something entirely different from what the AI assumed. What happened, and how did you catch it?

Big week for Sponkly: right after NVIDIA Inception, we have been accepted into the AssemblyAI Startup Program.

Every Sponkly feature starts with speech recognition. Before a call, a meeting or a face-to-face conversation can be translated, the app has to understand exactly what was said, in any accent and in real time. With AssemblyAI's support we will make Sponkly's live voice translation faster and more accurate, from live calls to meeting transcripts and video subtitles.

Thank you, AssemblyAI, and thank you to everyone here who supported our launch last week

Which speech-to-text do you use in your products?

Some founders launch their product once and then focus entirely on building and growing it. Other approach is to treat every major update as another opportunity to launch. A new MCP server, a Business or Agency plan, or even a new integration can all be reasons to put the product in front of people again.

I can see the logic behind both approaches. Launching once lets you focus on the product, but launching regularly gives you more chances to reach new audiences.

With over 60% of searches ending without a click, search performance reporting needs to account for what users see directly in search results.

Think about someone researching CRM software. They might ask ChatGPT for vendor recommendations, compare pricing, and evaluate features before visiting a single website. The brands appearing throughout those conversations have an opportunity to influence the buyer's shortlist.

To be honest, my unfair advantages are gone with the new AI models: Getting ideas and building them. And this scared me, for a moment, I thought that finally everyone around me would become a builder.

I went to all my friends who told me before: "I have an idea, can you help me build it?" and told them that they can do it now, with very little effort.

To demo Explain s power we present to you Reddit TV (redditwatch.tv) which automagically creates short videos of any Reddit post, summarising the discussion. The homepage pulls in Reddit's most popular posts which refresh every 15 minutes and you can also use the search field to tune into the most popular posts from any subreddit (subreddit feeds refresh once per day).  These explainer videos are generated on-the-fly within seconds the first time someone clicks a post and we cache them for instant playback for the next viewer  .

Apart from featuring in the Product Hunt newsletter and some awareness.
What measurable (long-term) outcomes did you get?

E.g. apart from increased traffic it could be:
XY long-term users
 collaboration claimed
investors who gave you XY amount of money

Share your highlights that stemmed from the ranking.

YC applications close in 11 days...

And every day I find one more thing I want to fix before submitting.

I tried switching to voice dictation for writing prompts last week because typing out long logic explanations gets exhausting. The funny thing is, when I speak out loud I end up giving the AI way more context about edge cases than I ever would when typing in a hurry.

The only catch is trying not to look completely insane while talking to your laptop in a quiet room. Has anyone made voice workflows a regular part of their day or is it just a fun gimmick?

Gabriel and I both open the repo in the morning and kind of... hope for the best hehe

Sometimes we're on the same file without knowing it.

That's literally the problem we're trying to fix with Murmell, a multiplayer workspace for coding with AI.

But the human side is still messy :)

I started out tracking what new launches on Product Hunt, Hacker News and PeerPush were built with. I had developers in mind when I built it.

Then I started finding websites beyond the launch boards. Shops, salons, trades, cafes, agencies. Businesses putting their sites live without posting about it anywhere.

For 3 months I shared weekly progress updates publicly. Numbers, learnings, screenshots, full transparency. Then I noticed something: I was choosing what to build based on what would look good in the update not what users needed. "Impressive" features got priority over "useful" ones and a dashboard screenshot performed better than a boring fix that actually mattered.

I stopped posting metrics publicly after that. The transparency didn't keep me honest it made me perform.

So for those who build in public: have you caught yourself optimizing for the audience instead of the user? And the harder question: if you removed the audience tomorrow would you still build the same things?

I came across this question in the YC application:

Please tell us about the time you most successfully hacked some (non-computer) system to your advantage.

Every new session...

I end up explaining the same project again hehe

Me and Raf have been working together for more than 12 years !

He was the founder of the animation studio where I joined as an intern.

Hey Product Hunt

A few weeks ago, you made Talanoa the #1 product of the day. Thank you again. A lot of you sent feedback, and the most common one wasn't about a missing feature: it was about the app feeling slow during big syncs, or an account that sometimes stopped catching up.

I m about to let visitors try my app before they sign up. The easy way is anonymous sign-ins: every visitor gets a real user ID, so limits, row-level security, and the saved result all keep working, and signing up later converts that same user.

Before writing any of it, I had Claude Code read the code, and I separately queried the live database. The two disagreed.

1. The code review said the results table requires a user id. The live column is nullable. The migration files said one thing and the database did another.

2. My users table requires an email, and a trigger copies the email from the auth table into it every time a user is created. An anonymous user has no email. I haven t tried it, but a trigger that fails inside sign-up usually takes the whole sign-up down with it. On launch day that would have looked like a vague database error , and I would have blamed the auth setting.

The fix is a new migration that lets email be empty, then a check of every query and job that assumes an email exists (the weekly email, for one). It isn t shipped yet.

Unit tests are great for making sure buttons click and forms submit but they are completely blind to a modal moving 5 pixels to the left or a background color bleeding into a header.

We started running automated headless browser screenshots before and after restyling PRs to catch silent layout breaks.

Dani here Today's other posts have been about the clever stuff: proactive messages, metadata, the JavaScript API. This one is about the opposite. Most people only want a chat box on their website that talks to their n8n workflow and looks like it belongs there.

For most setups, that's all it takes:

1. Paste your n8n webhook URL

2. Set your colours, avatar and welcome message, and watch the preview update as you go

3. Add it to your site with one snippet, or with our plugin on WordPress, Wix or Webflow. No code needed.

That's the whole setup: no CSS to fight with and no frontend build. If you change something later, Managed widgets update on your live site as soon as you save, so you never paste the code again.

Hi everyone

One of our power users has been running his go-to-market with a team of Grok Bot agents, all plugged into the same AI Context Flow memory. We loved it so much that we rebuilt it, tested it, and wrote it up.

The setup in short:
One memory bucket per agent, shared with your Claude and ChatGPT chats
One Todoist board where every agent tracks its work
A "Human" label that makes agents stop and wait before anything goes out

Full step-by-step guide with prompts: https://plurality.network/blogs/...

In August, Ross and I bought N8N Chat UI from its original creator. He'd built something brilliant: the widget, streaming, voice, file uploads and the JavaScript API were all his work. Our job was to take it further and get it in front of more people.

What we've learned in the 12 weeks since:

You inherit decisions you didn't make. Pricing, free tiers, naming. Some of it we kept, and some we're still working out. It's humbling to defend a choice you'd never have made yourself, then find out why it was made.

Plugins open more doors than features do. Our first big step was meeting people where they already are: WordPress, Wix, Webflow and an n8n community node. Most people don't want an embed snippet. They want a button in the tool they already use.

Most websites lose almost everyone who visits. People land, have a question or a doubt, don't find the answer quickly, and leave. A chat widget is the obvious fix, but one that sits in the corner waiting to be clicked usually doesn't do much.
What we've seen work better, working with agencies and shop owners, is treating chat as part of the sales process, not a support add-on:

1. Show up at the right moment. Open the widget when someone's been on the checkout page for a couple of minutes, or sat on a product page without adding anything to the cart. That's usually when a question is stopping the sale (delivery, returns, sizing).

2. Know why they're there. Pass the page, the product or the cart to your n8n workflow with every message, so the agent can answer the actual question and doesn't open with "How can I help?".

3. Ask different questions on different pages. Pricing-page prompts and returns-page prompts shouldn't be the same.

4. Don't lose the people who aren't ready to buy. Capture them as leads into your CRM (HubSpot, Pipedrive, Mailchimp or whatever you use) with consent, so you can follow up.

5. Book the calls. Let the chat book the meeting with Calendly or Cal.com, so a warm visitor doesn't have to wait for an email.

All of that can be wired up with n8n and a bit of JavaScript, and we've put ready-made workflows for the CRM, booking and order-lookup parts at n8nchatui.com/recipes.
What I'd like to know: which of these has actually moved the needle for you or your clients? And what have you tried that didn't work? I'd love real numbers if anyone's willing to share.

Purchasing a product has many different angles of decisions, from understanding the brand, figuring out if the product is worth it, and ultimately deciding if the price is right.
A lot of it comes down to the relationship the prospect has with the site, which is where N8N Chat UI comes into play - Allowing you to make modules as dynamic as you want, or as tame as you want.
We've just launched on ProductHunt, link is here https://www.producthunt.com/prod...