PH热榜 | 2026-03-04

Hey PH 👋

We're the team behind Notte, a browser automation platform, and today we're launching Anything API!

Most websites don't have public APIs. If you want the data or the workflow, someone has to build and maintain the automation. Anything API handles that for you.

Describe what you need done in the browser, our agent uses the Notte CLI to execute it, build the workflow, and ship a callable function endpoint. Deploy it serverless, schedule on Cron, or call it via API. No setup or infra to manage.

Who it's for:

🔁 Teams running the same browser workflows repeatedly

📦 Developers who need data from sites with no public API

⚙️ Anyone automating anything, from document processing to form submissions to data extraction

Try it free at: anything.notte.cc

We'd love to know - what site or workflow would you point this at?

This is pretty cool. Btw when the agent generates the API endpoint, how deterministic is the execution layer? Is the AI only used during the build phase and the runtime stays fully scripted?

Super cool product name, nails it right away!

Good luck for the launch 🙌

🚀

Great launch—this is exactly the gap most GTM teams hit when they need reliable data from tools that don’t expose APIs. Quick follow-up from my earlier thread: if I need session persistence + proxy rotation for long-running enrichment jobs, can Anything API pin identity state per endpoint (so retries don’t break auth flows)?

Great great idea aaaand mint UI as always! 🏆

Kind of amazing that you can generate an API like this. Does it suffer from the same problems as web scrapers if the DOM should change dynamically? Since it's an API I'm supposing that actually it could protect as a layer of abstraction above the changes on the page and then you just get it to self heal on a recurring basis? Perhaps you could even charge for that self healing cadence.

Love the update and congrats on the launch, @samatnotte!

Huge congrats on the launch, Andrea and the Notte team!

As someone building in the fintech space, the 'API gap' for regulated platforms is one of the biggest bottlenecks to true autonomy. I love the approach of bridging agentic exploration with deterministic execution to create APIs where they don't exist.

Great product guys! Congrats on the launch

This solves a real pain — so many useful sites still have no API. Turning browser workflows into callable endpoints could save teams a lot of brittle scraping work. Curious how it handles sites that change their UI often.

That sounds fantastic! I'll take a look at it.

Great work! Excited for the future!

Hey PH 👋, here are a few examples of use cases you could try with anything.notte.cc 🚀

→ Pulling competitor pricing daily from sites with no public API
→ Scraping real estate listings with filters no public API supports
→ Monitoring regulatory filings (SEC, FDA, etc.) and alerting on new submissions

→ Extracting structured data from portals locked behind logins

If you've got any idea, drop it below, we might just build the endpoint live 🔥

This is so cool!

I didn’t even realise this was possible but I can confirm after heavy usage, it very much is.

Really cool, highly recommend everyone gives it a try:)

Hey launch team, congrats on the launch!

This is a great idea for creating quick production grade API's. I can't wait to try it out, the interface looks very clean and easy to use as well. I particularly like the "what it does" section but would it perhaps be a bit more helpful to have a hovering "alt-text" when a user hovers over a function / curl command for more ease of use to explain what each element does, this would also help users be more confident in reducing bloatware.

Hunter @garrytan coming in clutch with these new product launches! XD

What about bypassing bot blocking? For example, if I need to get content from an e-commerce site and they don’t have an API. Such sites usually ban bots. Do you bypass that?

This is solving a real pain point. Building API integrations from scratch for every data source is exhausting — I've been through it connecting YouTube, Stripe, and multiple AI providers for my SaaS. How are you handling rate limits and auth differences across platforms? That's always been the hardest part for us.

Looks awesome, great work @ogandreakiro and team!

This sounds like the real "build for agents" tool. Will check it out!

I just used this for my platform One Pager and I gotta say wow!

This is much more than a standard chatbot.

One Pager rapidly creates websites for outreach, but does not have a public API at the moment. I entered my website, then anything API created an account, logged in starting understanding my platform and created an API to build website using my data structure.

This is something I haven't seen before and really impressed me.

Curios if there is a way to output the endpoints created into a document for external use?

@samatnotte @ogandreakiro How do you ensure the API delivers reliable and structured data, especially from messy or complex websites?

Intriguing. For teams running repetitive workflows, how customizable are the automations in terms of logic, triggers, and specific conditions?

Lets say this tool is applied to a site like Airbnb , how can it avoid detection or being flagged?

Very cool. A few of the sources we pull from don’t expose APIs, so we’ve had to rely on makeshift solutions. Can see this being super useful!

The 'fully editable layers' angle is what sets this apart, every other AI design tool gives you a pretty image you can't touch. Kodo actually respects the designer's workflow.

As a founder prepping launch assets for my own product (Fillix - a job application Chrome extension), I've been manually tweaking Canva exports forever. This could genuinely replace that. Adding to my stack. Congrats on the launch!

Hey @mg272011 , I tried Kodo and I love the fact that you can generate layers. Great work! As a suggestion: I think the elements design would be better on the left, like Canva, so it's less out of the way of the design itself

Hey Michael, congrats on your launch! How do you plan to compete with big players that have sort of unlimited budgets and teams and offer similar tools, for example Nano Banana?
P.S- I find the editing bar on top of the image convenient and also the UI looks familiar to what we are used to use on a daily basis which is a good thing

This is really cool! Btw what's happening under the hood when it generates the design structure? Is it creating something closer to Figma-style layers/components, or more like editable SVG groups?

Just tried Kodo. bye bye Canva

Congrats on the launch, Michael! The biggest pain point with current AI tools is definitely that 'wall' between generation and editing. Being able to manipulate layers and real text is a total game-changer for those of us who need speed without losing control. Do you have any plans to add direct export to Figma

Generating designs with AI is great, but having them come out fully editable instead of flat images is a huge win. What kinds of designs does Kodo handle best right now things like social posts, slides, or more complex layouts?

Is it possible to create a design for a website and import it into Figma?

Congrats on your launch Micheal. I just tried out the tool.

Let’s go!

Brilliant! I've been looking for a UI that I can use to just edit directly rather than type it out on a chat interface

Thanks for building a platform that can actually generate a structured design (plus, it's editable). Congrats on the launch@mg272011!

Cheers on shipping a cool product. Best of luck 👍🏻

To have built a platform like Kodo is a huge achievement in and of itself!

To have done it as 14 is beyond me.

You will go on to do great things Michael, well done!!

lets go michael!

When someone changes text length or moves elements around how does Kodo keep the spacing and alignment from breaking?

I’m a developer and CEO, and this product started from my own frustration. Over the years, I’ve used countless coding tools that only react after something breaks — after the bug appears, after performance drops, after architecture gets messy. But real development doesn’t work like that. When we code, we’re constantly thinking ahead. We anticipate problems. We refactor before things collapse. I kept asking: why can’t our tools think that way too? That question led us to build Enia Code.

There are already many AI coding tools — copilots, editors, chat-based assistants. Most of them wait for prompts. Enia is different. It’s proactive. It detects bugs, performance risks, architectural inconsistencies, and refactoring opportunities as you write. No constant prompting. No re-explaining context. No switching tabs. It works quietly inside your IDE, adapting to your coding habits and team standards over time. The goal isn’t to replace developers — it’s to protect their flow.

We believe coding tools are evolving from reactive copilots to proactive agents. The next step isn’t just faster autocomplete — it’s intelligent systems that anticipate, learn, and grow with your project. Software complexity is increasing, solo developers are building bigger systems than ever, and “flow” is becoming the most valuable resource. The future of AI coding isn’t about answering questions — it’s about preventing the need to ask them in the first place.

If you have any thoughts, ideas, or feedback, I’d truly love to hear them — feel free to drop a comment and let’s discuss.

Follow Enia Code on X and YouTube:

• https://x.com/EniaCode

• https://youtube.com/@EniaCode

I’ve been building SaaS tools for the past few years, and one thing that always slows me down is discovering architectural problems way too late. Refactoring a working system is painful. The idea of a tool that signals these issues earlier sounds really valuable. Curious whether Enia detects these patterns based on past projects or just the current repo context.

Would love to see some examples of issues Enia caught proactively. Real-world scenarios would help illustrate the value.

This reminds me a bit of static analysis tools, but with AI reasoning layered on top. Traditional linters catch syntax or style issues, but they don't really help with deeper design problems. Interested to know what types of problems Enia is best at detecting.

Interesting concept. Most AI coding tools today are essentially prompt-response systems. You ask → it answers. A proactive model feels more like an actual collaborator. Curious how Enia decides when to surface a suggestion.

Congrats on this launch!

Unable to download via your website.

Why should I use this over something like Claude Code or OpenClaw?

Congratulations on launch!

The plugin approach is very smart move. Plugins are the easiest way to get started with anything

I just checked the pricing and I got shocked because it offers so few requests. Maybe I don’t know how it works, but it seems you need to think very well about what you want to achieve and type it, because if not, you will run out fast.

Hey @jessica_miller_7 Congrats on the launch! i like no-prompt nature of this. Curious what the setup process looks like. Do you have to first give it details on what you’re building , what’s you’re building on, etc?

Absolutely needed Jessica! Does it work through my IDE or connecting my repo? Anyway it's super useful and a truly time/energy-saver. Wish you all the best here!

Code standards change over time as projects grow. How does Enia adapt when a team updates its coding rules or architecture patterns and how do developers review or adjust what the AI has learned so it stays aligned with the team?

The "proactive" positioning is what caught my attention. Most AI coding assistants wait for you to ask — which is fine for exploration, but doesn't scale when you're shipping under deadline. The "detects bugs and architectural issues as you write" claim is interesting. Two questions: 1. **How does Enia handle conflicting suggestions?** If I'm working on a prototype vs. production code, the standards are different. Does it learn context-aware thresholds? 2. **Team standards learning** — how long before Enia understands our codebase patterns? We're a small team with some unconventional architectural choices (multi-tenant customization engine). Would love to know if it adapts or forces conformity. Also curious: Does it work with monorepos? We have a shared utils package across multiple services.

I like the idea of AI focusing on prevention rather than generation. Most tools try to write code for you. But preventing bad code might actually be more valuable.

The plugin approach is smart. One challenge with new dev tools is adoption friction. If it integrates directly into existing IDE workflows, that's a big plus.

Congrats on the launch! Love the "proactive detection" approach—catching issues as you code is way more efficient than fixing them later. Best of luck! 🚀

I work on a small startup team where we ship features quickly, and technical debt inevitably creeps in. Having something that proactively points out potential architectural drift could actually save us a lot of cleanup later. Wondering how customizable the rules are for different teams:)

Something I’ve noticed with AI coding tools is that they help you write code faster, but they don’t necessarily help you write better systems. The proactive idea here is interesting because it shifts the focus from generation to guidance. Would love to understand what kind of signals Enia prioritizes first.

One of the most frustrating parts of development is realizing that a design decision made weeks ago is causing problems today. If Enia can help detect those early signals, that alone could save a lot of time.

One of the most frustrating parts of development is realizing that a design decision made weeks ago is causing problems today. If Enia can help detect those early signals, that alone could save a lot of time.

Looks sweet!

I wonder how proactive it will be, especially when I am working on a large project.

One interesting use case might be maintaining code quality in fast-moving teams. AI catching potential issues early could reduce time spent in PR reviews.

I like the positioning of this as a coding partner rather than just a tool. The best developer tools feel like they enhance your thinking rather than replace it.

Hi everyone!

I’ve been using Gemini 2.5 Flash API in my BYOK translation plugin. Switched to gemini-3.1-flash-lite-preview by literally just changing the model name — quality jumped, speed stayed the same at identical throughput, and the bill is still reasonable. Quite happy.

Official use cases like high-volume translation, content moderation, real-time image sorting, dashboard automation, UI generation and multi-step retail agents are spot on. If your app (or any slice of it) hits any of those, this one is definitely worth a shot right now in preview.

Grab it in @Google AI Studio or Vertex AI.

Multimodal was the right bet from day one, the ability to reason across text, image, and code in a single context window is something GPT-4 is still catching up on.

Using Gemini API while building Fillix - a Chrome extension that makes job hunting embarrassingly easy. The context window size alone makes it worth building on. Curious where the agentic capabilities go next.

When you're generating thousands of localized variations for Google Ads and Shopping campaigns, API costs usually eat up the margin) This price point for high-volume generation is insane. Are there any strict rate limits while it's in preview?

Awesome to see Gemini continuing to evolve! The multimodal capabilities and deep research features look really powerful. What’s the feature you think people are still sleeping on the most right now?

2.5X faster first token is the real headline here. For latency-sensitive apps (chatbots, real-time assistants), that gap is massive. Curious how it handles longer context windows under load.

I was waiting for it, I love it, for sure I am going to add it to YouScaleIt, nice work Google!

$0.25 input and $1.50 output per million tokens, while matching or beating 2.5 Flash quality is the number that changes the economics of high-volume AI pipelines. For anyone running thousands of generations per day, that pricing tier is the difference between a viable product margin and a problem.

The 2.5x faster first token is the other figure worth paying attention to. In real-time user-facing workflows, that latency gap is what separates an experience that feels responsive from one that feels like it's thinking.

I orchestrate multiple AI models, and the cost per generation is a constant pressure point at scale. A model at this price point that holds up on quality for tasks like content classification, asset sorting and UI generation is exactly what makes certain features economically feasible to ship. Curious how it handles multimodal consistency across a long batch run, does quality stay stable or drift?

I've been using Picsart for quite some time now, and I'm still in awe of how the platform constantly (and consistently) is still updating with new features. Congrats on the launch@davit_avoyan!

Can you do this with existing personas, I mean real people? Like put myself in different scenarios?

This is amazing, something I've been looking for, an easy UI/UX for video editing

Been using Picsart since ages, solid product! Good to see them building AI features into it. All the best team :)

Interesting idea, consistency is definitely one of the biggest problems with AI characters. If I create a character in Persona, can I lock its identity so the face and style stay consistent across hundreds of videos and different scenes?

What improvements make GPT‑5.3 “more accurate”? Are there changes in training data, algorithms, or model architecture?

What OpenAI got right wasn't just the models, it was making powerful AI feel accessible to a solo developer at 2am with a credit card and an idea.

Building Fillix on top of OpenAI's API, a Chrome extension that makes job hunting embarrassingly easy. Wouldn't have been possible to ship this fast without the ecosystem you've built. The real moat isn't the model, it's the platform.

Excited to hunt GPT-5.3 Instant today!

This is a meaningful upgrade to ChatGPT’s most-used model focused on the stuff people actually feel every day: smoother conversations, fewer unnecessary refusals, less preachy tone, better web synthesis, and more accurate answers.

What stands out:

• Fewer dead ends and defensive disclaimers

• Stronger judgment around sensitive topics

• Better balance between web results and reasoning

• Noticeably reduced hallucinations

• More natural, less “cringe” conversational style

• Improved writing quality and range

It’s not a flashy feature drop, it’s a refinement of the core experience. Faster clarity, better flow, and answers that feel directly responsive to what you asked.

These are real UX improvements at massive scale. What do you think?

Follow me on Product Hunt to be informed of latest and greatest launches in tech: @rohanrecommends

The novelty of fast text generation has worn off; what we actually need is execution. We need AI that can take on long-horizon tasks without needing its hand held at every single step.

The biggest friction point in AI right now is the gap between a prompt and a finished project. Most tools tap out after giving you an outline or a code snippet. MiniMax operates differently. It’s built as a multi-agent system meaning it actually plans, breaks down complex requirements and executes sub-tasks until the entire job is done.

It’s powered by their M2.5 model (packing a massive 1M context window), which makes it incredibly stable for heavy-lifting workflows that require extended reasoning.

Here is what it actually takes off your plate:

• Lightning Mode: Enhances all chat scenarios, handling daily tasks like Q&A, document summarization, and creative writing with greater speed and efficiency.

• Deep Research & Analysis: Synthesizes web-wide information into in-depth reports.

• Multimodal Content Generation: Effortlessly produces video, audio, and image to enrich your content.

• Create Pro Presentations: Generates polished, visually rich slideshows with just a click.

• Build Full-Stack Apps: Creates stable and visually stunning full-stack applications, from front end to back end.

• Coding & Debugging: Your smart programming partner for efficient coding and testing.

MiniMax Agent is a general-purpose AI that plans, researches, builds, and creates. From deep web analysis and pro presentations to full-stack apps, coding, debugging, and multimodal content (video, audio, image), it executes complex, long-horizon tasks end-to-end with speed.

Now available on mobile:

• Play Store (iOS): https://play.google.com/store/apps/details?id=com.minimax.ai

• App Store (Android): https://apps.apple.com/in/app/minimax-your-ai-agent/id6742651446

Follow me on Product Hunt to be informed of latest and greatest launches in tech: @rohanrecommends

Cool to see this on mobile, how is personal data handled on-device vs. in the cloud? Especially with it running across multiple messaging platforms, curious what the data retention and privacy policies look like.

Very cool! First implementation I've seen at least on the mobile. Will give it a shot.

The variety of languages does seem like a great base boon but reliability is the main stress point for tools like these. So what’s its general error margin?

Interesting, do you offer API?

The URL paste-to-transcript flow is really smart. Being able to drop a YouTube or TikTok link and get a timestamped, speaker-labeled transcript without downloading anything removes so much friction. The 120 min free tier is generous too. How's the accuracy holding up for accented speech or overlapping speakers?

Superb! Does it work on the mobile? Would love to try it out.

This is lovely! Is there a time limit for the audio being transcribed?

What is the difference between "Standard" quality and "High" when it comes to transcribing the video? (Currently testing and didn't find any explanation.)

What if I post a link from YT and would like to follow the script on top of the video or let's say another platform where I would like to have it on top of th original content, is it possible to have that or do I always have to jump between tabs? I think this would really be useful. Good luck

Building with AI coding agents is powerful and rapidly changing how we ship products. But the workflow around them is fragmented. We're either forced to use engineering-based IDEs or oversimplified no-code tools. And if you're not using either, you're bouncing between a terminal, a browser, a file manager, and a code editor, constantly switching context just to prompt, check the result, and make an adjustment.

I wanted something that felt as simple as a no-code tool but gave you the full power of any coding agent. Not a new AI, a better workspace for the ones that already exist. What's more, I wanted it to work with any project and didn't want to lock people in to a particular language or force them to pay for models they're not even going to use.

That's Projekt. One platform supercharging your agent/s: live preview, file browser, inline code editing, element selection, multi-agent tabs, git actions and many quality of life improvements you'll quickly realize you can't do without.

The best part is Projekt is agent-agnostic, aka BYOK, bring Claude Code, Codex, Gemini, Opencode, whatever you prefer. Your key, your agent, your workflow.

This is a free early alpha and I'm building the roadmap in public. If you build with AI and you're tired of tab-switching and duct-taping your workflow together, I'd love to hear what you think. Help me shape the future of building with Projekt!

Bring your own keys' is doing a lot of heavy lifting here, no lock-in is genuinely rare in this space.

How does it handle context sync when multiple agents are working on the same file simultaneously? That's the part I'd want to stress test before switching my stack. Building Fillix (makes job hunting embarrassingly easy) mostly solo so far, but this is tempting.

So cool! and really nice UI. Will explore using this

Really impressed with some of the design decisions here. Great work @bobbydesign

Congrats on the launch, @bobbydesign! This one's for the people who constantly bounces from one file manager to another.

Having one workspace to design, build, and experiment with AI coding agents sounds super powerful. What kinds of projects have people been building with Projekt so far?

@bobbydesign Congratulations on the launch. What is the advantage over Cursor?

Built by a solo dev with too many tabs and not enough discipline' might be the most honest product description I've ever read on Product Hunt. 😂

The midnight reset is genuinely clever - it's not blocking you, just creating a natural forcing function. Fellow Chrome extension builder here (Fillix - makes job hunting embarrassingly easy), and the 'everything local, zero data' approach is something more extensions should commit to upfront. Congrats on shipping!

After using for a couple of days this helped me start my days focused instead of scrolling through my millions of other tabs. Especially with exam season this was very helpful!

This is so cool! :)

ATM I use TabsMagic, but it doesn't have this feature, so I have all those tabs open anyway. Painpoint was spotted really well here.

As a bit of a messy person when it comes to opened tabs, I find it interesting, however, some of them I keep open to use as reminders. that I will check in the upcoming hours or days. I think that we should be able to sort of manually edit or ask for specific pages not to be closed, for example when you started a draft but did not finish it(did not press on save), on a website that you do not use daily. Would I lose that "draft"?

Nice idea, the morning summary sounds like the most interesting part. If a tab gets closed but I suddenly need it later, is there an easy way to restore it from previous days?

I swear I end every day with 30+ tabs open thinking I’ll come back to them later. The daily summary idea is super clever. Did people find it stressful at first having everything close automatically at midnight?

I'm a bit of an "a-type" personality, so my tab hygiene is pretty clean, but I can definitely see how this could be useful for a ton of folks I screenshare with and notice the spiderweb mess of tabs they have open. How do they even live like that?!

Curious if you have any plans for the next version of the extension?

This is awesome and I have like a lot of tabs and I have really clean UI later and Its perfectly archived. super.

Hey @aalabs congrats on the launch! This is a super useful extension 🚀 and I've already gone and added it to Chrome! Thank you for making it free, but honestly I would pay for something like this. Maybe not recurring haha but definitely as a one-time purchase.

One thing I would recommend is in the daily summary, is there a way to see how many times you visited a tab within a day? I could see on the video, you define what's popular by usage but would it be good to also see how many times a tab was visited or if it was re-visited a number of times each day?

Literally the best app to experience the latest @Qwen3 local AI models on your phone! 🚀

"Offline + private + no login = the holy trinity that 90% of AI apps ignore because cloud is easier to monetize. Locally AI is betting on the right side of the privacy conversation.

As someone building Fillix, a Chrome extension that makes job hunting embarrassingly easy, the 'no login required' UX decision hits close to home. The best tools get out of your way instantly. Apple Silicon optimization is the cherry on top. Congrats on shipping!

Great launch 👏

Running powerful models locally on iPhone and iPad is exactly where things are heading — privacy-first, fully offline, no logins, no cloud dependency.

Excited to see Qwen integrated here. Strong reasoning + vision capabilities, and having that fully on-device is a big step forward in user control and data ownership.

Curious about:

– inference speed across different devices

– memory usage and optimization

– how the model download and UX flow are handled

If performance holds up, this could be a serious alternative to cloud-based AI apps. Congrats on the launch 🚀

Qwen 3.5 2B vision on an iPhone 16 Pro is astonishing. This is absolutely the future of device AI. I can't wait for OSes that use AI as the kernel for everything.

curious to check it out

Great work launch team! 🚀 I love everything flight, so this is a fantastic concept for engineers. Particularly captains, crews and crates. Superb way to gamify project management.

Would be great if you could add "teams" -> Custom Airlines ✈️

Scheduled activities -> Scheduled Maintenance / Food & Beverage time

Custom Boarding Pass showing project inception to Project go live date with a loading bar which you can show incremental progress on as tasks are completed.

Backlog -> This can be ATC tower

Would be great it you clicked on an item or card similar to ADO and you could see the boarding pass in the card and even if you added a Departure/Arrivals after duration showcasing when a project is due to be completed and maybe a delayed status if it misses the window.

Also the landing page is great but I think the dashboard UI could be better, maybe if you used the same background from the landing page as opposed to just off white/grey.

Wishing you all the best with the launch! ❤️

Nice work Henrik very cool concept here!
"we should have fewer meetings" and then we'd schedule a meeting to discuss that... make me laugh out loud, this is also quite quintessentially Swedish.
Curious about if there is a "backlog" in a similar way to scrum, and how does the team decide on what goes into the flight? I always find that selection (eg what to work on) is the most difficult decision and can be costly if you choose incorrectly. How does it work here with the Flights methodology?

The "headwinds/tailwinds" language for communicating status upward is the best part of this. We burned months trying to get non-technical stakeholders to care about sprint metrics. Plain language that maps to something intuitive beats a velocity chart every time. How are you handling mid-flight scope changes when a captain wants to add crates after takeoff?

Love the thought behind the name, @henrikhussfelt. Congrats on the launch!

@henrikhussfelt Congrats in the launch! I love this ideology. You mentioned the captains are strict which is great cause we need someone focused in landing that plane! However how does a small team with only a couple managers/supervisors navigate this process with so many flights that need to take off? Or is this geared more towards bigger teams and enterprise?

The flight metaphor is doing real work here — a sprint implies circular motion (you always restart), a flight has a destination and a real cost to rerouting mid-air. The "no story points, just done or not" discipline is the key insight.

Curious how you handle scope creep: if a crate grows during the flight, do you reroute mid-air (extend the flight) or leave it on the tarmac for the next one?

good luck on the launch guys, using your api sometimes for voice to text and vice versa

Speech-to-text is a solved problem until you actually try to build on it, then you realize accuracy on accented English, speaker diarization, and real-time latency are three completely different challenges. AssemblyAI quietly nails all three.

Exploring voice-based job interview prep features for Fillix - a Chrome extension that makes job hunting embarrassingly easy. The LLM gateway + transcription in one API call is exactly the kind of DX that makes a feature go from 'maybe someday' to 'shipping this week.

Voice logging in FuelOS runs on streaming STT and the place it consistently fell apart was alphanumeric strings, things like "vitamin B12" or "omega-3" getting mangled mid-stream. How does Universal-3 Pro handle those in noisy kitchen environments specifically, where background noise compounds the problem?

Congrats on the launch of Universal-3 Pro! In the retirement planning space (Ready Aim Retire), accuracy isn't just a metric; it’s the foundation of user trust. Seeing a model that prioritizes the 'hard stuff' like alphanumerics and disfluencies without needing heavy prompt engineering is a big win for those of us building finance tools. Keep up the excellent work!

This is very cool! Looking forward to trying it out.

Very useful! Does it work inside perplexity?

memory popular these days

how about saving it into obsidian notes locally. Google chrome is vulnerable. Hackers stole my entire password chain from Chrome

This is hilarious ... but also super unique and innovative! Next step - add a ping pong table where your agents can take a load off and reset. Maybe a beer fridge for those Friday afternoon sessions. Water cooler anyone?

Haha wow cool concept - can you have multiple openclaw agents hanging out together?

very cool! We're doing something similar, but focused on Web3. All the best!

The Auto-Heal feature is the real hook here. Most of us waste 40% of our dev time just acting as a human clipboard between the terminal and a LLM. If Nova can intercept a traceback and just apply the fix, it moves the AI from a consultant to a co-pilot.

Creating any product takes real perseverance. Creating something that supports relationships takes even more devotion. Humans are complex—and two humans trying to understand each other even more so.

I respect the commitment it takes to step into that space and build something that helps close the gap between where people are in their relationships and what’s possible for them.

Excited to see the many relationships this will support. Congratulations!

Hey @keith_gould I really like the concept behind this, it encourages being better in relationships and I haven't personally seen anything like this before, which is awesome. Best of luck with the launch! Love from India.....

hahah that is very 2026

Congrats on the launch! Just downloaded and started going through the onboarding. LOVE the little quiz question to make sure users are super clear on the purpose of the app before getting into it. You've got me thinking about adding something like that to an app I'm working on.

One piece of feedback: when I hit the reflection portion and was forced into typing at least 50 characters, I felt a little discouraged becuase I didn't "feel" like it at that time, and was more interested in diving into the app and exploring more before committing a bunch of time. If it's important to keep this, maybe even a little note could go a long way, acknowledging that users might nto feel like doing it right now and/or why it's important at this time to fill it out?

Hey @keith_gould I really like the concept behind this, it encourages being better in relationships and I haven't personally seen anything like this before, which is awesome. Best of luck with the launch! I love the avatars, I would say one improvement, the UI could use a bit of work, particularly in the actual chat functionality.

It does the same with audio files, I saw videos being mentioned in the description, how does that work? Kudos on having it as a pop-up so that we do not have to jump between tabs, that takes some frustration away

Hey PH! 👋 I’m Apostolos, the person behind ClawPane.

I built ClawPane after my first day on @OpenClawturned into a $50 bill from “simple testing” (Claude was the main culprit). OpenClaw is great at agent orchestration, but I kept running into the same gap: agents need a smart way to choose which model should handle this prompt without me hand-tuning every step.

ClawPane is an extension to OpenClaw that plugs in as a single custom provider. You set priorities (cost, speed, quality, carbon), and ClawPane automatically routes each request to the best-fit model. Your agents don’t change. Your prompts don’t change. Only the model selection does—per request.

Why ClawPane (instead of “just OpenClaw”)? OpenClaw orchestrates agents and tools; it doesn’t try to optimize model choice across providers and prompts. ClawPane adds that missing layer: per-prompt model selection, enforcement, and visibility—built specifically to fit OpenClaw’s provider pattern and drop in fast.

Here’s what you get:

• Automatic routing per request: stop overpaying on easy prompts, reserve strong models for hard ones

• Multiple router profiles: cost-first support agents, quality-first coding agents, latency-first realtime flows

• Automatic fallbacks: keep running when a provider degrades or goes down

• Real-time cost + latency per request: see exactly what each agent call costs and how long it took

Who it’s for: anyone building agents on OpenClaw who’s tired of manual model picking, brittle defaults, and surprise bills—especially with mixed workloads and production reliability requirements.

What models are you using with OpenClaw today—and where does model choice hurt the most?

Interesting idea, automatic model routing could save a lot of cost. Сan developers see or override which model was chosen for a specific request if they want more control?

Super useful, following this and will look to integrate with our platform moltin.work

Looks cool man! Is it mainly for enterprises or for smaller teams as well?

Hey @tirell_arzu super glad I stumbled upon this! Next step in automation for sure. Love the performance dashboard showing confidence score and analytics. Wishing you all the best with the launch 🚀

Questions:

1. How do you determine next actions to take for the model in terms of predicting what comes next?

2. Is there a fallback option on the off chance the model incorrectly defined and carried out steps but you would want to reverse a task?