PH热榜 | 2026-02-24

Google is quietly turning Stitch into something much bigger than a UI generator 👀

The new Hatter agent hints at multi-step, reasoning-driven design workflows, not just one-shot mockups. If this connects to their “Deep Design” system, we might be looking at AI that actually thinks through product structure, not just visuals.

What stood out to me?

• Auto-generated App Store assets (screenshots + descriptions + icon) 🔥

• Native MCP export directly into Cursor, Claude Code, Gemini CLI 🔌

That’s design → store-ready → dev environment in one flow.

Google is clearly building Stitch into a serious end-to-end product design engine.

Would you trust an agent to handle full multi-step UI flows or do you still prefer manual iteration?

very cool

Can you make an Design of the app here and import that design into Antigravity project to build full app? @Stitch by Google <3

Love this app. Have been using it and Lovable to prototype UI ideas.

Love seeing the speed at which new AI products are shipping. The challenge now isn’t finding tools — it’s evaluating them properly. Curious how you tested this against alternatives.

Design → App Store assets → dev environment in one flow is kind of wild. That removes so much friction between idea and launch. Feels like Google is playing the long game here 🔥

Would've been amazing if Stitch were around when I first had the idea for CoreSight!

congrats google 😎

Hey PH 👋

Aram and Eduard here - co-founders of Modelence. We previously built and scaled a startup where we kept solving the same infrastructure problems over and over: auth, database, APIs, cron jobs, deployments. So we built an open-source full-stack TypeScript + MongoDB framework to never do that again.

The problem: AI coding agents (Claude Code, Cursor, etc.) are great at writing application logic. But they constantly fail at backend setup from scratch - wiring up auth, database connection, infrastructure. That's because most platforms were designed for humans reading docs, not agents writing code.

We built Modelence from the ground up as an open-source framework for agentic development, unlike other app builders that simply use existing third party frameworks and platforms not designed for AI agents.

• Built-in guardrails for agents to catch and auto-correct errors before you deploy.

• Automated database schema and index management, so agents don't have to attempt to solve these tricky problems on their own every time. Agents focus on your product logic instead of burning tokens on boilerplate and failing at infrastructure setup.

• Cloud that actually closes the loop - persistent containers, dev environments, one-click deploy, and built-in observability around every operation. Because building is the easy part - running in production is where things actually get hard.

• And there is zero lock-in: you fully own all source code and data.

AI App Builder included - Type a prompt on modelence.com and get a working full-stack app. Then pull it locally, continue in your own IDE, and deploy back to Modelence Cloud.

What's coming next - A built-in DevOps agent that lives in your cloud environment, knows the framework end-to-end, and uses observability data to act on errors, alerts, and incidents automatically.

Our bet is that the real challenge in AI coding isn't the builder tool - it's the framework and platform underneath. If your agent has a solid foundation, it ships real apps. If it doesn't, it generates impressive demos that break in production.

Modelence is open-source - tell us what's missing, what you'd want to see, and how your workflow actually looks - we want to hear it.

Try it now: modelence.com 🚀

Congrats, team! I’ve seen agents generate great logic but completely struggle with auth, DB wiring, and deployment setup. If Modelence truly abstracts that reliably (with guardrails + production-ready infra), that’s a big change!

Going to test it out 👀

Really nice! how do you handle context drift for web scale projects?

Congrats on the launch. Who is Modelence built for? Are you targeting devs and builders or largely non-coders?

How does it differ from something like @Solid?

Congrats on the launch, @artahian! How does the auto-fixing of the DevOps agent look like? Curious to know

Congratulations. What is the DevOps agent and how will it work?

What does it specialise for? Websites, web app, mobile app or chrome extensions?

Combination of @Anima App + @Modelence App Builder + @Stitch by Google would be so powerful!! All three launched today :D

Looking forward to giving it a go myself. Congrats, guys! ~James

Everyone’s talking about AI building apps faster, but almost no one talks about what happens after the demo works. 

Who do you see as your core user right now? Indie hackers experimenting with AI agents, or teams shipping production products?

Nice concept! How does Modelence handle more complex cases like custom middleware, background jobs, or multi-tenant setups? Is it easy to go beyond the happy path?

Congrats on the launch @artahian . Product looking sharp.

Congrats! Wondering how this stacks up vs Replit, Lovable or other app builders. What's your take?

Smooth experience with auth and DB pre-wired. On a first quick test, it feels similar to others. Real edge (production-readiness, observability) is invisible until later. Worth making that visible earlier :)

This actually hits a real pain point. AI agents are amazing at logic, but backend setup is where things quietly fall apart. Love that you're focusing on the foundation instead of just another shiny builder. Solid thinking 👏

Congratulations, looks like a great platform! It was about time somebody developed something that is more turnkey than the rest and more AI based. I like the fact that its open source as well, such that it can keep developing.

damn! nice!

The framing of 'framework designed for AI agents' is the most interesting angle here — most infra tools were built for humans reading docs, then retrofitted for agentic use. The production gap (auth + DB + deployment working day 1) is exactly where Claude Code and Cursor agents burn most of their retry loops. Curious whether the open-source framework is usable standalone without Modelence Cloud, or if the cloud is load-bearing for the production guarantees.

Hey PH 👋
Avishay here, co-founder of Anima.

We built Anima the UX design agent because we hit the same wall everyone hits with vibe-coding: coding agents ship fast, but they’re design-blind. You get inconsistent UI, broken patterns, and the “AI slop”.

What Anima does: it’s a design-first playground where you can go from ideas to runnable, on-brand code, with your brand + design system in the loop. 

Start from a prompt, clone your site, paste Figma links, import a design system, then iterate with the agent. The output is real code you can hand off to engineering or to other agents via MCP. 

What makes Anima different:

1. Design-first, not design-blind - Anima understands your brand and design system as context.

2. Works with your existing stack - Figma, design systems, and coding agents via MCP.

3. Real code, not just prototypes - Runnable, connected to data, and ready for handoff/publish.

4. Built for teams - Shared workspaces, shared context, consistent output.
   

Who is it for:

Product teams building with AI who are tired of inconsistent UI and design-handoff friction. Designers who want to ideate faster without breaking the brand. Engineers who want clean code from day one.

Anima has been leading the design-to-code space for 8 years, and is used by teams at Amazon, Samsung, Apple, Disney, Deloitte, and more. This launch is the next chapter: design as code, with AI that actually gets it.

We’ll be here all day. Ask anything, share your feedback, questions, and hot takes 👇

- Anima team

Amazing work team! can you explain more how I can take it into Claude code when I'm done?

For publishing, do you allow adding a custom domain?

Can you explain more about the design system integration?

The biggest friction in AI-built products is that weird gap between prototype and production. If Anima really delivers runnable, on-brand code that engineers don’t have to clean up, that’s a big deal 🔥

Loved it team! Design made easy and friendly. Wish you all the best here!

We need to either feed the figma file or we can start with a prompt, yes?

Interesting launches today. Combination of @Anima App + @Modelence App Builder + @Stitch by Google would be so powerful!!

this is just gold!

Great tool to clone and evolve nice websites design! Just curious - could it generate high end design for banner/posters in future too?

Congrats on launching. How is this different from just using Figma + a coding agent like Cursor or Claude?

This reminds me a bit of the direction many AI tools are going — specialized workflows instead of general assistants. Interested to see how this evolves compared to the larger models.

Loved the experience! We are actively using this in AutoSend 🚀

I love the idea of this -- having troubles connecting though.

I’ve been analyzing a lot of AI tools recently and the biggest thing I notice is people struggle to compare them properly. Curious how you position this vs the current options.

Orchid's is great. It really helped me save time when I wanted to mimic the design of a website I liked a while back. Cool to see them on PH!

Congrats on the launch guys! Looking forward to trying this out. Question for you: Does it support OpenRouter API keys and if so, how reliably does it work?

Congrats on your launch!

I’ve been using your app for ~2-3 months, and over the last month I’ve been really disappointed last month.

Example: 50 minutes ago I started a task (Opus 4.6) and the agent froze. It stayed “running” with no progress or output. After I reloaded the page, it resumed for ~20 minutes, used ~95k tokens, then froze again. When I stopped the task and started it again, it began from scratch. Over the last month, my agent consumed ~3M tokens in these loops with no expected output.

Support hasn’t replied at all.

I really wanted this to work (I like the design), but right now it feels unreliable.

Can you share when you plan to fix these issues (support responsiveness, freezing, and token burn with no results)?

Congrats, Kevin!

I have two questions regarding the long-term side of things:
1. Vendor Lock-in: If I build a project with Orchids, can I export the full source code to host it independently? It’s crucial to know if we truly own what we build.
2. Maintenance: How does Orchids handle updates months later? Does the AI maintain the full context of the existing codebase to ensure that adding new features doesn't break old ones as the app grows?

Thanks for the insights!

Awesome! Great work guys... All the best!

@kevinlu12 Great that you can integrate ChatGPT, Claude, Copilot, and more how do you ensure smooth performance when combining multiple AI APIs?

Any feedback on Liner Write would be greatly appreciated ❤️.

Congrats on the launch! Super sleek product — love how it extends the Liner experience. I’m building a similar editor in Korea too.

Does it work for copywriting? Blog-post, for example?

The Cursor analogy is apt but I'm curious how it translates to writing in practice. With Cursor, the interaction model is clear: tab-complete, inline diff, accept/reject. For code that works because changes are structural and self-contained.                            

Professional writing is messier. If I'm halfway through a 10-page proposal and I ask the agent to "make the competitive analysis more specific," how does it handle that? Does it rewrite the section in place with a diff view? Suggest edits inline? Or regenerate and let me compare?

The 95% SimpleQA accuracy is impressive for factual grounding, but the co-writing UX is what would keep me using it over just prompting Claude and pasting into Google Docs. Would love to hear more about how the agent interaction actually feels mid-document.

The "documents are the new source of truth" framing really clicked for me — I've been thinking about this a lot in the context of writing vendor proposals and internal strategy docs, where a single hallucinated stat can quietly tank credibility. Curious how Liner Write handles source attribution inline — like, can you actually trace a claim back to where it came from without breaking your writing flow? That transparency layer feels like it could be the thing that makes this trustworthy enough for high-stakes professional use.

Cool product, Jinu.

Curious: Does it work cross-device?

I use different devices (phone, laptop, tablet), so it would be cool if my context followed where I went.

keep making such products like these!

just interested, is there any information on age?

This is seriously impressive work. I've been meaning to dig into YC trends like this for a while.

If there's any way I can contribute or help, I'd love to be involved. I'm especially interested in understanding underlying YC patterns more deeply.

Great analysis for YC portfolios! now question is here - how could my start up successfully break into YC? lol

It is interesting to note that almost 70% of the teams are only two-person.

Hey nice work, interesting insights! What is your definition of "wrapper"?

Cool, I will be sharing this with the team, they are trying raise money, some stats will be helpful for them. Thank you for sharing :)

Amazingly helpful tool for me while making the first steps with CoreSight. Congrats on the launch and the product!

Thanks :) I've been wanting an analysis like this, but never got around to doing it.

And a sidenote, I'm not surprised education is at the bottom. Over the past five years, AI has been killing informal education platforms; Udemy, Coursera, Thinkific, etc., have all lost ~85% of their stock price. I've heard universities have been struggling as well.

Hey PH 👋 — Joe (@rolodexter), Aleks, and Matt (@matias_papaleo) here, and we made ToggleX.

We built this because we kept watching @OpenClaw agents give stale, hallucinated recaps of actual work. The agent wasn't dumb — it was blind. No visibility into what had actually happened across sessions.

ToggleX fixes that. It streams real browser sessions into your agent as structured context — projects, focus scores, context switches. Your Claw stops guessing and starts knowing.

The moment it clicked for us: the agent came back with "594 minutes tracked, 519 min work, 75 min breaks — here's exactly what I can't confirm without the API key being set." It knew its own data limits. That's when we knew this was real.

Happy to answer anything — especially from OpenClaw users already running always-on setups. That's exactly who we built this for.

PS One more thing worth mentioning — ToggleX is built by GLIK AI, a privacy-first AI research lab. We're SOC 2 Type 2 in progress, working with Vanta for continuous compliance monitoring. Real-time security posture publicly visible here: https://app.vanta.com/rivalz.ai/trust/8ye1fzqonaxfgotf2l6xr2

you guys will crush it

What's the lag behind data consumption by the extension and delivery to the agent?

This is awesome, how well does it understand what i do during the day?

Cool idea but I’m worried about the privacy aspect. What do you do for privacy concerns? Who owns my data?

I was a bit skeptical going into this but it generated a pretty nice vid for Product Hunt!

https://imgur.com/a/product-hunt-brand-vidoe-created-with-bazaar-gsqAnAU

Hey Hunters, Jack here - co-founder of Bazaar.it

We built Bazaar because we needed an easy way to create launch videos for the apps we were building.

You shouldn’t need to spend thousands or wait weeks for a motion designer just to explain what your product does. AI should handle that.

So we built an AI video generator focused on creating accurate software product launch videos — recreating and animating your actual website or app.

Would love to hear what you think and see what you create.

Congrats on the launch! This is a huge pain point for someone like myself who is more of a technical founder but needs to create content.

Did you create your promo video (slide one) with the product?

I've been using Bazaar for my micro Saas startup and freakin love it. Videos that used to take days of tedious editing are now done in minutes, and sometimes they are even better than what I can produce manually. I have a million things to do for my business and the last thing I want is to be burning hours to make crappy motion graphics. Bazaar is saving me tons of time that I can put onto more important things and it helps keep the momentum going instead of hitting the wall and getting frustrated on a long edit.

I met Jack and Markus randomly on a beach in the Philippines and they are definitely visionaries living in the future. I think they have created something really special with Bazaar and I can't wait to see where it goes. Incredible work guys!

Did you create your demo video with Bazaar? It looks really amazing !!

Hey, I'm Markus. CTO of bazaar.

first off, I wanna upvote the bazaar community. 8,000 people making videos, breaking things, telling us what sucks, and somehow sticking around anyway. you guys made 30,000+ videos on this thing. v4 wouldn't exist without that.

So what is v4? We took everything we learned from those 30,000 videos and built a new agent around it.

Paste a link, upload a screenshot of your UI, and just tell it what you want. product demos, animations, explainers, whatever.

multi-track timeline, voiceover, the whole thing. it actually works now!

Tons of people are building in this space right now, But we think Bazaar is the best way to go from idea to finished motion graphics. 9 months of community feedback baked into everything.

oh, and the community asked for an API, so we shipped a CLI. Agents are already making videos with it. But we'll save that story for the next launch...

Try it. roast it. tell us what you love or hate. everyone gets $4 to test it our before you hit the paywall.

We're here all day. (we're actually in the philippines right now.. timezone is rough but the cappuccinos are cheap so)

keep making great products like these!

Markus and Jack you guys are incredible! V4 is an insane step up, this upgrade is next level. Awesome work guys! 🚀

Hey Product Hunt! 👋

Jason here, co-founder and CEO of Collective OS - the AI-powered matchmaking platform that helps agencies, consulting firms and freelancers generate new deal flow through strategic partnerships.

The backstory: My co-founder and I spent years in the agency and consulting world watching the same problem play out. Referrals had become so core to our businesses, yet there was no platform that let me easily build new referral partners, or get the most from my existing ones.

Meanwhile, the old playbook for growing a services business (cold outreach & paid ads) was getting less effective by the day.

There had to be a better way.

For the last 2 years, we worked hand in hand with hundreds of users and prospective users to build a different approach - one that felt more human, and that helped me maximize the things that were actually working - referrals.

Enter... Collective OS - the AI referral network for agencies and consultants to match with complementary partners to co-sell and refer each other into their respective clients.

Our AI analyzes each firm and consultant's expertise, industry, past clients, case studies and growth goals, then surfaces complementary, vetted partners you'd never find on your own.

In the last year alone, we've already facilitated $5M+ in shared deal flow across our early network, along with thousands of matches.

Our users have told us that they love the platform because it gives them the ability to "say yes to more revenue" without the overhead of hiring or becoming full-service themselves.

Award winning independent agencies like Laughlin Constable, Fuse Create, Moroch, Power Digital, Tinuiti, and Stoltz Marketing Group are already on the platform partnering in ways they couldn't before. Not to mention another thousand firms and consultants from around the world who are discovering new partners every day.

It's completely free to sign up and start getting matched. No credit card, no catch. For firms that want to go deeper, our Pro tier unlocks better AI, more matches, more opportunities, unlimited messaging, and search.

🎁 For the PH community: Use code PRODUCTHUNT at signup to get a free month of Pro on us.

We built this for the independent firm that's tired of outdated channels to drive new business, or leaving money on the table because they didn't have the right partner at the right time.

If that sounds like you, we'd love for you to check it out.

We're a small but mighty team, and we take our user feedback very seriously. Anything you share with us will get added to the roadmap, so we want to hear from you.

Drop your thoughts below 👇

Love the idea!

The promo code doesn't work for a free month, unfortunately. Looking forward to checking this out!

Such a smart concept to shift the way agencies like mine identify verified partners who can fill our service gaps!

Great app. thanks a lot.

I’ve been one of the first customers and still am. It’s a fantastic product and team. Congrats @jason_flack1!

Just tried it out!

I enjoy reading long form articles! This is definitely something I see using often.

nice product mate

love the idea bro

Solve this easily on mobile browsers and you're onto a winner.

Good stuff!!

Did I do anything wrong? Because I can see the image.

keep crushing it guys!

@maiquangtuan Love this direction 👏

The “just give me the skill without building it” insight is very real. Most non-technical users don’t want flexibility — they want leverage.

Curious how you’re thinking about quality control as the marketplace grows? Do you see this becoming more curated (like an App Store) or more open (like a plugin ecosystem)?

Congrats on the launch 🚀

Looks great! Have you guys considered evaluating your skills? You can look into this with tessl.io/registry/skills/submit keen to hear if you find this helpful, and whether you want to add this as part of your review process with your community

Awesome work! @maiquangtuan

Hey Product Hunt! I'm Maxi, co-founder of Falconer. Dave and I met at Uber, where we spent years wrestling with a problem every fast-moving eng team knows too well: the knowledge lives in your codebase, your Slack threads, your teammate's head—but never in the same place at the right time.

You know the drill. Someone asks "why did we build it this way?" and the answer is somewhere buried in a six-month-old PR comment, a Slack thread with 47 replies, and a doc that three people know exists. You spend 30 minutes archaeologizing your own company's decisions before you can write a single line of code.

We tried everything. Wikis that went stale before anyone could find them. Expensive tools that quickly became digital graveyards. READMEs that lied to our faces. None of it worked because none of it was connected to where the actual work happens.

So we ditched it all and built something that actually works. Falconer is a memory layer for fast-moving engineering teams. It connects to your GitHub and Slack, watches where decisions actually get made, and turns that into living documentation that updates itself. The context you need is just there, and you don't even need to leave Slack to get it.

Every decision, every bit of context, every "why did we do it this way"—captured, connected, and current. Without anyone having to maintain it.

It's the thing your team's most talented engineer knows off the top of their head—except now everyone has access to it.

We built Falconer because we kept losing entire afternoons to problems that had already been solved, and somehow every bloated tool we tried just added more noise instead of fixing it. We think Falconer actually nails it—and we want you to think so too.

Give it a try, tell us what clicks, tell us what's missing. All feedback is fair game.

Congrats Dave and team, excited about your launch! Seems like you're the perfect team to be working on this!

Before joining Falconer, I spent several years at large tech companies. I loved writing code and building products for millions, even billions, of users. But writing and maintaining internal documentation was always a pain point for me and the teams I led.

A few reasons why it felt so painful:

• Keeping docs up to date in a fast paced organization is incredibly hard. They become outdated almost as soon as something changes.

• It often feels redundant to write documentation for other engineers, such as API docs for internal teams, after already investing so much time building and validating the code.

• Organizing documentation is also tricky. Finding the right or related docs for a feature can be frustrating. Over time, doc debt accumulates, and cleaning it up requires precious engineering time.

I believe many engineers feel the same way. We care about documentation, but keeping it accurate and well structured is much harder than it should be. And that is why we built Falconer. We wanted to remove this pain for ourselves and for other builders who share the same frustration.

I'm on the ops and marketing side at Falconer, so I'm not the deep in the codebase—but I am the person who needs to know exactly how things work at any given moment. I'm constantly pulling from our knowledge base and taking that information as truth to do my job.

Here's what using Falconer looks like in my actual workflow:

• Customer conversations: I ask @falconer in Slack "how does SSO work?" and get a cited answer I can trust for a sales call with out interrupting engineers.

• Product demos: I search across code, docs, and Linear tasks in one place to pull accurate feature descriptions and workflows.

• Internal decisions: I can find prior decisions and understand why they went that way, instead of hoping the right person remembers.

I've been using Falconer for a few months now, and I love it. I've used it to help draft blog posts, create bespoke onboarding plans using my internal docs and a job description, create docs for a CLI tool just from reading the code, and more. It's become a tool I use every day, every way.

Well done! Excited for the world to start using Falconer. If you care about moving fast without losing context, it's time to give it a shot!

Before building Falconer, I’d gone through multiple onboardings where understanding even a single service meant jumping between Slack threads, outdated docs, PRs, and chasing people for answers. Important context was buried in DMs, system design was hard to track down, and docs started falling behind as soon as the code changed.

That frustration is a big part of why I joined this team. We wanted to build something that stays connected to how teams actually work: pulling context from code, projects, and conversations, so documentation and diagrams don’t become another thing engineers have to manually babysit.

It’s been incredibly rewarding to help build a product that turns scattered knowledge into something living and searchable. Grateful to everyone trying Falconer today— would love feedback from fellow builders.

What's the core value proposition here?

Do you see this expanding to different teams or areas that need documentation? Ie sales rules of engagement etc?

This is wild. I made a shirataki noodle co in a few minutes... or at least got started. :)

The live inbox transparency is a bold move — curious how Polsia handles situations where the AI makes a commitment on your behalf that you'd want to walk back. That edge case feels like where the "runs while you sleep" pitch either earns total trust or breaks it completely.

Hey Product Hunt! I'm Mauro, an independent developer from Pergamino, Argentina.

I created Lemonade because all the password managers I tried seemed too complex or didn't fit my workflow. I wanted a single place for my passwords, .env files, API keys, and secrets, without having to learn a new system or install multiple apps.

So I created it. It took a long time to perfect, but I'm proud of how it turned out: AES-256-GCM encryption, access keys, TOTP with QR scanner, emergency access, extensions in the Chrome and Firefox stores, and the Env Vault that I use daily.

You can understand exactly how it works here: https://docs.lemonadepass.app/

The free version is generous; I want you to try it without any pressure.

It's intentionally simple. I'd love to hear what you think.

you guys will nail it

Hey Product Hunt! I'm Paul, co-founder and CEO of Quilt.

I started Quilt on paternity leave with my second son. I'd been a partner at Google's Area 120 for a few years and knew I had another company in me but was struggling. I made a rule that I would only start something if I would commit at least a decade to it. Tech moves so fast – how can you get confidence you're betting on the right thing, that you won't spend years attached to some wrong bet? I witnessed this going wrong first hand early in my career. It scared me.

The solution sounds quite simple, but took me years to find and was personally very significant: If I worked on whatever problem I genuinely believed was the most important to my kids' generation, there was no way to fail. The worst case? You spend a decade trying to make the world better for them. Zero regrets. The best case? You help solve the defining problem of our time, because by definition that's what you're working on.

It didn't take long to land on climate — and specifically on homes, which account for 20% of U.S. emissions, mostly from burning fossil fuels for heat. Heat pumps are genuinely magical technology — 400% more efficient than gas furnaces — but the products on the market were bulky, confusing, and stressful to adopt. That's not a technology problem. It's a product problem.

So I did what any reasonable person would do: I went and got a heat pump for my own house. And it was a disaster. Multiple contractors just to find someone willing to sell me one. The unit was huge and ugly, controlled by a remote from 1985. No app. No intelligence. No joy.

I grew up in Alberta, Canada — the heart of the fossil fuel industry — and spent my career in consumer tech. So the question kept nagging: why hasn't anyone applied what we know about building great products to the single biggest emissions lever in the home?

My co-founders Matt and Bill (ex-Google, Nest, Google X) and I started building in 2022. At the time we wrote a fake Wirecutter review (Amazon style) to capture the vision (and just two weeks ago the real one came out!).

Here's what Quilt is today:

• A deeply integrated SW/HW home climate system with industry-leading efficiency (SEER2 25, HSPF2 12) — 20% more efficient than the next best ductless system on the market

• Indoor units only 7⅞ in. tall — they fit above doors and windows where traditional mini-splits can't

• Room-by-room control through the Quilt Dial and App, with occupancy detection that automatically saves energy in empty rooms

• 500x more processing power than traditional HVAC. Last fall, we shipped an OTA update that boosted heating capacity 23% and cooling capacity 22% to every installed system. No service visit. No new hardware. No cost.

• Now in 21 U.S. states and 6 Canadian provinces — and we're just getting started

For the PH crowd: think of Quilt as what happens when you build HVAC the way you'd build a great consumer tech product — beautiful hardware, intelligent software, continuous improvement. We're building what we hope is the Model S moment for home climate — something people are genuinely excited to bring into their lives because it's a better product, full stop.

As Quilt has gained momentum, the decade commitment now seems much too short. Something I learned a year in: the youngest of the major HVAC manufacturers is over 100 years old. 100! (Trivia: Rheem, founded 1925.) Like so many other industries being reinvented by technology, HVAC is overdue — and the opportunity is massive. Humans spend 90% of our lives indoors. These systems (air, water, heat, cooling) are what make buildings livable. There is so, so much to do here when viewed through that lens.

Happy to go deep on anything — the product, the tech, the HVAC industry, or the founding journey. And if you're curious about upgrading your own home, check out quilt.com.

— Paul

Quilt is a Silicon Valley startup, founded by a handful of former Google employees (and alum from Nest, Tesla, and Apple), is building "the smartest and most efficient heat pump on the market". They announced a $20M raise last December.

Wirecutter says, "The Quilt heat pump’s handsome fit and finish is an obvious departure from the mini-split status quo."

As a homeowner with solar, I've become interested in heat pumps as my next home upgrade. Energy is one of my biggest expenses, and improving my home's efficiency would unlock great savings.

Here's how Quilt compares:

With Auto-Away automatically adjusting empty rooms, real-time energy usage tracking, and over-the-air updates that make your system smarter over time, Quilt is at least:

• 20% more efficient than the next-best ductless heat pumps available

• 80% more efficient at cooling than what most North Americans have in their homes today

• 400% more efficient at heating than what most North Americans have in their homes today

nice!

Been a pleasure building this! Excited to have you guys try it. Would love any feedback!

This is definitely one of the most exciting product launches recently, I can't wait to use this!

Love the product

nice product mate!

Would be nice if you can set "avatars" for people that inspire you in life. Have a morning wakeup discussion with Messi on why you should get up CHAMP)

Love this so much!! <3

Haha, I love this. So creative, and I can't believe it hasn't been done before! How effective has it been for you? Any way to load up custom voices?

It might be hard to read the tea leaves on this one (so goes developer-focused launches), but the fact that WebMCP is being advanced by both Google and Microsoft is a big deal, and even moreso now that Google is shipping it in Chrome in early preview:

Developers can test WebMCP by enabling the “WebMCP for testing” flag in Chrome 146 Canary at chrome://flags. Google’s Chrome Early Preview Program offers access to documentation, demos, and updates on API changes. With working code already shipping, W3C institutional support, and Microsoft’s co-authorship, WebMCP has cleared the most difficult hurdle any web standard faces: moving from proposal to production software. The foundation is now in place for websites to speak the same structured language that AI agents understand natively.

Expect to see this more widespread by Google I/O in June.

This fills a real gap. Most agent tooling assumes APIs exist, but a huge chunk of useful functionality is trapped behind web UIs with no API. The collaborative model where user and agent share the same interface is the right mental model too - agent as co-pilot, not separate automation. What is the approach for handling auth state across agent sessions? Cookie persistence or re-auth per request?