PH热榜 | 2026-03-20

Happy launch team! Quick question: How do you handle context and prioritization when aggregating feedback from so many different sources? For example, how do you distinguish between loud but low-impact requests and signals that actually represent broader customer demand, and how reliable is the deduplication when similar feedback is phrased differently across channels?

Congrats on the launch! But how is it different from, say, ProductBoard, Canny, airfocus, and the likes?

The "closing the loop" part is what I care about most here. We've tried a couple feedback tools before and the collection part is usually fine, but actually telling users "hey we shipped the thing you asked for" always falls through the cracks.

$24/mo flat is solid too. Most tools in this space charge per seat which gets painful fast when you want the whole team to have access.

How does the AI handle feedback that's more of a rant vs an actual feature request though? That's always been the tricky part for us.

Congrats on the launch! How does ProductBridge handle conflicting signals? (example: when a feature is largely recommended by free users but paying customers never mention it). Does AI score accounts by revenue impact, or is prioritization purely vote-based?

One of my biggest challenges with customer feedback is trying to filter out which ones were real feedback and which ones were from bots/fake. Are there ways that ProductBridge help with this?

We collect client feedback across several channels at once — and deduplication is what interests me most. The same request often arrives three times, worded differently, and it's hard to tell if it's one problem or three. How does ProductBridge decide two pieces of feedback actually belong together?

Congrats on the launch! @hareesh_vemasani @rohithreddy
Honestly, this is something most teams just deal with instead of solving.
Feedback keeps coming in, but it rarely turns into clear product decisions.

Really like how you’ve made it more structured and usable.

Curious, what kind of feedback patterns surprised you the most so far?

Great work @hareesh_vemasani 👌really love how you’ve tackled such a real and messy problem. As someone working in growth and SEO, I’ve seen how scattered feedback across channels often leads to weak prioritization and missed insights. What stands out here is the full loop from collecting feedback to actually closing it with users through changelogs and notifications, that’s where real trust and retention are built. Also the flat pricing is a smart move in a space crowded with seat based models. Curious to see how it performs at scale, but this looks genuinely useful for product teams...

really like this because feedback usually ends up scattered everywhere and teams lose a lot of time just trying to piece it together. the closed-loop part stood out to me since users rarely know what happened after sharing feedback. which source is giving you the most valuable insights so far: support tickets, reviews, or slack conversations?

Hey Hareesh, congrats on the launch! Interesting tool solving a real problem.

Question: Feedback online is highly skewed and biased (as it takes particular types of personas to post, with no proper way to 'control' via experimental design). Is a product roadmap built on online feedback the best path forward for builders?

I like the direction of the product. As a PM, I always use Dovetail to aggregate user reviews from different channels. But the "feature brainstorming" and "prioritization" part I do with my co-pilot, but it's a second surface (outside Dovetail) so I have to switch between them. Kudos for solving these problems as I see it and both important parts can be done with the same tool.

Bringing feedback from multiple channels into one place and actually turning it into roadmap decisions sounds really useful. I like the focus on closing the loop with users after features ship. How does ProductBridge detect and merge duplicate feedback across different sources without losing important context?

Congrats on the launch!

Quick one: Most feedback tools assume users know what they want and can express it clearly, but that's rarely true for non-technical users. What does your product do when the feedback is vague, emotional, or indirect? How do you turn 'this is confusing' into an actionable roadmap item?

Great product which solves a real pain point! Congratulations on your launch!

How does the deduplication work when users describe the same issue in completely different words? Congrats on the launch!

This is a very interesting idea. In our business, we receive a lot of feedback from multiple channels that never really gets processed as data as such, so this idea could actually be relevant, but I have a couple of doubts that came to mind:

The actionable insights sound great, but how does the app process contradictory feedback from clients to decide which side to lean to? Is there a process of prioritizing certain types of feedback over others? It would be super interesting to get a bit more info about this.

Anyways, congratulations for the launch!

Feedbacks are like goldmine, just curious how will ProductBridge will filter out feedbacks from tons of other contents. Congratulations on the launch. This looks a great product that will genuinely help businesses.

Congratulations on the launch 🎉 🎉

a great product @hareesh_vemasani and team. much needed today!

Excellent looking app for building a customer feedback loop & offering transparency!

Congrats on the launch!

Congratulations

Feedback is everywhere — support tickets, Slack, emails, user calls, but turning it into clear, actionable insights is still a big challenge for most teams.

ProductBridge seems to be tackling exactly that gap. If done well, this could really help teams prioritize better and build what users actually need.

Curious, how does ProductBridge handle deduplicating and prioritizing feedback across different sources?

Congrats on the launch and excited to see how this evolves 🚀

Congrats on the launch! The closed-loop piece, auto-notifying users when their feature ships, is something most feedback tools completely ignore. Smart move on flat pricing too. Rooting for you guys

Hi Product Hunt! 👋

I’m Oscar, co-founder of AdsTurbo AI.

After years in video production and digital marketing, we kept seeing the same gap: AI made ads faster to produce, but most outputs still felt synthetic—flat expressions, stiff delivery, and that “AI vibe” that hurts trust.

So we built AdsTurbo AI around AI Talking Actors trained on real creators—real expressions, gestures, speaking rhythm, and on-camera presence—so the result feels closer to real UGC and real ad performance.

What you can do with AdsTurbo today:

• Create UGC-style video ads with more natural, believable actors

• Produce more variations and test creatives faster

• Scale production without losing the human feel that helps people stop scrolling

We originally built this for our own ad workflows, and after other marketers and brands asked to use it, we decided to launch publicly.

Would love your feedback: What’s the hardest part of making high-performing video ads right now—speed, cost, performance, or “looking real”?

I guess it's obvious, but I wanted to still ask that since you trained it on real creators, and I hope that it was ethically done with their permissions, etc., yes?

Congrats on the launch! What’s your key differentiation from Heygen please?

Congrats on the launch!!

Do you have an API? Looking forward to integrating it into my app!

Congratulations. If there is a demo video for the agent, would love to see.

Congrats on the launch!!

Where are the top performing ads sourced from? Is it from the ad library of our competitors or the platform curates by itself based on the major industries?

AdsTurbo's AI Talking Actors are aimed at the right problem. For agency workflows, the bigger hurdle is approval risk, since clients notice fast when one take feels even a little off-brand or uncanny. A shot-level review flow would matter a lot.

This looks sick, I have been in the ads industry for a while now (9 years) and have been using arcads, bandoffads and also n8n automations to now start using sora and veo3, so this actually looks really sick. Great work man! Would definitely give this a go

If you don't mind me asking how much time did you spend on leveraging real creator to train this product?

Interesting approach. Improving realism in AI-generated video ads is a big gap, especially when trust depends on how natural the content feels. The focus on expressions and delivery makes a lot of sense for performance. How customizable are the AI actors in AdsTurbo, can users control tone, pacing, and gestures for different ad styles?

How do you handle different brand voices across campaigns? Like if a client wants a more serious tone vs a playful one, can the AI actors adapt to that? Congrats on the launch!

Are there ways to customize the type of tone or specific messaging that's used within the ad?

The facial micro-expressions are unsettlingly good - I kept waiting for the uncanny valley to kick in but it never did. Have you tested what happens when you feed it a script with deliberate emotional contradictions (like someone smiling while delivering tragic news)? I'd love to see if the AI can navigate those subtle human inconsistencies that separate convincing from just technically correct.

Composer 2 by @Cursor is a frontier-level coding model designed to solve complex, long-horizon programming tasks with high efficiency and strong benchmark performance.

It tackles the problem of limited coding accuracy and high costs in AI dev tools by combining improved intelligence with optimized pricing.

What makes it different is its continued pretraining + reinforcement learning on multi-step coding tasks, enabling it to handle hundreds of actions with better results across benchmarks like Terminal-Bench and SWE-bench Multilingual.

Key highlights:

• Strong coding performance (61.7 on Terminal-Bench 2.0)

• More cost-efficient ($0.50/M input, $2.50/M output)

• Fast variant with same intelligence but quicker responses

• Built for real-world, long-horizon dev workflows

Great for developers, teams, and builders working on complex codebases, automation, and AI-assisted programming. If you're building with AI, this is worth checking out!

P.S. Here's an interesting comparison between Composer 2 vs Opus 4.6 vs GPT 5.4 (unscientific). Composer 2 is 10× cheaper than Opus 4.6 and supposed to rival it.

P.P.S. I hunt the latest and greatest launches in tech, SaaS and AI, follow to be notified → @rohanrecommends

While Windsurf confuses their pricing model, Cursor keeps trucking with their own tech. Inspiring stuff.

Is this a fine-tuned Kimi 2.5 model?

That's fascinating. Cursor isn't just an app or AI model company; it's both. I think this dual identity is the biggest differentiator Cursor has among hundreds of coding agents and editors.

Just gave it a spin. Loving the speed and the cost efficiency, even if it still needs a lot of hand-holding. That's great for planning though, and to carry out simple tasks :) It will totally become my daily driver

The pricing is what gets me. $0.50/M input is wild for a model that's beating Opus 4.6 on coding benchmarks. Been burning through tokens on long refactors and this could cut my bill in half.

Curious how it handles multi-file edits across a full monorepo though. That's where I've seen most coding models start to lose context and make weird decisions. The "long-horizon" claim sounds promising but I'll believe it when I see it on a real 50-file refactor.

the token efficiency angle is interesting - most coding models optimize for correctness first and leave efficiency as an afterthought. curious what the tradeoffs look like in practice. do you find it handles multi-file refactors well or is that still where longer context wins?

My early tests of Composer 2 look very promising. It feels like using Claude 4.6 Opus, but faster and more cost-efficient. I was considering switching to Zed or Windsurf before this update, but this release has kept me on Cursor(for now). That said, Cursor is still a heavy RAM consumer in my workflows, and I’d prefer a more memory-efficient IDE that offers the same level of capability.

Hi PH, cofounder of Assembly here. Thanks for the hunt @benln

We're calling this 2.0 because it includes most of the top-requested features from the past year:

1. A new Client Home — Full redesign + you can now set up different homepage variants for different types of clients.

2. App folders — our users love customizing the sidebar in their portal with embedded tools. Now you can group them into collapsible folders. My favorite use case is putting a bunch of analytics embeds into one "Analytics" folder.

3. Better project management — you could already assign tasks to team members or clients, but now you can assign tasks internally and still associate them with a client. And you can mark those tasks client-visible so that clients can follow along with progress.

4. Time-based and recurring automations — we have our own built-in automation engine and it now supports time-based triggers and recurring rules.

5. Desktop app for Mac with native notifications. Windows is coming in a week.

6. A whole lot more — check it out!

We read every message. Would love to hear what you think, good and bad!

So much hard work and hours went into this! Congrats to the whole team! Bringing more and better experiences for service businesses

I like the design. Congrats on shipping. Could we automate payment collection based on certain triggers, for example, if an invoice is not paid in the next X days, there's an auto follow-up about it. I believe simple automations like these would also be available?

Congrats on the launch team! Been using the platform for almost 3 years now :)

This is a really cool idea, I recently decided to dig deeper into my website hosting and maintenance business offerings and this could be a nice addition to the offerings for clients!

The whole team worked hard to get this over the line, wouldn't have been possible without everyone involved!

Huge congrats to the team on the launch 🚀
I know how much work went into this behind the scenes. Amazing to see it all come together.

Would one use Assembly to replace an existing tool (like Notion) or is it meant moreso to be complementary to other tools (like how one would use Jane App for billing and scheduling)?

Client-visible tasks caught my attention — in practice it's always a balance between transparency and what clients don't need to see. Can you control which tasks are client-visible and which stay internal at the individual project level?

Nice, finally the client portals are now powered by AI truly. Congrats!!

Congrats on the launch team! In the age of vibe coding it's definitely important to emphasize the humans involved in making thoughtful decisions and improvements that make our product so good.

Shoutouts to:

@adam_schwartz7 and @ana_eremina for leading the charge on the customizable client homepages

@ellie_spigelman1 @dovid_baum1 and @pinkstrings for App Library

@ari_quinones for scheduled time-based automations
@pinkstrings and @foleyatwork for the desktop app
@mysterysal @dovid_baum1 and commerce team for the contextbar, payments page

I'm definitely missing folks but this was a whole team effort! ❤️

I'm so excited to see this go live! We're really looking forward to people getting their hands on the new features we've been thinking about for so long

Good job team! Great hustle!

Nice portal.
But experience > interface.

Clients stay for results, not dashboards.

Nice update. The focus on client portals and better organization with things like app folders and task visibility feels very practical. The addition of recurring automations also sounds like a strong improvement. How flexible are the automation rules in Assembly, can users combine multiple conditions and triggers in one workflow?

Hey PH! 👋

We built GitAgent because the agent ecosystem was a mess — and three problems kept showing up everywhere:

① No single source of truth Everyone had agents, nobody had a canonical place they lived. Built in Claude Code? Stuck there. Move to CrewAI? Start over. Every tool, its own format, its own lock-in.

② Portability was half-solved Tons of solutions for sharing agent skills — but nobody was solving portability for the entire agent. Memory, identity, behavior — all stranded in whatever tool you built it in.

③ No real versioning No branching, no rollback, no diffing. You'd tweak a prompt, break something, and have no idea how to get back. Agents had no git — and they desperately needed it.

So we asked a simple question: what if the agent just lived in git? Every developer already knows how to fork, branch, PR, and tag. Git already solves versioning, collaboration, and portability for code.

We just mapped that onto agents.

GitAgent is our answer to all three — one repo, one source of truth, runs anywhere from Claude to CrewAI to OpenClaw without reformatting a single file.

Try it right now — run any agent directly from GitHub:

npx @open-gitagent/gitagent@latest run -r https://github.com/your/agent -a claude

Swap -a claude for -a openai, -a crewai, or -a openclaw — same repo, any runtime. Would love to hear what runtimes you want us to support next! 🚀

Support the project: https://github.com/open-gitagent/gitagent

Support projects built on GitAgent Standard

Clawless: https://github.com/open-gitagent/clawless

GitClaw: https://github.com/open-gitagent/gitclaw

The segregation of duties aspect is interesting. Are there ways that GitAgent helps enforce that or remind me to improve SOD when any one agent begins to take on too many steps within one process?

Interesting approach. Using the repo itself as the agent's memory makes way more sense than dumping everything into a system prompt or building separate context management.

How does it handle large monorepos though? Any filtering or scoping so the agent doesn't drown in irrelevant code? Also wondering about private repos, is there a self-hosted option?

Congrats on the launch, gonna try this on a couple of our projects.

Does the agent work well with monorepos or is it better suited for single repo setups? Congrats on the launch!

Finally - someone built the thing I've been duct-taping together with cron jobs and prayer. The idea of turning your actual repo into an agent that understands your architecture patterns feels like the natural evolution of "docs that nobody reads" to "docs that actually do something when you're not looking.

How does it handle the inevitable drift between what the agent thinks your code does versus what it actually does after six months of "temporary" fixes?

Hey Product Hunt! 👋

We're thrilled to share something we've been obsessing over. @GitAgent

Here's the backstory: We kept seeing the same problem over and over again. Teams would pour weeks into building AI agents, defining tool chains, decision logic, memory, persona, only to realize that all of that IP was trapped inside whatever framework they chose on day one.

Want to switch frameworks? Rewrite everything. Want to version-control your agent the way you version-control your code? No clean way to do it. Want to hand off your agent to another team or deploy it somewhere else? Pain.

We thought, why can't agents work like code? Store them in Git. Version them. Port them. Run them anywhere.

That's GitAgent.

It extracts the "soul" of your agent. The config, logic, tools, everything that makes it yours. It stores it as a portable, version-controlled definition. Then you deploy it to any framework with one command.

We're calling this #OwnYourAgents because we genuinely believe that if you built the agent, you should own it. Not the framework. Not the platform. You.

We'd love your feedback, your upvotes, and most importantly, your honest takes on what we can do better. This is just the beginning.

Let's make agent portability the standard. 🙌

curious how GitAgent decides which parts of the repo are relevant context for a given task. do you embed the whole codebase or is there a smarter chunking/indexing step? I ask because I built a tool that does code security analysis on vibe-coded repos and the context selection is probably the hardest part to get right.

Hi everyone!

The new Channels support for Claude Code is a massive workflow upgrade, especially if you run long or complex tasks.

Instead of being tethered to your terminal waiting for Claude to finish a heavy refactor or CI pipeline, you can now bridge the session directly to Telegram or Discord via MCP servers. It’s a true two-way street: you can push events into the running session, and Claude can send progress updates, ask for permission approvals, or deliver the final results right to your phone.

The setup process is very straightforward. You just install the official plugins, drop in your bot token, pair it via a DM code, and lock down the allowlist.

This might put a little pressure on @OpenClaw :)

this is actually a workflow gap I run into a lot - Claude Code does the work but then you want to push updates or get notified without staring at a terminal. how does the Telegram/Discord integration handle long-running tasks? like if a coding session takes 10 minutes, does it stream progress or just send a completion event?

I've been using Claude for parsing messy legacy codebases - it's surprisingly good at understanding the intent behind uncommented functions written by developers who apparently believed variable names were a luxury. The constitutional AI approach means it won't just blindly refactor everything into unreadable one-liners like some other models I've tried.

trying hard to clone Openclaw it they got acquihired by ClosedAI :p

Google is absolutely on a tear this week. After launching Stitch 2.0 yesterday, Google just upgraded AI Studio into a full-stack “vibe coding” platform. It looks like a @Claude Code competitor inside AI Studio powered by the @Google Antigravity agent.

It turns simple prompts into production-ready apps, solving the gap between quick prototypes and real, scalable products. You can go from idea to working app without leaving the environment. Feels like a shift from “AI-assisted coding” to AI actually building usable software.

What stands out:

• Builds real apps with multiplayer, auth, and databases (@Firebase)

• Auto-detects needs like login, APIs, and installs libraries for you

• Supports modern stacks like Next.js, React, Angular

• Lets you connect real-world services + manage secrets securely

• Remembers your progress across sessions

Great for:

• Indie hackers & builders

• Startup founders validating ideas fast

• Devs who want to ship faster with less setup

If this keeps improving, “prompt to production” might become the new default.

I hunt the latest and greatest launches in tech, SaaS and AI, follow to be notified → @rohanrecommends

the Antigravity + Firebase combo for full-stack vibe coding is interesting - most vibe coding tools stop at the frontend and leave you to figure out the backend yourself. how opinionated is it about the stack? like can you swap Firebase for a different backend or is it pretty locked in?

Super excited to see what people build with these updates! 🙏

love it

The auto-detect for login, APIs and library installs is a big deal. That setup phase is where most vibe coding sessions fall apart for me. You spend 20 minutes prompting and then realize the AI forgot to add a dependency.

Firebase-first makes sense for Google but I wonder if that locks you into their stack. Would be nice to see Supabase or Postgres support eventually.

How's the code quality compared to Claude Code or Cursor though? That's usually where the real differences show up.

Built for Devs is the result of productizing a service that drove the greatest results I've ever seen in my career.

I brought real developers in to screen record themselves naturally trying a client's dev tool—no scripts, no hand-holding. Just honest, unfiltered first experiences. Those recordings shaped findings reports that told founders exactly what was broken and what to fix. Red flags for what needed to be addressed first. Quick wins for low-effort opportunities. The full story of how developers experienced every stage of the product.

The results were unlike anything else I'd produced.

One client fixed a handful of friction points and hit Product Hunt #1 product of the day and week. Another completely pivoted their market using insights from 10 developer segments. The recordings alone kept roadmaps full for months.

The service was such a success that it became a platform—and then some.

Dev tool founders now get continuous journey tracking so they always know where developers slow down and disappear. Real developer evaluations matched to their exact ICP—screen recorded, unscripted, and incredibly revealing. And an AI engine that analyzes every data point including video, drafts findings reports richer than anything I could produce by hand, and recommends exactly what to fix.

Not a one-time audit. A living system that gets smarter over time. The more data it collects, the more precise the recommendations get. Founders stop guessing. They know exactly what's broken and exactly what to do about it.

Built for Devs is developer adoption intelligence for dev tools. It shows founders exactly where developers drop off, why they leave, and what to fix—continuously.

If you're a developer reading this—those screen recordings don't record themselves. Built for Devs pays developers to try dev tools. No meetings, no scripts, no hand-holding. Just your honest first experience with a product.

HOW IT WORKS
Built for Devs leverages a tracking script that captures the users entire journey—from first visit through interaction—tracking pageviews, clicks, form submissions, time spent, errors, rage clicks, and scroll behavior to measure the dev tools TTV (time to value). It uses that data to provide a fully-detailed developer journey with your touch points mapped to the right stage. It leverages everything that it learns to constantly provide recommendations for improvement.

When developers use the dev tools during an evaluation, the system processes three layers of data: a full transcript of what they say, video analysis revealing where they struggle or get confused, and interaction data captured by the tracking script that records navigation, clicks, and time spent. AI synthesizes these three layers into a findings report, identifying patterns in how developers approach problems and where their product could improve the experience.

Congrats on the launch, looks helpful!

I love the data-driven approach. Feedback to the team becomes easier to quantify and break down into work that shows actual results. Super helpful.

This looks really useful. The screen-recorded evaluations with real devs are a great idea, getting unfiltered first impressions before launch sounds way better than guessing. Curious about the time-to-value tracking too, how do you define 'conversion'? Is it something I configure or does it detect it automatically? I'm building a desktop app with Electron so also wondering if this would work for that or if it's web-only.

Do you track where developers get stuck during onboarding or is it more focused on the overall experience? Congrats on the launch!

Congrats Tessa on the launch!

So excited to see this out there!

Congrats on the launch! 🎉 The insight about user interviews being unreliable is spot on — people describe a smoothed-out version of what actually happened, not the real confusion. Screen recordings of unscripted first experiences are a completely different signal. About to launch OceanMind, an AI-powered breathwork iOS app, and the onboarding drop-off problem is exactly what keeps me up at night. Curious whether the platform works for mobile app onboarding too, or is it primarily focused on web-based dev tools and SDKs? The ICP-matched developer evaluation piece is the part I find most compelling — getting real first impressions from people who match your actual user before you’ve burned your launch day on guesswork.

Looks great Tessa! Congrats on the launch 🚀

The 6k evaluator network is the real moat here, not the tracking script. How do you keep evaluation quality consistent as you scale that pool? Marketplace supply-side quality tends to collapse once you stop hand-picking participants.

Here's an example of a developer adoption score!

Drop one script tag into your site. That's it.

bfd.js tracks how developers actually move through your docs and product—pageviews, clicks, scroll depth, time on page, rage clicks, copy events, JS errors, and form interactions. No fluff. No PII. Sensitive fields and params are automatically redacted.

Pair it with screen-recorded evaluations from ICP-matched developers in our 6k+ network, and you stop guessing what's broken. You see it.

Three products. One goal: turn drop-offs into adoption.

→ JS tracking script — measures TTV and the full developer journey
→ Screen-recorded evaluations — real developers, your exact ICP, paid to do a thorough job
→ AI recommendations engine — tells you what's broken and what to fix. Gets smarter over time.

Built for dev tool teams who are tired of shipping docs into a void.

Do you find a difference in data quality between the developers that are being paid to test a tool versus actual users of the dev tool? For instance, I know that my behaviour is different when I'm filling out a survey for a contest versus one that I genuinely am interested in.

the "how developers actually use it" angle is underserved - most UX tools optimize for non-technical users and then try to bolt on developer modes. what kind of signals do you surface that typical session recording misses? I am thinking things like rage clicks on APIs or copy-pasting error messages.

What happens with responsive? Figma designs are usually at one breakpoint. Does VisDiff only match that specific size, or does it do anything to make sure the output doesn't fall apart at other screen widths?

good job guys, can you use the product to update existing frontends ?

I’ve run into this a lot working on frontend projects. The generation part is fast, but getting things pixel perfect still takes time. Curious to see how well this performs in real-world use.

The Figma URL approach with no setup is a big deal. I've used Cursor to build UIs from Figma screenshots and the spacing is always slightly off. Does it work with any frontend framework or are there specific ones it handles best?

How does it handle responsive designs where the same component looks different across breakpoints? Congrats on the launch!

so sick! congrats on the launch!!

Does it work with my existing design system ?

How is this different from other Figma-to-code tools ?

As someone that has experienced their design come out completely different when it gets implemented as code.... I love this idea. Are there certain differences that Visdiff have trouble detecting versus ones that it is best at?

congrats on the launch ! very cool tool, and very intuitive to use!

Hi, I’d definitely use this! I have 2 questions though. How do you map elements from design to implementation under the hood? And a real friction point for lots of us is that, unless given incredibly specific instructions, AI tends to just throw a magic number or an !important to pass a visual check, which over-time adds up to crazy tech debt. Does Visdiff address this?

Congrats on the launch! When you say it integrates with existing codebases through MCP, what does that look like in practice?

Hey man, this is a really cool product. Looking forward to trying it out! :)

This solves a real annoyance. I'm using Claude Code with custom skills and every time I want to test the same skill in Cursor or another agent, it's manual file copying and adjusting paths. Having one place to manage and push skills across agents is a workflow I didn't know I needed.

Question: does it handle skill conflicts or versioning? For example, if I install the same skill from GitHub but it gets updated later – does Skills Manager detect the newer version and let me update, or is it a one-time install that I'd need to manually re-pull?

Cool Idea. A dashboard to view all active global skills is very useful. I believe all the coding agents say they should auto-detect skills in .agents/skills folder but skill detection has worked better in agent specific directory. Auto-complete for skills doesnt work when in .agents which is annoying. Did you experience the same thing? Does this solution put skills in agent specific directory?

Congrats on the launch! 🚀

I always end up losing my best system prompts and agent instructions in a messy graveyard of scattered text files. Using a centralized hub to organize and tag these AI skills would definitely speed up my workflow when switching between different development environments. I would love to know if you plan to add a local API so we can dynamically pull these structured prompts directly into our codebases.

the skills discoverability problem is real - I end up rediscovering the same prompt patterns across projects. curious how you handle versioning when a skill gets updated but some agents were built around the old behavior. do you pin versions or is it more of a live dependency?

Hey Product Hunt, I built Room Service after getting frustrated with how generic most Mac cleaners feel.

On my machine, the problem was never just “junk.” It was Xcode build data, package manager caches, Docker data, generated folders like node_modules and .venv, app leftovers, duplicates, privacy traces, shared game engine caches and logs, and a general lack of visibility into what was actually happening on disk. Most tools I tried either missed too much or reduced the whole experience to risky one click cleanup.

I wanted something that felt more transparent and more useful day to day. That is what Room Service became, a Mac cleaner built for developers, with a real home dashboard to keep disk usage and reclaimable space easy to follow, developer focused scan coverage, a review first cleanup workflow, a dedicated Performance workspace for live system metrics, Startup Item management, Applications and Leftovers cleanup, Privacy Mode, PIN protection, Touch ID support, Smart Alerts, Quarantine and Undo for safer deletes, and a shared desktop and CLI workflow.

The goal was not to make another generic cleaner. It was to build a cleanup tool that fits the way developer Macs actually fill up, and gives you enough visibility and control to trust what happens next.

Thanks for checking it out.

As a small thank you for the Product Hunt launch, I set up a 50% discount for the community, valid for the next 2 days: https://bit.ly/3NjKctQ

Hey everyone, as a small thank you for the Product Hunt launch, I set up a 50% discount for the community here: https://ardacankirko.gumroad.com/l/zfsjob/ymsfqth

Would love to hear your feedback if you end up trying it 🙏

Damn, the idea is great and the app overall is too. I scanned, found the files I wanted to delete, but the monetization style of "we'll hit you with a paywall on the last screen and block the actual action" is a killer. In the end, I just ran rm -rf on the files I needed, so thanks for that anyway!!

Yeah sorry not a fan of the illusion of a free tool and then when you want to actually use it you have are hit with a paywall... it should be UPFRONT/ transparent.

I wasted my time after getting all excited about a potentially cool new tool.

This gets a down vote in my book just because of that - even if the tool seems qualitative.

Most Mac cleaners just nuke browser caches, so seeing one that specifically targets abandoned node_modules and Xcode derived data is refreshing. Setting up automated weekly sweeps of dormant Docker images would be a killer use case for this. That alone would save me from having to do a manual disk space panic cleanup every few months.

This seems perfect for Mac users! I’m not sure why, but every time I hit “Clean Up” after selecting all the junk files found in the scan, the app crashed 😔

Still, this feels like a long overdue app for Mac Users, and I’m excited to see where you take it.

Does it detect Xcode derived data automatically or do you need to point it to specific folders? Congrats on the launch!

Finally something that understands dev Macs are a different beast. My 512GB fills up fast between Xcode derived data, stale node_modules in forgotten projects, and Docker images I forgot about. The review-first approach is smart — I've been burned by cleaners that nuked things I actually needed. The CLI workflow is a nice touch too.

developer machines accumulate junk differently than regular users - node_modules alone can get out of hand fast. curious what the most common space hogs you find are. I am guessing it is a mix of build caches, old simulator runtimes, and abandoned Docker layers.

Congrats on the launch, Matheus! 🎉 The eye contact problem is the one that kills most teleprompter setups — the moment you’re clearly reading, you’ve lost the viewer. Solving for natural delivery rather than just text display is the right angle. About to launch OceanMind, an AI-powered breathwork iOS app, and recording pitch videos and app store content is exactly where something like this would come in handy. Is there an iOS app, or is it currently web/desktop only? Would love to use it directly from my iPhone while recording.

Congrats on the launch!

Hey guys, I decided to give people who supported me here a 10% discount :) Just use the code PRODUCTHUNT10 at checkout.

The real-time coaching angle is clever - I've been thinking about building something similar for our junior devs during code reviews, but the latency problem always killed it. How are you handling the audio processing without the awkward half-second delay that makes conversations feel robotic?

this is a smart idea because speaking naturally while keeping eye contact is still such a pain point for presentations and videos. i like that you’re solving for confidence and delivery instead of just showing text on screen. have you found it works best for live presentations, recorded videos, or both?

The session-amnesia problem is real and underrated. Been running OpenClaw agents on complex tasks and the knowledge loss between sessions is one of the biggest time sinks. The framing as a community memory layer is smart because isolated per-project memory doesn't compound value. The IP/sensitive code concern raised in the comments is worth addressing prominently in your docs. The environment context piece (versions, deps) is crucial for signal quality. Congrats on supporting OpenClaw natively!

knowledge sharing between agents is a problem I run into constantly - each agent starts cold and rediscovers the same things. how does Context Overflow handle conflicts when two agents have contradictory knowledge about the same topic? and is the knowledge graph per-project or shared across projects?

Can the generated plugins access Figma APIs like variables and components or is it limited to basic layout? Congrats on the launch!

Very useful product

Nice product

love this idea making figma plugin creation feel accessible through plain language is such a strong unlock for designers who have ideas but never cross the coding barrier. the part about building a whole suite of custom plugins is especially interesting. what kind of plugins are people generating most often so far?

I think the non-blocking approach is a refreshing take. How does GentleLimit define “mindful” interventions — are they timed prompts, intention-setting dialogs, or something else entirely?

This is really cool. I think that I saw something similar to @focusedOS

What kind of permissions do I need to give this tool, and what data leaves my computer?

Is there any pricing available?

The Mermaid diagram preview sold me instantly. I write a lot of docs in Markdown and always have to open a separate editor just to check if the diagrams render correctly. Does it support live-reloading, like if I edit and save a file while Quick Look is open?

Nice idea !! extending Quick Look for dev files makes a lot of sense. @parcse

Would be cool to see support for log files or very large files too : is that something you're planning to add?

Hello Michael, happy launch day. The "replaces a stack of tools" part hit me. Every creator knows that struggle.

Was on your site for a bit. Sarah's testimonial caught my eye. "Setup took an afternoon and we were collecting payments the same day." That's what people want.

One thing I noticed on the homepage. Your headline says, "Build the membership platform your audience deserves." That's safe. But every membership tool says the same thing. The real differentiator is in the subhead: "Gate content, manage members, collect payments, and grow your community — all in one place."

That's the actual value. One tool instead of five.

Flip them. Lead with "All in one place." Then explain what that means.

This is just honest feedback from someone who spends way too much time analyzing homepages and finding gaps in positioning. Hope it helps, @michael_okedigba.

Nice idea. Keeping documentation, help centers, and product knowledge in sync is a real challenge for many teams. I like the focus on making content easier for users to actually find and use. How does Gately keep documentation automatically updated when changes happen in the codebase?

Hey! This is really cool. Does Gately handle content gating, community features, and payments natively, or does it rely on third-party integrations for any of those pillars?

Congrats. No data collection is a huge win btw! How does it handle threaded replies or templates to keep that joy going?

love this kind of product because gmail is powerful but not always pleasant to use day to day. the centered compose alone feels like a small change that probably makes a big difference. are users loving the writing experience more or the cleaner inbox improvements more?

I'm loving the improved readability already! Just a quick question: is the "mark all as read" button supposed to be in the bottom left instead of the top right, like in the video?

I really appreciate the “no data collection”. Does Joy work with Google Workspace (business) accounts, or is it currently limited to personal Gmail accounts?

What improvements do you offer on the search function? Current search function quite bad compared to outlook... very frustrating

hi Product Hunt 👋

im excited to introduce Chat

when building services, we often spend weeks building a frontend before users can even try the product, tools especially AI services, internal tools, and APIs; the frontend is mostly just an interface to send requests

so i built Chat

Chat is an MCP client that lets you connect your own service’s MCP server and instantly expose it through a chat interface

instead of building a full website or UI, you only need to focus on:

• backend logic

• APIs

• automation

• AI workflows

once connected, your service becomes available directly through chat

this makes it much easier to:

+ launch MVPs faster

+ experiment with new ideas

+ ship internal tools quickly

+ expose APIs as usable products

think of it as a chat interface/ ChatGPT UI layer for your backend

id love to hear your thoughts, feedback, or ideas for how you might use something like this 🙏🏼🙏🏼

Does Chat support authentication and user-level access control for the exposed chat interface, or is it purely open once connected to the MCP server?

really interesting idea turning a backend into something users can actually try through chat feels like a smart shortcut for mvp building. i like that it removes the pressure of building a full frontend just to validate an idea. how are you thinking about customization for teams that want the chat experience to feel more branded or product-specific?

This feels like the moment AI stopped being a tool and started becoming a true collaborator.

Super fast, surprisingly smart, and actually useful in real workflows — not just demos.

If this is the direction, we’re entering a whole new phase of AI 🔥

I work from cafés a lot so the shoulder surfing detection is really appealing. Curious how fast the blur kicks in when someone walks up behind you. Is it instant or is there a brief delay?

This is very cool. How does Gaze Guard handle dual-monitor setups — does it track gaze per screen independently, or does it blur all screens simultaneously when you look away?

Check if I understand the product. So, if the person whose laptop it is, it will not blur then, but if in front of the camera somebody else comes, then it will auto-blur the screen. Is that correct? And then it unblurs whenever the original owner arrives in front of the laptop. Is that correct? Is it using some kind of facial recognition?

Can it also blur specific parts of the screen?