Connect with the Product Hunt community

We spent weeks tweaking our pricing copy and redesigning tables, but nothing moved the needle. Then we decided to simply drop our lowest plan, and surprisingly, conversions on the higher tiers jumped right away.

It turns out having a super cheap option was just giving buyers an easy way to avoid committing to the plan they actually needed.

Have you ever tried removing a plan instead of adding new features or tiers to fix conversions?

So, I'm about to launch my second product, which in my head I felt was a unique offering. As part of the work to try and get community engagement pre-launch I started exploring certain Red-colour forums to get feedback and understand user engagement, but what actually happened was a complete pile-on on the apparent unoriginal-ness of my product, etc.

I get that everyone has their own opinion, and I do not generally engage with trolls, but sometimes it's hard to ignore it, especially when you're 1-2 weeks out from your product launch, and then doubt kicks in.

I have a two-person SaaS business, and for months I responded immediately to all support emails like there was a fire. I had to drop what I was doing and respond in the hour, saying sorry even when the bug was small. Then I was sick for a week and couldn't respond like that.

Retention did not fall. It increased slightly. My take: immediate responses were teaching the users to ask me instead of looking at the two-line documentation I had already written; the stress of "somebody might need me right now" was taking up my valuable fix time.

I open Product Hunt regularly and every product looks incredible.

Perfect landing page. Perfect Loom video. Perfect story "Built in 7 days with AI, 1000 users in 48 hours."

I try 5-6 a day. And honestly, most of them are impressive for the first 5 minutes. The onboarding is slick, the UI is clean, the demo makes you go "wow."

Then I never open them again.

I've been a full-stack dev for 4+ years (React, Next.js, Node, NestJS), so I went into vibe coding thinking it'd just be "coding but faster." It wasn't quite that.

I run an independent practice building for small agencies and fashion/retail brands, and I started leaning on AI tools (Cursor, Claude Code) to move faster on client work. The first 60-70% was genuinely magical scaffolding, boilerplate, even decent architecture suggestions came together way quicker than writing it by hand.

Where it got messy was the last stretch auth edge cases, a deployment config the AI kept "fixing" back to broken, and a couple of times it confidently rewrote logic I hadn't asked it to touch. Being a dev already, I could catch it and course-correct, but I kept thinking about how differently this would've gone for someone non-technical.

Curious what this community's experience has been:

Not naming names for my own product here, someone else's tool burned me last week. 14 questions before I even got to see a real screen. Company size, use case, "how did you hear about us", team size again but phrased differently. By question 9, I was just clicking random answers to get to the app.

Never even got there properly. Closed the tab, opened it again two days later out of guilt, got hit with a slightly different set of questions because apparently my first run didn't save. That was it for me.

The customer wrote to ask why our plan down had not included the one thing she uses every day, and our plan up included six things that she would never use. She was not complaining. She was confused by whom we designed our plans for.

I had invested countless hours to make sure our pricing tiers were priced correctly, but zero effort into seeing if our pricing tiers told good stories about our different types of customers.

68% open rate on day one followed by a cliff on day three that no tweaks to the subject line have managed to change. I used to equate open rate with interest level. Now I realize that all it means is that our subject lines were good and that we sucked at onboarding.

Anyone who has ever had to deal with a problem like this, was the solution something to do with the email, the timing, or something about the product itself that emails couldn't solve?

Two runs of the same request on the same commit, two diffs that do not match, and most of us file it under luck. A leaderboard published this month put a count on that luck, because it prints every trial: eight models, ten tasks, eight runs of each. We tallied the eighty cells. Six passed on every run, thirty-three failed on every run, and forty-one, more than half, landed somewhere between.

Under it are two sources of variation, and they stack. The model hands back a probability for every token and a sampler draws one; temperature reshapes the distribution, and at zero the draw is gone but the runs still differ. Anthropic's API reference says so in one sentence, and its models released after Opus 4.6 will no longer take a temperature value. The rest comes from the serving stack. A research post from Thinking Machines ran one prompt a thousand times at temperature zero and got eighty distinct completions, traced to floating-point sums added in a different order as the batch around your request changes. An agent loop turns that into a different first file to read, and from there the two runs walk different paths.

The figure people quote from a leaderboard is pass@1. On this grid the top row sits at 38.8% for k=1 and about 70% for k=8; a row further down goes from 28.8% to 90% and never solved any task on all eight runs. The k is the part that goes unsaid.

How many runs do you give the same task before you believe the result, and does the count get written down anywhere?

Over the past 24 hours, my agent IDE session has been repeatedly crashing right in the middle of writing code.

The generation will start, run for a few seconds, and then the workspace completely drops or terminates the agent process, forcing a full IDE restart just to continue.

I work on a B2B AI product, and one problem I keep running into is deciding how much of the product s underlying concept users actually need to understand.

Some things can be simplified through UI, but simplifying too much can also hide information users need to make the right decision.

I ve started thinking that the goal might not always be make everything intuitive. Sometimes the product may need to teach the user something first.

For people building complex B2B products, how do you decide what should be simplified in the UI and what users actually need to learn?

A while ago, I started a quick note titled "apps to check out." It has now grown into a massive list of around sixty items, and I have realistically tested maybe four of them.

It starts with good intentions you see an interesting launch, save the link so you do not lose focus during the day, and promise yourself you will test it over the weekend. Then the list just sits there accumulating dust while new tools keep getting added to the pile. At a certain point, it stops being a helpful reference and starts feeling like digital clutter you quietly ignore.

Is saving links like this an actual discovery process for you, or is it just another form of hoarding? If you actually clear out your list, how do you decide what is worth keeping?

Over the past 24 hours, my agent IDE session has been repeatedly crashing right in the middle of writing code.

The generation will start, run for a few seconds, and then the workspace completely drops or terminates the agent process, forcing a full IDE restart just to continue.

last month a user hopped on a call and went on for 20 minutes about how we saved their team. named people. named the old process. said they'd write something that night.
i sent a thank you. waited. nothing.
followed up two days later. "yeah yeah i'll do it this week."
that was 3 weeks ago. i still have the call recording. i still have zero words i can put on the site.
this keeps happening. the people who are happiest are also the busiest. the people who write fast usually send "great product, highly recommend" which is almost useless.
so now i'm stuck between sounding desperate and having a homepage that looks like nobody uses the thing.
what actually works for you when you ask? not the theory. the message you send. or do you just stop asking and live with it.

I spent twenty minutes yesterday clearing out random extensions and temporary bookmarks I had added over the past month.

I always tell myself I ll test a tool for five minutes and remove it if I don't like it. Instead those trial extensions just sit in my top bar consuming memory and cluttering my tabs.

It made me realize I need a better system for trying out new launches without ruining my primary browser setup.

What s your current routine when you want to try a new app do you use a temporary guest window, a distinct browser, or just keep it simple in one window?

I have been thinking about a failure mode that permissions alone don t really solve.

An agent might correctly decide:

Refund this customer $25.

I've gone back and forth between these two more times than I should probably admit.

Coda always pulls me in because I can make tables, buttons, little automations, etc. It's fun until I've somehow turned a simple notes system into a project of its own.

Then I go back to Notion because I just want to write things down and find them later.

And then, inevitably, I miss something Coda did better.

I used to think a thorough pricing page was essential for building trust. Our had multiple tiers a 14-row breakdown, and answers to every question imaginable.

Out of frustration I cut all the noise and left just one straightforward offer on the page. The result? Our sales went up significantly.

It felt like users were previously getting overwhelmed by the choices and leaving to "think about it."

Every week a dozens of new tools drop on Product Hunt, promising to "10x your workflow."

I used to test every new launch out of FOMO until I realized half my workday was spent configuring apps instead of building.

How do you evaluate new launches without wrecking your focus?

Finding good tools here is easy, but sharing them with people in real life is hit or miss.

My non-tech friends don't really care about new productivity apps and my teammates usually prefer sticking to what they already use.

I m curious if you guys post your finds on social media share them at work or just keep a private list for yourself?

I posted my repo and got downvoted instantly. People called it pointless.

Six weeks later: 400+ stars, 5 pull requests from strangers, and a sponsor.

No viral thread, none of it came from aggressive self-promotion.

My honest take, online comments are heavily biased toward critics who like to tear things down fast.

I spent most of my budget building the product and have almost nothing left for marketing.

For anyone who has been in this situation, how did you get your first users without spending money?

I keep finding new AI tools on Product Hunt that look useful but after trying a few I usually end up going back to the same 2 3 tools I already use. I m curious if other people have the same experience. Are there any AI tools you discovered here that actually became part of your regular workflow and what do you use them for?

We read both vendors' documentation and the recent work on looped transformers with three questions: what does a model's reasoning cost, what can you steer, and what can anyone read afterwards. The answers sort into three places.

On the page, as visible text. You pay the output rate, you steer it with the prompt, and you can read it, though studies show these visible traces are often unreliable.

Off the page, as hidden tokens. The example in OpenAI's own guide shows 1,186 output tokens charged, 1,024 of them reasoning; you pay for every one and the text stays with the vendor. Anthropic's docs show the same: you're billed for full thinking tokens whether shown or not. The rate is the same, the steering is an effort knob in place of a prompt, and what you can read is at most a summary. The exact count is in the usage object, and on the newer models earlier turns' hidden reasoning stays in context and is billed again as input.

Inside the network, as extra passes through the same layers, which is what recurrent depth means. Compute rises per token, the token count does not move, and there is no text for anyone to read, the vendor included.

Imagine an infrastructure agent detects a failing service.

It decides to restart instance A.