🔥 Roast my idea: drop your ideas and get brutally honest feedback 🔥
by•
It's simple. Drop your next big idea and get some brutally honest but hopefully valuable feedback. The rules are:
Drop your idea, you don't have to go super in depth but give a sentence or two about what it is and does.
Get roasted, duh
Roast someone else's.
That way we create a cycle of feedback
2K views
Replies
I’m building a tool where any clinic can launch their own assistant, on WhatsApp or their website, in seconds.
They just scan a QR or paste a script, and it's live.
No setup, no credentials, no tech knowledge needed.
Behind the scenes, the system generates a dynamic n8n workflow per client,
so it’s not just a chatbot, it can actually understand context, book appointments, answer real questions, and adapt to the patient’s needs.
Curious to hear what you think.
Is this useful? Overkill? Not solving a real problem?
@lautaro_nievas It sounds good, but how do you overcome the lack of reliability in an LLM's output? Just make your users eat up that additional cost for iterative failed runs while it brute forces it's way to completion?
@kalebautomates Great question, and honestly, that’s been one of our main concerns from the start. We don't rely on LLMs to “figure things out blindly” or brute force responses. Instead, we keep the agent's role tightly scoped: it detects intent, pulls structured info from our tools (like FAQs or available slots), and reformulates the answer naturally. Most of the logic, booking, validating data, even retries, is handled outside the model (in our backend or n8n). This way, we get the benefit of a human-sounding response without giving the LLM too much responsibility or room to fail. Also, we cap token usage, avoid complex reasoning tasks, and use hard rules for fallback: if the model can’t confidently respond, it just asks the user to clarify. No infinite loops, no surprise bills, and no “guesswork” pretending to be smart.
@lautaro_nievas If all else fails, you can try out TSCE and get a quick accuracy or task adherence uplift ~+30pp over baselines. Will save you a ton on iteration or failures. Won't let me send you the github in a reply, but its an open framework and I'm just in it for the research. Feel free to check out my "product" and incorporate it into your workflows! P.S. I'm a data scientist and work in healthcare! Feel free to hit me up if you need some domain info
@kalebautomates Oh man, that's gold. I'm not that advanced with that level of data. It's definitely something I'll have to take a look at. I'd love to get in touch with you to show you the app and get your feedback!
@lautaro_nievas How do you handle the HIPAA part of it all?
@bismich4 Yes, that is something I am still trying to resolve, since for example WhatsApp is not valid for Hipaa, I would mainly start launching it in my country, which does not have problems with those regulations, and if not, also start with beauty salons, since this application can be adapted to any type of business.
@lautaro_nievas That's what I was thinking too...start with other type of businesses first - then slowly move into healtcare
DataTable.dev
A local city app with all possible food options existed, where users type an ingredient, and it shows dishes that contain it. Like sometimes I really want to eat some fried sweet potato in Kyiv and I make a whole research to find it :)
@angelinashv no way to get the data in a reliable way
Product Hunt
@angelinashv @_tijs If you can get the data it'd be cool
DataTable.dev
@angelinashv you’ll have a classic critical mass problem though. How will you get business to signup without users and vice versa
@angelinashv Definitely not easy to start with as @_tijs said, but that would be quite fun if you could get the data in any way. But you'd need a lot of users to make it relevant for both restaurants and consumers..
@angelinashv Nice Idea, just might be difficult to source good data for building a nice recommendation system.
@angelinashv Your servers would die if you type "Salt" or "Water"
A wizard-style platform that assists founders in creating optimal pricing tier structures through AI-powered recommendations, cost calculations, and industry benchmarking. It features an intuitive pricing tier builder.
Product Hunt
@pradomaker This is really good. Unique and really needed.
@steveb you should take a look https://www.producthunt.com/products/tierwise
Product Hunt
@pradomaker on waitlist
Product Hunt
@pradomaker Love this actually
@pradomaker who are the first customers?
Product Hunt
I'll start: A Tiktok style app that lets people vertically scroll through different property tours and allow users to directly contact the agent selling the listing.
@aaronoleary There are many platform where you can see property pictures and details, contact agent, what's new ?
Prototypr
@aaronoleary it would be good for the real estate agencies I think - you can sign up the agencies so it can be very curated, or it might get full of scammers. Then it can be like a property search tool but a lot cooler and engaging
Product Hunt
@graeme_fulton scammers is a big concern. The big goal really is just to make property search more engaging and more of a novelty. One issue though is APIs, real estate industry is so antiquated in how the handle them
Prototypr
@aaronoleary what do you need APIs for? Actually I thought property sites might have RSS feeds of recent listings, but just tried searching for their feeds, and they seem to be non-existent too..that's really odd, I think you're right about antiquated
Product Hunt
@aaronoleary Reelty?
Product Hunt
@steveb good name lmao
DeepTagger
@aaronoleary vertical swiping encourages skim-reading. Key details like square footage or year built get ignored in favor of flashy visuals 👀
@aaronoleary "We use what we find in search to refine our preferences." I like what you're building. Maybe swiping right gives more pictures from the same property.
TradeX
An AI messaging app where people can talk in groups and use AI which will read your messages and replied in the group. Can be used to create storyline’s, do group discussions, brainstorming sessions, couple therapy and more more.
Product Hunt
@divijgoyal If AI is doing the writing, is this productive for therapy, brainstorming, etc?
TradeX
@steveb its not AI is writing. Imagine a group chat where AI agent is also a group member. So this group chat be used by a group of people for multiple use cases.
Haimeta
@divijgoyal That's a great thought! People often talk a lot in group chats, but many decisions or ideas are often forgotten or left unfinished.
TradeX
@gin_6078 I guess so. Launching soon :)
InterviewBee AI
How about "live AI assistance during actual job interviews?"
Most candidates prepare well, but freeze at the right moment to recollect and structure their answers confidently. This AI assistant reads the CV and the job requirements and generates talking points/ notes in less than 2 seconds.
Open to feedback.
Haimeta
@bismayy_mohapatra Is the target audience the interviewer or the candidate?
InterviewBee AI
@gin_6078 It's for the candidate only - students/ professionals/ freelancers.
For the interviewer, it will be a different product.
Deamoy
@bismayy_mohapatra This will make the opposite side come out with an anti-AI AI that summarizes the real capabilities behind the interviewee's package. lol
@bismayy_mohapatra Isn't that cheating?
InterviewBee AI
@nayan_goenka1 I don't disagree, but I am sure that the interviewing process needs an overhaul, considering access to ChatGPT and similar tools in the market. I believe we should be adopting open web and open books interviews, where we test how the candidate can extract information and build their narrative in a quick span. And obviously, the smarter interviewers will stop asking surface-level questions and dive deeper to evaluate. What are your thoughts?
Haimeta
An app designed to guide new-beginner authors, using AI to help organize world-building, timelines, and outlines.
@gin_6078 will it offer customizable templates?
@gin_6078 Is this specific to a genre? I would be interested in Non-fiction and Autobiography
MeetSpace
An app to help builders/founders work out their potential cash flow before they start a new business, assessing the financial feasibility of their idea? Currently in progress at https://flowforecast.app
MeetSpace
Could do with some of the features @pradomaker mentions above!
@lamplightdev we can do something together, like a partnership or something like that, i think the both apps could improve each other.
MeetSpace
@pradomaker Great idea - just joined your wait list
Product Hunt
@lamplightdev Looks great!
@lamplightdev could save a lot of headaches down the line 🙌
Great idea. Here’s mine:
A browser plugin that takes notes and creates summaries from meetings, emails, and docs, all in one place.
Happy to hear what’s wrong with it and I’ll roast one in return.
@olivia_bennet I like the idea! However, don't a bunch of these already exist and have issues with reliability and "hallucinations"? How do you plan on overcoming that?
An AI-powered tool that converts basic product images into a full product ontology—auto-generating detailed descriptions, enhanced visuals, and optimized listings for platforms like Google Shopping, Amazon, and Shopify—streamlining eCommerce content creation from scratch.
@raja_j_kumar This sounds amazing! Do you have an alpha or beta version available for testing?
@nick_moure Thanks for your kind words. I’ve compiled the necessary literature and detailed functional requirements and all the solution needs—including rationale and edge scenarios—for building such an AI-driven content generation ecosystem. Beyond just automating current workflows, the goal is to enable capabilities that go further than today’s tooling allows, especially in the areas of product ontology enrichment, image-to-attribute mapping, and omnichannel optimization.
Happy to discuss further if you're interested in collaboration.