Self-Promotion
Show off what you're working on
Start new thread

What are you building? Drop your AI product below

by•

Doing one of these because I get more from reading what others are building than from any newsletter.

I'll start.

Building AI Hive, an enterprise AI agent platform that helps mid-market and enterprise teams get from AI pilot to production in weeks instead of quarters. The hard problem we keep solving: compliance, model flexibility, and the lack of in-house AI talent on the customer side.

Product Hunt page:

Your turn. Drop:

- What you're building

- Who it's for

- The hardest part you're solving right now

Will read everything and upvote what resonates.

2.4K views

Add a comment

Replies

Best

Thanks for this thread.

A year ago I started teaching an AI to cook, on evenings and weekends. This week it went live on the App Store and Google Play. Solo build: services, web, native mobile, and the pipelines in between.

CuocoMitra is an AI cooking and nutrition companion. Plenty of apps bolt a chatbot onto a food database. I wanted the inverse: a rigorous system where the AI is the interface, never the authority. These are the design decisions I would defend in any architecture review:

1. The LLM never touches a number.
Models are great at intent, bad at arithmetic. My agents can only emit validated, schema-checked actions (swap this ingredient, adapt to my diet, plan my week). Free-typing an answer is not an option the model has.

2. Deterministic domain engines do the real work.
All food math lives in pure, isolated engine packages: no network access, no storage access, same input gives the same output. Ingredients resolve against a real nutrient dataset through a matching scorer, and macros are computed, never generated. 34g of protein is a calculation, not a hallucination.

3. The AI plane is isolated from the data plane.
Agent sessions, tool orchestration, and live token streaming run in their own service behind their own auth boundary. Model latency or an outage never blocks a normal read or write. Worst case, the product degrades into a great manual cooking app, not a brick.

4. Every AI edit is a versioned mutation.
An adapted recipe is a new version with full history, revertible in one tap. The agent has write access to nothing it cannot undo.

5. Retrieval over memory.
Culinary and nutrition answers are grounded in a curated knowledge base through an embeddings pipeline. Traceable to sources, not improvised from whatever the model memorized in training.

6. One brain across every surface.
Photograph a plate and a vision model recognizes the dish, then the same deterministic engine attaches the macros (I call it MitraVision). Ask for a diet swap and a fusion of culinary knowledge and your learned taste profile keeps the dish authentic: a Thai curry should still taste Thai after the adaptation. One profile of your palate, constraints, and goals drives search, meal planning, and personalization everywhere.

7. Models are replaceable, guarantees are not.
The AI layer is deliberately thin. When a better model ships, it slots in, and every correctness property above stays untouched.

Quietly proud of one more thing: a single typed API contract powers both the web app and the native mobile app, and agent progress streams live to both.

If you build LLM agents, retrieval systems, or you just cook, I would genuinely love your take on the approach.

 Really appreciate you laying out the architecture like this, not many makers go into this much detail. The "LLM never touches a number" rule hit home for me, I've seen so many AI cooking/nutrition apps just let the model guess macros and it's scary how confidently wrong they get.

Separating the AI plane from the data plane is smart too. Feels like the kind of decision people only make after getting burned once. Curious how you handled the matching scorer for ingredients that have weird regional names though, that part always trips up nutrition datasets in my experience.

Btw if you ever want to swap notes on agent architecture, I'm building AI Hive, an automation flow platform for teams. Would genuinely love your take on it if you have 2 mins, and a review would mean a lot too:

 Сooking is a nice AI use case because the output has to be useful :)

I’m curious what you’re focusing on most right now? helping users decide what to cook from what they already have, adapting recipes to preferences and restrictions, or making the cooking guidance easier to follow?

I'm building Get Styli — an AI stylist that lets people virtually try on clothes before buying them online.

Who it's for: Anyone who shops for fashion online and wants to know how an outfit will actually look on them before making a purchase.

The hardest problem we're solving: Creating realistic, identity-preserving virtual try-ons that accurately maintain a person's body shape, skin tone, facial features, and clothing fit across a wide variety of garments. We're also working to make the experience fast enough for everyday shopping.

Would love your feedback if this resonates!

Product Hunt:

 Thanks Gaurav, virtual try-on that actually preserves identity (body shape, skin tone, facial features) is way harder than most people realize. Every early try-on tool I've tested makes everyone look weirdly airbrushed or gives you clothes floating on a mannequin. Solving realism plus speed for everyday shopping is a legit technical challenge.

The consumer angle is tough because you're competing with attention, not budget, but if you nail the "wow" moment on first use, this kind of tool goes viral fast on TikTok and Reels. That could be your biggest distribution unlock.

Wishing Get Styli continued growth and hope you nail the realism at scale. Just upvoted and followed 👗 If you have time, would love your eyes on AI Hive too, we're building for enterprise AI agent workflows: . Any support really helps.

Hello!

The What: Sela - a caretaking task management app built by a caregiver for those family members providing free home health aid/care to/for a loved one(s).

The Why: There are over 60million people who provide unpaid health care to a loved one in the US alone. These loving family members have access to multiple apps that are meant to make staying organized easier, however they only provide tracking of specific items forcing those caregivers to use multiple apps or pay for the features they truly find helpful or use pen/paper. Adding yet another inconvenience to the already overwhelming list of daily tasks most caregivers work through daily.

The Challenge: At present my current challenge is getting users who will also take the time to report issues/provide a review of the app so that gaps and risks can be identified and resolved. And to also show proof of concept for funding to expand the app.

Go to to try the app and leave a review.

 60 million unpaid caregivers in the US alone and most of them are duct-taping together 3-4 apps to stay organized. That's a real problem that deserves a real dedicated solution, not a workaround. The fact that you built this as a caregiver yourself gives it a kind of credibility that most health apps don't have.

The user review challenge is tough early on, especially for something this personal. Worth reaching out directly to caregiver communities on Reddit or Facebook groups — those folks are vocal and loyal when a tool actually fits their life. Rooting for Sela

Uptiqr

What I'm building:
Uptiqr is a simple uptime monitoring platform that helps developers and businesses know when their websites and APIs go down. It monitors services, tracks performance, and sends instant alerts when something breaks.

Who it's for:
Developers, indie hackers, startups, and small businesses that rely on their websites or APIs and need an easy way to catch downtime before their users do.

The hardest part I'm solving right now:
Building a monitoring system that is reliable enough to catch real issues quickly while avoiding false alarms. The challenge is balancing fast detection, accurate alerts, and keeping the experience simple compared to more complex monitoring tools.

 Uptiqr is one of those tools where the value prop is obvious the moment something breaks at 2am and you're the last to know. False positive fatigue is real though — that's the thing that kills adoption on monitoring tools faster than anything else. Curious how you're handling alert tuning. Upvoted both! If you have a sec, I'd love a review or upvote on AI Hive — we're building automation workflow infra for enterprise teams:

ChatForger

What I'm building:
ChatForger is a white-label AI chatbot platform that helps agencies create and resell branded AI chatbots for their clients. Agencies can train chatbots on client content, embed them on websites, and manage client analytics through their own branded experience.

Who it's for:
Marketing agencies, web development agencies, and freelancers who want to add AI chatbot services to their client offerings without building the technology themselves.

The hardest part I'm solving right now:
Making AI chatbots reliable enough for real businesses while keeping the setup simple for agencies. The challenge is balancing accurate answers from client data, easy deployment, and a seamless white-label experience.

 Both of these are solving genuinely painful problems. ChatForger hits a gap I've seen agencies struggle with for years: they want to offer AI chatbots but building the infra themselves is a non-starter. The white-label angle makes total sense.

Building MySitePilot () — a platform that automates business workflows across websites and web applications using natural language.

Instead of requiring APIs or custom integrations, MySitePilot interacts with existing web interfaces to execute repetitive operational tasks.

Right now we're focused on helping teams automate browser-based workflows that are still performed manually every day.

We recently launched our live demo where anyone can try it on a real e-commerce website: ()
You can see the details how this demo site works here: ()

The hardest challenge we're solving right now is making workflow automation reliable across dynamic web applications while keeping setup simple enough for non-technical users.

 Really cool approach going the interface route instead of forcing APIs, that's such a practical fix for teams stuck with legacy web tools that never had integration in mind. Tried clicking into your live demo and it's a nice way to actually show the automation working instead of just talking about it.

Keeping workflow automation reliable across dynamic web apps is genuinely the hard part though, DOM structures change all the time and most bots break the second a site updates its layout. Curious how much manual tuning MySitePilot needs per site or if it adapts on its own.

Hope this keeps scaling well for you guys, upvoted and following along. Would love it if you could check out AI Hive too when you have a sec, it's built around helping teams set up automation flows more efficiently:

 Hi! Thanks so much for the feedback and the upvote, really appreciate it!

You hit the nail on the head regarding DOM volatility that’s always the hardest with web automation. Traditional selector-based bots fall apart the second a class name changes. With MySitePilot, standard HTML/modern layouts are usually straightforward, but platforms where HTML is dynamically generated (or heavily customized canvas/shadow DOM elements) are where the real challenge lies. We combine visual element anchoring with semantic intent so the agent adapts even if the underlying layout shifts slightly, though complex web apps still require occasional fine-tuning.

Just checked out AI Hive as well and love how you guys are tackling the efficiency side of setting up automation flows! Dropped an upvote your way as well. Best of luck with the launch!

 Really appreciate you breaking that down, the visual element anchoring plus semantic intent combo makes a ton of sense as a way to survive layout drift without needing constant rewrites. Good to know standard HTML holds up fine and it's really just the canvas/shadow DOM stuff that needs the extra tuning, that's a pretty honest tradeoff to admit rather than claiming it handles everything.

Thanks so much for checking out AI Hive too and for the upvote, means a lot coming from someone building in the same space. Wishing MySitePilot a smooth scale from here, will keep following your progress.

What I'm building: BurntToastHub — AI prompt kits and Excel tools for corporate professionals, plus my book Burnt Toast & Boardrooms.

Who it's for: FMCG, HR, trade marketing, and field sales people drowning in repetitive reports, reviews, and admin work.

Hardest part right now: making the tools specific enough to actual corporate workflows that they beat generic "AI productivity" prompts — not just another ChatGPT wrapper.

Live on PH today, would love feedback from this crowd.

PH Listing :

producthunt.com/products/burnttoasthub

 Congrats on the launch today, that's always a nerve wracking day. The "not just another ChatGPT wrapper" problem you mentioned is so real, feels like half the AI tools out right now are generic prompts dressed up as products.

Making it specific enough to actual corporate workflows like FMCG or trade marketing reviews is the right call though, generic never sticks with people buried in repetitive admin work. That's usually where the real value hides, in the boring specific stuff nobody else bothers to solve.

Hope BurntToastHub crushes it today, upvoted and following to see how the launch goes. If you get a chance, would really appreciate your thoughts on AI Hive too, we help teams build automation flows for their workflows:

 Appreciate this, Nolan — and well said. "Boring specific stuff nobody else bothers to solve" is basically the whole thesis behind BurntToastHub, glad that landed.

Thanks for the upvote and follow, means a lot on launch day. Heading over to check out AI Hive now, the pilot-to-production gap is such a real problem for mid-market teams. Upvoted 🙌

 Haha glad that landed man, and congrats again on the launch numbers so far. The pilot to production gap comment means a lot coming from someone who's clearly thought hard about the same "not another wrapper" problem from a different angle.

Really appreciate you taking the time to check it out on launch day of all days, that's not a small thing. Wishing BurntToastHub keeps climbing

What you're building — TrailSmoke () , a wildfire smoke and air quality tracking application centered on a live AQI map, pulling in NOAA smoke modeling data (near-ground air quality vs. upper-atmosphere/trajectory models) and US wildfire statistics for seasonal projections.

Who it's for — People who are planning outdoor activity around wildfire smoke: hikers checking trail safety, runners/cyclists seeing if they want to risk breathing heavy, and generally people trying to think about what the smoke will be like 1-2 days from today.

The hardest part you're solving right now - Not too hard. My product is free and for fun. So I'd just like to get people to use it!

 Thanks for sharing, wildfire smoke tracking for hikers and runners is such a practical, unglamorous problem that actually matters a lot in the western US right now. Pulling in NOAA data instead of just a generic AQI number is the detail that makes this useful vs just another weather app.

Love that it's free and just meant to be used, that's a refreshingly simple goal in a sea of monetization-first launches. Upvoted and following, hope this gets in front of the hikers/runners who need it.

If you don't mind, would appreciate a look at AI Hive too, automation flows for productivity, happy to return the upvote/follow:

What I'm building


Linkwise: an AI-powered read-later app that turns saved articles, newsletters, RSS, emails, PDFs, and YouTube videos into a usable personal knowledge base. It includes instant AI insights, chat with your entire library, and Creator Mode (which generates quizzes, flashcards, mindmaps, and reports from your content).

Who it's for
People who save a lot of content but actually want to learn from it and create with it: students, researchers, writers, creators, and lifelong learners.

The hardest part I'm solving right now
Turning passive “read later” saving into active knowledge and creation. Most tools just store links. We’re building the AI layer that makes your saved content actually useful and actionable.

know more here : linkwise.app

 Thanks for posting this, the read-later graveyard problem is so real, I have hundreds of saved articles I will probably never open again. Turning that into an actual usable knowledge base instead of just a bookmark pile is the right problem to solve.

Creator Mode generating quizzes/flashcards/mindmaps from saved content is a nice touch for actually retaining what you save instead of just hoarding links. Curious how it handles video content specifically since that's usually harder to summarize well.

Thanks , supports youtube videos transcript generation as well and it summarize the video content using that transcript.

 Thanks for the quick answer, that's actually really solid. YouTube summarization from transcripts is one of those features that sounds simple but saves a ton of time once you've got hundreds of saved videos sitting there unwatched. Curious how it holds up on longer form stuff like 1-2hr podcasts versus a short clip.

What I'm building: ChainPoints, a Chrome extension that sits on top of Google Flights and shows you the cheapest way to book any flight with the points you already have. All the math, right on the page.

Who it's for: Anyone with points and no plan. I built it because I was that person: 17 cards, 500,000+ points, and no clue if I was redeeming any of them right.

Problem I'm solving: Most people are sitting on credit card points they never use, even if they do, its for statement credits or travel portal, cause redeeming them well is a nightmare. 15 tabs, transfer partners, award charts, napkin math... so the points just sit there and lose value.

Hardest part right now: It's literally launch day, and I'm doing this with zero ad budget as a solo, self-taught founder. 100% organic: PH, communities, and word of mouth.

PH listing: (Pro access free for 30 days)

And if you want proof it works: drop where you want to fly + roughly how many points you have in the comments, and I'll personally find you the cheapest way to book it.

 Thanks for sharing, and respect for shipping this solo with zero ad budget. 17 cards and 500k points with no clue how to redeem them is basically every points hoarder I know, myself included.

Doing the math directly on the Google Flights page instead of another dashboard to check is the right call, less friction means people actually use it instead of bookmarking and forgetting. Bold offer to personally book flights in the comments too.

Upvoted, and following to see how launch day goes. Would really appreciate you checking out as well (automation flows for productivity), happy to return the upvote/follow

First
Previous
•••
567
•••
Next