in 
p/windsurf
p/windsurfWindsurf drops credit-based pricing for "industry standard" quotas; users revolt
Chris Messina:@mazula95 it does beg the question — what hope does Windsurf have to stay competitive?
in 
p/self-promotion
p/self-promotionEatSpoiler – Fridge and Expiry Tracker
eatspoiler:Feedback means a lot to me guys, thanks
in 
p/general
p/generalHow do you decide what features should be free and what should be paid?
Michael Beckett:Great question, Nika. I think about this a lot because I'm building in this exact space. I built Signbee — an e-signature API designed for AI agents. So literally my product lets agents handle legally-binding documents. Which means I've had to sit with this question deeply: where does the trust boundary actually sit? My take: trust the agent with execution, never with authority. What I mean is — an AI agent can draft a contract, send it to a signer, and track the status. That's execution. But the human still has to read and sign it. That's authority. The agent never holds the pen. That distinction matters. Same principle applies to your examples: Finances: I'd let an agent prepare an invoice or flag an anomaly. I wouldn't let it approve a transfer. (The Sapiom news is interesting but notice — it's agents buying tools, not making investment decisions.) Health: Let an agent summarise your lab results. Don't let it decide your treatment. Communication: Let an agent draft a message. Don't let it send without review. At least not yet. The pattern is: AI handles the workflow, humans hold the checkpoints. That's actually why I built Signbee the way I did — the agent can orchestrate the entire signing flow via API, but there's always a human at the signature step. The protocol itself enforces the trust boundary. As for the Sapiom raise — I think it's less about "trusting agents with money" and more about infrastructure catching up. Stripe just launched MPP (Machine Payments Protocol) this week. Cal.com has Cal.ai for scheduling. AgentMail handles email. We handle signatures. The stack is forming, and each layer has its own trust model baked in. The real question isn't if we'll trust agents — it's whether the tools they use are built with the right guardrails. That's on us as builders.
in p/self-promotion
p/self-promotionPitch your product with max 5 words. Can you?
Michael Beckett:One API call. Document signed @Signbee
in 
p/introduce-yourself
p/introduce-yourselfNewbie to ProductHunt
Martyn Johnson:@sai_tharun_kakirala Appreciate that, Sai! yeah it definitely feels like a lot of us are going through a similar shift at the moment. Right now I’m working on something around turning ideas into actual structured plans. That gap between “I know what I want” and “what’s the best way to actually do it”. Still early days, but it’s been interesting seeing how differently you can approach building now compared to even a year ago.
in p/introduce-yourself
p/introduce-yourselfNewbie to ProductHunt
Martyn Johnson:@dr_simon_wallace Haha nice, Geocities takes me back. My night builds now are basically short loops... idea/concept > prompt > prompt > working feature > test > iterate. I’m shipping a lot smaller pieces, a lot faster than I used to. Less “sit and code for hours”, more “get something working and refine it”. Feels like a completely different way of building compared to a year ago.
in p/general
p/generalHow do you decide what features should be free and what should be paid?
Wambugu Gichuki:I think the decision is actually a simple one, and it all depends on if you can afford to self fund the free service out of pocket, which if you can then go ahead. if you can't then monetize from day one and only offer the most minimal amount of freebies to get users to convert to paid. ive learnt this from my own product which i monetized from day one and tbh, you'd be surprised that people are willing to pay if it solves a true pain point and actually adds value to them.
in p/windsurf
p/windsurfWindsurf drops credit-based pricing for "industry standard" quotas; users revolt
M. Aziz Ulak:I think after Cursor’s Composer2, it’s becoming increasingly difficult to compete—their benchmarks look very promising. If those numbers translate to real-world performance, Windsurf will need a true breakthrough to keep up.
in p/general
p/generalTo hard paywall or not — that is the question!
Chris Messina:@umairnadeem Have you ever tried a peanut butter and apple sandwich though? 🤤
in 
p/vibecoding
p/vibecodingBuilding SaaS in 2026: Are you vibecoding your own product or engineering it the "old way"?
Mykola Kondratiuk:I have 5 shipped products and I still wrestle with this on every new one. My current take: vibecode the UI and happy-path flows, engineer the parts that are hard to change later (auth, data model, billing). The test suite can wait, the schema really can not. The over-engineering trap is real. I spent 3 weeks setting up a "proper" backend for a tool that got 12 users and died. The next one I vibed in 3 days and it has actual paying customers. The difference wasn not quality - it was whether I got real signal before investing. The one thing I refuse to cut corners on even for MVP: how data is stored. UI is replaceable. A messy schema with years of production data on top of it is not.