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: https://www.producthunt.com/products/ai-hive
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.
4K views
Replies
What you're building: FasalVision — a live AI farming assistant app available in 9 languages. It gives real-time crop disease detection, weather-based advisory, mandi (market) prices, and personalized farming guidance — all powered by AI, built for farmers in rural India and beyond.
Who it's for: Small and marginal farmers who don't have access to agronomists or agricultural experts. Farmers who speak local languages (Hindi, Punjabi, Bengali, Tamil, Telugu, Marathi, and more). Anyone in agri-tech or rural development looking for an AI-first farming tool.
The hardest part we're solving right now: Getting farmers (who are largely offline or low-literacy) to trust and use an AI app. We're at 40+ installs organically — trying to reach 100 this week without paid ads. Distribution is the real challenge, not the tech.
Check it out: https://fasalvision.com
@rajnandan_r_kushwaha "Distribution is the real challenge, not the tech" - yeah, this hits. Getting low-literacy users to trust an app enough to open it a second time is a completely different problem from getting them to install it.
Curious whether you've tried working through local agri extension workers or village-level NGOs as a distribution layer? They already have farmer trust baked in, which is basically the hardest thing to build from scratch.
Hey everyone! I'm Arjun, a solo technical founder building from India.
I spent months staring at my own app and slowly went blind to it - every clunky screen looked normal because I'd seen it 500 times. The stuff that actually lost users (a confusing step, a dead end, a slow screen) I just couldn't see anymore.
So I built TestSting: you give it your Android app and an AI uses it like a brand-new user, then writes up the UX problems it ran into - each with a screenshot, plus a health score. A fresh pair of eyes, on demand.
It's early and Android-only for now, and I'd genuinely love feedback from this community. Free sample report, no signup: teststing.com/sample
@arjun_h_n The "gone blind to your own app" problem is so real, and honestly it's worse for solo founders because there's no one to tell you the onboarding is broken. TestSting sounds like it's solving exactly that.
Does the AI test on actual user flows or just visual/UX issues? Asking because I'd love to know if it catches logic gaps too, not just slow screens.
@nolan_vu Yeah — honest answer is somewhere in between. The AI actually walks the app: taps, types, logs in, moves through real screens like a new user. So it's testing actual flows, not eyeballing a screenshot — it catches the flow-level stuff: dead ends, confusing steps, screens it gets stuck on.
Easiest way to see if it catches the kind of gaps you mean — throw one of your apps at it , no signup: teststing.com. Genuinely curious whether it holds up on the logic-gap stuff for you.
@arjun_h_n sure, let me examine your suggestion later on, thank you very much though
Hello, my name is Seán and I am the founder of Ontoworks.
We are building software for automated video production, where we deep-analysis Footage-databases in order to accelerate production pipelines.
Our focus are production companies and TV broadcasters, with whom we are launching initial projects at the current moment.
The hardest part is improving performance based on user feedback and making our system adaptable and demonstrable in order to reach integration within our customer´s workflows.
@sean_mcfadden thanks for your sharing Sean
Automated video production for TV broadcasters is one of those use cases where the technical bar is genuinely brutal. Footage databases are massive, metadata is inconsistent, and broadcasters have very specific workflow requirements that change between channels. The fact that you're already running initial projects with real production companies is solid traction for that space.
The "adaptable and demonstrable" part is what most enterprise AI tools underestimate. Broadcasters don't want a magic box, they want to see exactly what the system did and why. Have you found that early customers care more about speed gains or about creative control over what the AI surfaces? That answer usually shapes the entire roadmap.
TAM Network. an AI-native alternative to linkedin. the AI part is the agent identity row. when an AI agent does real work, it gets a row on the receipt next to the human contributors. recruiters see exactly which line item was AI and which was human. transparency, not hiding. v2 launches aug 12. thetamnetwork.com
@thenameisarian Interesting, alternative to LinkedIn you said
"Transparency, not hiding" as the positioning is sharp. The whole industry is moving toward pretending AI involvement isn't happening, so calling it out directly and making it a feature instead of a flaw flips the script in a really smart way. Recruiters needing to know what was AI vs human is going to matter way more in 2026 than most people realize.
Aug 12 launch is close. How are you planning to handle the early skepticism from candidates who'll worry that flagging AI work makes them look less qualified? That's usually the friction point that decides whether transparency-first platforms actually take off or get rejected by the people they're designed to help.
@nolan_vu this is the right friction to name. the answer we are designing toward is opt-in per receipt, not mandatory across the profile. a builder publishes some receipts that flag the agent contribution and other receipts that were 100 human. the market self sorts. some recruiters will value the verified AI disclosure. others will not. over time the receipts with agent transparency outperform on credibility because the alternative (resumes claiming pure human work that we all know used AI) has already lost its signal. the platform does not force the choice. it lets the market price it.
@thenameisarian opt-in per receipt is the right call. forcing full transparency kills adoption, but letting the market reward it over time is a much smarter bet.
the interesting part is what happens when enough receipts with agent disclosure start outperforming the ones without. at that point it stops being a choice about ethics and starts being a competitive signal. builders will opt in because it works, not because they're told to.
curious whether you're seeing early data on that credibility gap yet, or still pre-launch hypothesis at this point.
DevCleaner
Just shipped the biggest @DevCleaner update yet.
📦 Project Hibernation lets you archive projects you are not actively working on, off your disk into a single file, and restore them in one click whenever you need them. It handles dependency reinstall on restore, is iCloud aware, and verifies every archive before touching the original.
👻 Also added Ghost Projects (per project Xcode DerivedData breakdown) and Project Artifacts (stale node_modules and build folders ranked by age).
✨DevCleaner stays free for manual cleanup and scanning. The automation layer is now Pro, with a 14 day free trial. Lifetime licenses are capped for early supporters at 29 dollars.
Would love feedback from the dev community here. What would you want a tool like this to do next?
Download here
👉🏼 devcleaner.app 👈🏼
@dawedeveloper thanks for shipping this, the Project Hibernation feature is exactly what i've been wanting for dormant projects that i can't bring myself to delete.
the iCloud-aware part is a nice touch, a lot of disk tools ignore that and you end up with half-archived projects that sync back and confuse everything. one thing i'd love to see next: a scheduled hibernation mode where projects auto-archive after X days of inactivity without me having to remember to do it manually.
DevCleaner
@nolan_vu That's a great idea! Thank you for that.
Building thematicanalysis.ai
It's an API that turns piles of open-ended text such as survey answers, reviews, interview transcripts, support tickets into actual themes. You send the text, you get back the themes, the quotes behind them, sentiment, and a confidence score as clean JSON.
The itch I'm scratching: everyone can prompt a model to "find themes," but it falls apart at scale. The theme names change every run, the counts never reconcile, and you can't point to the quote that justifies a finding. So I'm building the boring-but-hard part (consistency, persistent codebooks, evidence, confidence) as one endpoint, grounded in the actual six-phase thematic analysis method.
Sandbox is opening to the waitlist in waves → thematicanalysis.ai
Curious what you'd point it at if you had it
@unclej thanks for sharing this, Olajide, the consistency problem you described is the one nobody talks about honestly.
prompting a model to "find themes" works once in a demo and falls apart the moment you need to compare run 1 to run 2. the persistent codebook approach is the right call because reproducibility is what turns a toy into a research-grade tool. i'd point it at customer interview transcripts first, that's where the "counts never reconcile" pain shows up most visibly in practice.
Building MartinLoop.
It’s for founders, developers, and teams running AI coding agents like Claude Code, Codex, Cursor, OpenCode, or custom agents in real repos.
The hard problem we’re solving: AI agents can write code, but they still run wild. They loop, waste tokens, change files, wait for approval, and leave teams guessing what happened.
MartinLoop adds the missing control layer: budgets, safety checks, verifier gates, rollback, notifications, and run receipts so every agent run is easier to trust.
Product Hunt: https://www.producthunt.com/products/martinloop
Website: https://martinloop.com
appreciate any upvotes and Stars on github, (download the open source and if its useful first ofc)
@keesan12 congrat on your launching mate. I have followed and upvoted your product. Hope that you can stil upgrade the product and earn more clients and followers later on
If possible can you do the same, upvote or give some review for AI Hive also, thank you very much in advance
https://www.producthunt.com/products/ai-hive/
What you're building
ReadyToPlay24 is a team management app that replaces the WhatsApp group chaos for sports organizers. The captain creates a group, adds a match — the app automatically sends push notifications to all players and manages the waitlist. When someone drops out, the first person on the waitlist gets a push notification and jumps in automatically. No more manual counting, no more "who replied?", no group chat scrolling. Available as a PWA (web + iOS) and native Android app, in 8 languages.
Who it's for
Three main users:
- Casual group organizers — someone who runs a weekly 5-a-side game with friends and is tired of managing it all through a chat group
- Coaches — running training sessions for 2–3 groups, needing attendance tracking and PDF reports per session
- Amateur league organizers — managing 10+ teams, needing rotation logic, stats, and season reports for sponsors
Players (non-captains) join for free — only the organizer pays. That's a key part of the model.
The hardest part you're solving right now
Getting organizers to make the switch. The product solves a real pain, but people are deeply habitual about WhatsApp groups — it's "good enough" even when it clearly isn't. The friction isn't the app, it's the moment you have to convince 12 people to install something new.
What's working: a 14-day free trial (no card required) for both Pro and Max plans, and the fact that players don't need to pay — that removes the biggest objection. But distribution is still the core challenge.
@jaroslaw_prazmo thanks for dropping this, Jarosław!
the "convince 12 people to install something new" problem is genuinely the hardest part of any group coordination tool, and you've already identified the right answer: don't make the players pay or install anything heavy. the captain bears the friction so nobody else has to. that asymmetry is what makes group adoption actually possible. distribution is still the grind, but the model is solid.
What you're building
TAM Network. the alternative to LinkedIn for everyone who builds and serves. coders, plumbers, nurses, designers. instead of titles and endorsements you publish receipts of work that customers and peers countersign. recruiters filter by 'who actually did this' instead of 'who has been at the company longest.'
Who it's for
the 92 percent of workers who do not write code, and the recruiters who need to verify them. solo founders and trades who have no portable credential they can carry between jobs.
The hardest part I'm solving right now
day-one cold-start for non-coders. an empty profile feels worse than no profile. early answer: a 'receipt builder' that turns one customer text message into a signed entry in under 30 seconds.
aug 12 launch. site: thetamnetwork.com
@thenameisarian appreciate you sharing this, Mustafa, we've talked about the credential layer before and it keeps coming up for good reason.
the cold-start problem for non-coders is the right thing to be obsessing over right now. an empty profile that mirrors LinkedIn feels like a step backward, so the receipt builder converting a customer text into a signed entry in under 30 seconds is exactly the kind of wedge that makes the first session feel worth it. looking forward to the Aug 12 launch.
What we're building
Nuvela AI is an AI-powered healthcare companion built by a registered nurse. It transforms complex medical documents into plain-English care guides with personalized explanations, medication insights, red flags, and questions to ask your provider, helping patients feel informed and confident, not overwhelmed.
Who it's for
Anyone trying to understand their healthcare: from patients and caregivers to parents and older adults. If you've ever left a doctor's office with paperwork you couldn't make sense of, Nuvela AI is for you.
The hardest problem we're solving
Making complex medical information both accurate and genuinely understandable while maintaining strict privacy and security. We're building an experience that feels less like using AI and more like having a knowledgeable nurse explain everything in a calm, human way.
🎉 Product Hunt Exclusive: Get 30% off the Basic Plan forever when you sign up through our Product Hunt launch.
Product Hunt Page: https://www.producthunt.com/products/nuvela-ai?utm_source=other&utm_medium=social
@nuvela thanks for sharing this, Samantha, really well framed.
the "built by a registered nurse" angle is probably your strongest trust signal and i hope it's front and center in your marketing. the accuracy + understandability tension you described is the exact thing most health AI tools get wrong, they optimize for one and tank the other. the fact that you're holding both at the same time is hard and worth talking about more openly.
@nolan_vu Thank you, I really appreciate that. That was exactly my goal from day one.
I never wanted Nuvela to just summarize documents. I wanted it to explain them in a way that's accurate, approachable, and genuinely useful without creating unnecessary fear or false reassurance.
That's a difficult balance, and it's something I think healthcare AI has to get right. Thanks for recognizing that, it means a lot coming from someone who picked up on that nuance.
@nuvela Thanks for coming back to this. The "accurate AND approachable without false reassurance" constraint is genuinely hard to hold, and most tools quietly drop one side of it to make the UX cleaner.
The fact that it came from a nurse background shows. That clinical instinct for what patients actually need to understand vs what just sounds thorough is not something you can prompt-engineer your way to.