We started as a builder for AI websites. Good product. But the more we talked to users, the more we realized they didn't want better websites, they wanted better work. So we went bigger.

Agentplace lets you create specialized AI agents for real tasks and workflows. Think AI teammates that actually help you get things done. Generative UI, voice, browser memory, agents that adapt to each user over time. A unified workspace where you can switch between agents and get real work done.

And we built around one core insight: the trick isn’t “build the perfect agent.” It’s to ship, use, fix, and repeat fast.

Work mode gets your agent running. Edit mode brings you back the moment something breaks or a better model drops. Republish in minutes.

We're genuinely excited to hear what works, what breaks, and what we should build next. Every comment here shapes the product.

Looking ahead, we’re doubling down on this idea of AI teammates.

We think the future isn’t just better agents, but a shared workspace where agents and people work together.

Agents will handle more work end-to-end, talk to each other, and run tasks autonomously. People stay in the loop, see what’s happening, and step in where judgment matters.

Over time, this becomes a new kind of work environment, where humans focus on decisions, and agents handle execution. That’s the direction we’re building toward.

What's the first agent you'd build?

Congrats with a new launch @polina_semina @vlad_yanch
This is cool bc it feels closer to how adoption actually works inside companies. Does one person usually own an agent or can a few people manage it together? That part matters a lot for teams.

I keep seeing agent products but most of them still feel kinda abstract.
This one feels more grounded tho
Is there still a learning curve for non technical people, or is setup actually lightweight?

Congrats on the launch! 🚀

how do you see your main audience at this stage? more developers building custom workflows, or non-technical users exploring agents for the first time?

Congrats on the launch! Agents instruct agents… Does you tool work over a codebase to tailor an agent for it? Asking because generic agent instructions is something I believe Claude itself can generate, wondering how it works in your product

Hey! Polina here, I run operations at Agentplace.

I’ve been actively using our agents for day-to-day tasks, and it’s been a huge help with routine work.

Would love to hear your feedback if you get a chance to try it!

Hi all, CTO of Agentplace here, feel free to ask me anything. And would really appreciate your feedback!

Hey Product Hunt! I'm Boris, AI Engineer and one of the makers behind Agentplace. Super excited to finally share what we've been building. Happy to answer any questions about the product, the tech under the hood, or anything else - just drop a comment!

It’s been 4 hours since launch, thank you all so much for the support!

Me and the team are always open to comments, feedback, and ideas.

Hey! I'm Andrei, founding engineer at Agentplace.

Excited for launch day! If you need help with anything or have feedback - just drop a comment. I'm here!

This looks genuinely useful. Can you build something really specific for one role, like recruiting or sales ops?

Congratulations on your launch, @systerr . Looking forward to playing around with it!

Can Agentplace plug into existing pipelines without restructuring, or do teams inevitably have to adapt their stack?

Congrats on the launch! its very nice to have a free plan so that the people can actually try the product before purchasing it.

Congrats on the launch! The tool is easy to use and launch!

Congrats on launch! Feels like real AI teammates, not just another agent toy

Interesting that this is built around workflows and not just 'we have AI now.' Quick question: how easy is it for an ops person to pick this up without looping in engineering every time? That's usually where these things break down.

Congrats on the launch! What happens when a model gets updated or replaced? How much work is it to re-test and adjust an existing agent?

Hi there!

2 hours in, and we’ve already collected a lot of valuable feedback for improving the product. Thank you all so much for the support. If you have thoughts to share, we’d love to hear them. :)

Congrats on the launch! Wishing you lots of traction on Product Hunt and beyond. Excited to see where you take this!

Congratulations

Congrats on the launch! Pivoting away from something that was already working takes guts! respect for that.

Curious: what's the wildest agent someone's built since you made the shift?

A few hours left until the end of our launch, and we’re currently #1.

If you’d like to support us, we’d really appreciate it.

Thank you so much for being with us today! ^-^

i can actually imagine using this for repetitive ops work or routing stuff internally. finally something i can picture in a real workflow

Love the pivot from website builder to agent workspace!! The work mode / edit mode loop is such a smart pattern. We've been building AI into our recruiting workflows and the hardest part is exactly what you described: agents break on real edge cases, not the happy path. Excited to see where this goes!

this is so cool, Vlad. why do you think agents need UI? why not to use claude code in terminal?

Congrats on the launch! Love that you pivoted from websites to workflows based on what users actually wanted. That's how it should be done.

Congrats on a launch!

Been using AngetPlace for a while now - awesome product! Saved so much of my time

Huge congrats on the #1 spot, Vlad & team! 🚀

Just a heads up ,the mobile site is still hitting a Connection Timeout for users in my region. I'm a QA Tester and I've been monitoring the uptime.

Once you're back up, I'd love to run a quick Stability Audit to make sure the surge didn't break your onboarding flow. Rooting for you guys to get back online!

Claude’s latest “Аuto Мode” might be the smartest update yet by @Claude by Anthropic.

It bridges the gap between AI thinking and action by letting Claude handle file writes, commands, and workflows on your computer. With permission and without constant approvals. Safe actions run automatically, risky ones get blocked and handled differently.

Set it once, let Claude manage repetitive tasks, scripts, or reports, and free your attention for higher-level work. Perfect for devs, operators, and founders who want AI to actually do, not just suggest.

Available on Team plan now; Enterprise and API coming soon.

The classifier basically automates what I already do most cases are obviously safe and get approved without much thought. That’s helpful, but the real challenge is the small percentage where things get weird and need context. Not sure if it handles those well or just falls back to a generic “blocked.

Umair is right on, the 90% rubber stamping was never the real friction, it was just the visible friction. The actual problem is the 10% where you need to understand what Claude is trying to do and why before you can make a good call. If the classifier just blocks those with a generic message and no context, you've traded one interruption for a worse one.

What I'd want from auto mode is not just safe/blocked but a third state: "proceeding but flagging this for your review." Something that lets the session continue without stopping but surfaces the decision for you to audit after. That way you're not context-switching mid-flow but you're also not flying completely blind on the edge cases.

Use Claude Code daily and the constant approvals do break the flow, especially on long agentic sessions. The right trust model here isn't yes/no per action, it's more like a pilot and autopilot relationship. Autopilot handles the cruise, the pilot takes over when conditions get genuinely tricky. Curious how the classifier is trained and whether it improves over time on the user's specific codebase patterns.

This feels like a deeper shift than just removing approval clicks.

The real challenge was never the obvious safe actions, it’s the small % where you need context to trust what the agent is doing.

Auto mode starts to look less like convenience and more like a trust layer between human and agent.

Curious how you think about that boundary evolving.

Do you see this staying as a permission system, or becoming something closer to dynamic trust based on behavior over time?

Phenomenal launch, Anthropic team! 🚀

We've been auditing Claude Code and the new 'Auto Mode' architecture. The standout feature is the Guardian Classifier—it’s the most elegant solution we’ve seen to the 'permission fatigue' problem, allowing for true delegation without sacrificing system security.

For developers curious about the 'Project Scope' guardrails or how this fits into the 2026 agentic landscape, we've published our full technical audit here:

👉 Article Link

The 'Coffee Break Test' is officially passed! Looking forward to seeing how this evolves.

I use Auto mode since it came out and it is great.

This sounds amazing, finally an AI that actually does things instead of just talking about them

The permission classifier framing is interesting. It's essentially teaching the model to internalize your risk tolerance rather than defaulting to ask. I'm curious how it handles drift over time - if your codebase or usage patterns change, does the classifier retrain, or is it more of a snapshot of your initial preferences?

the classifier is doing the same thing i already do mentally when i hit yes/no on approvals. 90% of the time its obviously safe and im just rubber stamping it. nice that they automated the rubber stamp but the real problem was never the safe actions, its the 10% where claude wants to do something genuinely weird and you need context to judge it. curious if the classifier catches those or just lets them through with a generic "blocked" message

Hey PH 👋

I'm Dan, cofounder of Pendium. We help startups market their products to AI agents.

TL;DR — We're a new kind of AEO/GEO platform that helps businesses monitor and increase their visibility to AI agents, so you get more leads and customers from ChatGPT, Claude, Gemini, Perplexity, and more. It’s an easy-to-setup continuous visibility monitoring system that ties into your content engineering strategy, with a platform designed for humans, agents, and mixed teams.

Your first full visibility audit is free and getting started is as easy as entering your website URL and letting the system run:

🧑 For humans 

⤷ https://pendium.ai/industry/startups?producthunt

🕵️ For AI agents

⤷ https://pendium.ai/mcp

At my last company, we helped enterprise marketing teams at companies like P&G, Red Bull, and Microsoft create and distribute content designed for humans. In this new world, marketers need to also treat AI agents with the same level of attention and care.

Humans are turned off by AI slop — and turns out so are AI agents.

At their very core, AI agents care about finding truth. What they decide to recommend draws from what they've been trained on, what they can find, and ultimately what helps them do their jobs.

That's the core of what Pendium does: we help AI agents do their jobs — if you genuinely help AI agents do their jobs, they'll recommend your products and services more often to the right people

⋅.˳˳.⋅ॱ˙˙ॱ⋅.˳˳.⋅⋅.˳˳.⋅ॱ˙˙ॱ⋅.˳˳.⋅

You get a scan, YOU get a scan, YOU GET A SCAN!

Run a full AI visibility audit for your business at http://pendium.ai

*** If you post the report in the comments here we'll go through it with you (and give you free credits)

1. Enter your URL (dooo it)

2. Confirm the conversation topics where you care about being recommended

3. Confirm the personas that map to your buyers

4. We run a ton of parallel LLM calls on GPT, Claude, Gemini, Google, etc

5. We monitor the agents' thought processes and all responses and citations + extract brand mentions, sentiment, competitors, more

6. We process all of the data and present it in a clean explorer, separating out brand sentiment vs. how often you're recommended in conversations related to your core business vs. in broader aspirational growth areas

7. We generate insights and recommendations for what kind of content to create to help AI agents do their jobs when they're researching in or around your category

8. Send the whole thing to your AI agent via our MCP or API to give it AI visibility superpowers (access to on-demand scans, competitor data and more to help your own system create content that moves the needle)

9. Or — use Pendium's hosted platform to engineer content, grounded in your brand voice and tied into your existing knowledgebase and web presence (with human-in-the-loop approvals and input)

10. Push that content to your own CMS, or host it in a dedicated feed for AI agents

⊹ If you genuinely help AI agents do their jobs, they'll recommend your products and services more often to the right people

Fully free to start, with seat-based plans that scale based on your usage

⋅.˳˳.⋅ॱ˙˙ॱ⋅.˳˳.⋅⋅.˳˳.⋅ॱ˙˙ॱ⋅.˳˳.⋅

Woof. That’s a lot! The good news is that we designed Pendium to be agent-native and as autonomous as you want it, so increasing your AI visibility is as easy as entering your website and copying a snippet to your coding agent (Claude, Cursor, OpenClaw, etc), or letting everything run through our hosted web app.

We're giving 250k free credits for Hunters here, which will get you started with a full visibility scan and a full month of content engineering:

⤷ https://pendium.ai/industry/startups?producthunt

If you're already working on marketing to AI agents and want to compare notes, I'm in the comments.

I'm finding myself using Chat increasingly for purchases. There's a lot of experimentation in AEO/GEO, which could expand heavily if/when OpenAI and others introduce ads.

@dgreenberg what's the biggest surprise or learning building Pendium so far? I'm also curious to hear about the counter-intuitive things founder uncover when building something new. :)

Congrats on launching. The idea of marketing to AI agents instead of just humans is something I hadn't really thought about until now. But it makes total sense given how many people are using the ChatGPT and others to research products before buying. Really curious to try the free visibility audit. Does the platform show you which specific AI models are recommending you more, or is it more of an aggregate view?

Hey PH 👋 Rob here — cofounder at Pendium.

@dgreenberg covered the what. I want to share a little bit of the why.

From the start, we wanted to build something that helps people harness the power of AI to simplify and augment their lives (and definitely not replace them). With our background — 15 years at Sharethrough — we learned firsthand just how hard it is to get your message, story, and brand out to the world, and it’s only getting harder! We built Pendium to make it easier and faster to get your story visible.

Here’s how it works:

• We run hundreds of parallel LLM calls across ChatGPT, Claude, Gemini, and AI Overviews — simulating real buyer personas asking real questions in your category. Then we show you exactly where you show up, where you don’t, what gets cited, and what to do about it.

• We feed those learnings into a content engineering system so you can harness AI to quickly flesh out your message — grounded in real visibility data, in your voice, targeting the specific gaps that matter.

We built this to work for everyone — small business or Fortune 500. Quality should determine visibility, not budget.

Run yours and drop your report in the comments — we’d love to go through it with you and learn from your feedback!

Congrats on the launch @dgreenberg ! Would you say one has to do a (tough) trade-off between optimizing one's page for AI agents vs for humans or is it possible to optimize for both at the same time?

Congrats on the Pendium launch. Targeting AI agents is a great idea.

I visited the homepage of your product, @dgreenberg. One thing caught my attention in the hero section.

You say: [Reach more customers on....recommend your products and services to the right people]

Then you show a ChatGPT response where your business is listed as #1 with a note: [Your Business Should Be Here — Use Pendium to help AI agents and chatbots recommend your products and services.]

That's a demo. But it's also a contradiction.

How?

Lemme tell show you. You're showing a recommendation that was influenced by Pendium, but the example itself is promoting Pendium. It reads like a placeholder. A person landing on your page might think "is this what my business will look like, or is this just an ad for them?"

The middle section says "73% of users trust AI recommendations over traditional search."

That's a strong stat.

But it's buried under a wall of copy. A visitor scanning might miss it.

And the "how it works" section is six steps. That's a lot for someone to digest before they understand the value.

I attached a screenshot to show what I mean.

Spotted other things that could cost you. Happy to share if you are open.

These are just valuable insights from me because I analyze product pages daily and spot the loopholes. Anyhow, good luck!

As someone who spends most of my week auditing martech stacks and trying to nail down attribution, the shift from traditional SEO to AI agent visibility is officially my new headache. Tracking how agents actually cite and research products is definitely the next frontier for us.

Really love the approach you guys are taking here. The platform looks incredibly clean and solves a very real, emerging problem for marketing teams. Congrats on the launch.

Congrats on the launch! The core loop feels familiar, understand how the system works and optimize accordingly. Curious though, how stable are the visibility scores week to week? LLM outputs can be pretty noisy so wondering how you normalize that into something actionable.

SEO is dead, long live GEO. Now I need to optimize my content not just for Google's algorithm but also for Claude's feelings about my brand. We really are living in the future. 🤖

This is such a smart angle!! In recruiting, we deal with this exact problem: the right candidates exist but the systems don't surface them to the right people. Curious how you handle nuance in matching context to audience, especially when the signals are messy and unstructured?

This is a really interesting shift.

It feels like we're moving from SEO to something closer to “agent discovery infrastructure” — where the goal is not just ranking for humans, but being selected by AI systems during decision-making.

If that plays out, distribution itself starts to change quite a bit.

Instead of optimizing for clicks, you're optimizing for being chosen inside an agent’s reasoning process.

Curious how you think about this long term.

Do you see Pendium more as a marketing tool, or as infrastructure for how products get discovered and selected by AI agents?

Congrats on the launch! GEO is going to be a huge thing. Most people are still thinking about traditional SEO but AI recommendations are already changing how people discover products.

Nice tool !

Do you provide outreach on top sources ?

Congrats on the Launch. This looks quite nice. I was wondering, what you do different to all the other services. Is it the personas? I think that makes it especially useful. Just wanted to try it out but now I'm stuck on the Persona step and I'm too tired now to thoroughly fill all the relevant personas.

Anyways, last summer I was onto building such service myself but stopped working on it in favor of my current SaaS.
What I am wondering: how can you reliably detect citations? I mean you probably just use their APIs, but ChatGPT is not equal GPT, it adds some flavor on top of the API. So if you run a prompt I guess the response will be different from what I would prompt in the UI. Or what am I missing?

Wow, really cool! Our score is 0 😄 But I hope it is because we just added few days ago content to address this gap and it has not been indexed yet. @dgreenberg how log does it take from your experience to see results once you publish content?

It's amazing Dan! It's key to understand what agent is the best option for some specific task, even more for specific businesses. Hope many founders can take the most of it. Really congrats!

how is it different from otterly and other geo tools these days?

Google is on a roll recently, do you think with TurboQuant we can now run powerful LLM models even on a 16GB RAM device?

What is TurboQuant?

TurboQuant turns one of AI’s biggest hidden bottlenecks, memory, into a solved problem. Probably one of the most important efficiency breakthroughs for large scale AI systems?

It closes the gap between model performance and system limits by massively compressing the vectors that power LLMs and search engines without sacrificing accuracy.

TurboQuant works by rethinking how data is stored and compared. Instead of keeping bulky high precision vectors, it compresses them into ultra compact representations while preserving their meaning and relationships. This allows AI systems to run faster, cheaper, and at much larger scale.

It combines two novel techniques. PolarQuant restructures vector data into a more compressible geometric form, and QJL uses a tiny 1 bit correction layer to eliminate errors. Together, they deliver near lossless compression with almost zero overhead.

Compress once, and everything improves. Memory usage drops, retrieval speeds increase, and long context performance becomes far more efficient.

Key capabilities:

- ultra low bit compression down to about 3 bits

- near zero accuracy loss

- 6x or more reduction in KV cache memory

- faster attention and vector search up to 8x speedups

- no retraining or fine tuning required

In a world where AI is hitting hardware and scaling limits, TurboQuant feels like a fundamental unlock for making models smaller, faster, and more deployable everywhere.

How do you think this will change the game?

This is an absolute game changer! I couldn't wait to run the algorithm on our custom models.

Hey Product Hunt 👋

I'm Nathan, and together with my team - Chan, Neel, and Ha. We built LayerProof because repurposing content across social platforms is one of those tasks that sounds simple but eats hours every week.

You write a blog post. Now you need a LinkedIn post, an X thread, an Instagram carousel, and something for Facebook. Each platform has different formats, different lengths, different vibes. So you either spend an hour manually adapting each one, or you copy-paste the same text everywhere and wonder why engagement is flat.

LayerProof turns any URL into platform-native content in seconds.

How it works:

1. Paste a URL: either blog post, article, product page, whatever you want to share

2. AI generates content tailored to each platform (not generic reposts, actual native formats)

3. Every claim stays cited → trace any point back to the original source.

It's free to try and we built this for content teams, founders, and marketers who are tired of the repurpose grind.

Try it at https://layerproof.app/social-content-creator/ and tell us what sucks 🙏

Hey PH 👋 👋 👋 Neel here.

I took two years off to be a full-time dad in 2023. When I came back to work, AI content was basically everywhere, with all its telltale signs. One thing really annoyed me me: most tools just make things up confidently, and nobody seems to care.

We built Matte because we kept running into the same problem. You paste a URL, the AI generates social posts, and half the stats in the output don't exist in the original article. That's not a tool that helps you do things better or faster, it's just pain.

So we made accuracy the whole product. Paste a URL, get ready-to-post content for LinkedIn, X, Instagram, TikTok, Facebook. Every claim traced back to the source. If it's not in the original, it doesn't make it into the output.

We're early and building in public. Tell us what's broken, what's missing, what would make you actually use this every day.

That's what we're here for!

I am a (french) podcast host and also work as a Spartan!

Congrats on the launch.

LayerProof helped me create an Instagram carousel post out of a YouTube video in minutes. It one shotted it everything, including using by brand (font, logo).

You can steer it as much as needed and gives you variations to pick from. Then I extended my post to the other platforms and again it one shotted it.

Will keep using it in my workflow to save time.

Thank you Chan, Neel, and Ha!

PS: Feedback you could autogenerate / suggest hashtags and caption, would be helpful. Now I am using another tool to do so but would be cool to have it integrated here.

Congrats on the launch! Repurposing content across platforms is one of those things that sounds simple but is so painful in practice. Cool that it formats natively per platform.

The interesting part here feels like what happens right after generation. Getting multiple versions per platform is useful, but it can also turn into more decisions instead of fewer. Early on, I’d be curious how often people actually take one and post it as is, vs going back to edit or not using it at all.

This is super handy. Does it keep the original tone?

I’ve actually been using LayerProof Matte for a bit now, and it’s one of those tools that just quietly makes life easier.

Usually I’d have to rewrite the same idea 4 - 5 times for different platforms, but now I just drop in a link and it does a really solid job adapting it for LinkedIn, X, IG, etc.

What I like most is that it sticks to the source instead of making things up makes me feel way more confident posting.

Simple idea, but executed really well. Definitely worth trying if you post content often.

Lovely, I'm trying this today

Any specific feature that you wouldn't want me to miss as a new user?

Big congrats on the launch! Repurposing content across 5+ platforms usually eats up half of my week. This is exactly the kind of tool I didn't know I needed, especially during launch prep. Can you schedule the posts directly from LayerProof, or is it export only for now?

Hey everyone 👋,

I’m Chan, founder of LayerProof!

We’ve spent a lot of time building with LLMs, and one thing became clear: great output doesn’t come from better prompts alone, it comes from better input. And better input is a system: data > patterns > insights. The insight is where the real leverage is, because it compounds over time. Instead of building another content generator, we took a different path and focused on curating the right data before it ever reaches the LLM, so what comes out is actually meaningful, not just polished. This is especially important for teams trying to consistently produce high-quality content across different formats.

What we believe:
- LLMs are powerful, but only as good as the data you feed them
- Insights > raw content (because insights compound)
- Great content starts before generation, not after

We built LayerProof because we needed it ourselves. We’re a team of engineers, and like most engineers, content isn’t our strength. Instead of working around that weakness, we decided to solve it. It’s still early (about 3 months in), but we’re shipping fast and improving every week. And we’d love to work as closely with our customers as possible to make it better.

If you get a chance to try it, please DM me. I’d really appreciate your feedback. 😊!

Congrats! Very cool. :)

One thing I'm curious about that hasn't come up: does the brand voice setting persist across sessions and URLs, or do you have to re-configure it each time you start a new campaign?

With Omma you can create 3D interactive experiences and tools with AI, straight in the browser. With native WebGPU support and parallel agents, your wildest ideas are just a prompt away. 🤯

Do you plan to add team collaboration features where multiple people can prompt and edit the same project in real time, similar to how Figma works for traditional design?

Wow, 3D and apps together? How well does it handle complex builds?

The parallel agents approach is interesting, but I'd want to know how coherent the output actually is when you're combining code, 3D, and media in one build. Each of those is hard on its own. How much back and forth does it usually take to get from first generation to something actually usable?

Congrats on the launch! Parallel agents for building is a really interesting approach. The 3D generation part is what caught my eye. What kind of 3D assets can it handle?

Hi everyone 👋 I’m the developer behind Flowershow.

Flowershow came from a pretty simple frustration: we wanted publishing Markdown to be easy, without giving up Markdown itself.

A lot of tools make publishing easier by asking you to move your content into their system, change your workflow, or adopt some special format. We wanted the opposite. We wanted Markdown to stay Markdown, and publishing to be a separate layer on top.

We were also tired of setting up and maintaining sites every time we wanted to publish something. We could do it, but it always felt like too much overhead for something that should be straightforward. We also wanted the result to feel polished — not like a generic generated site, but something genuinely nice to share.

So we built Flowershow — a hosted way to publish blogs, docs, wikis, and knowledge bases from Markdown, whether from GitHub, Obsidian, the CLI, or drag and drop.

Curious to hear what you think and whether this fits your workflow.

This is owesome service!

I've also thought for a while that it would be convenient if Markdown could be published directly.

Quick question: any plans for other markdown apps support?

Can you connect it directly to an Obsidian vault or does it require a separate markdown repo? Love the simplicity of this, well done on shipping it!

Congrats on the launch!! love the philosophy here.

Quick question: any plans to support multiple sites under one account? Thinking about use cases like agencies or teams who'd want to manage several projects from a single workspace.

Markdown to website instantly? Does it support custom domains?

Hello, I discovered a very important vulnerability on your platform. Basically I’m able to read all files from public or private Github Repository of every single one of your users.

Do you have a bug bounty program?

Looks good, have you thought of creating FAQ- or Wiki-themes?

It seems all the templates are for blogs.

Congrats on the launch! No vendor lock-in and Obsidian import is a big deal. Definitely giving this a try.

Congrats For Publishing This Fantastic Tool ✨
boor Question 😅

is it free all-time ? or i have one (flowershow) domain for all docs ?
supports the linked words like obsidian ? what if i uploaded a file that has a linked word to another file -i didn't uploaded it- ?

what if i edited a file ?
Expand the free space a bit 😄 a 100mb will make a fear for personal using for one like me 🙄😂

Really cool idea. How do the AI teams handle complex tasks?

Finally an app that automates my crippling obsession with checking Google Flights at 2am. My therapist will be thrilled. Or unemployed.

Congrats on the launch! When a fare alert fires, where does it take you? Does it deep-link into a specific booking platform like Google Flights, Kayak, or Skyscanner?

Finally! How far in advance does it start tracking prices?

This is such a good idea. I always feel like I’m checking flights over and over hoping they drop. Does it actually catch the best deals early enough to make a difference?

The cutest website I have seen this week.

It's clear that the builder has a good aesthetic sense (judging by GitHub as well) :)

Running vision-based agents for web monitoring is something I've thought about a lot - the fragility of DOM-based approaches is a real pain. One thing I haven't seen addressed much: how do you handle the token cost at scale if you're running continuous frame capture across multiple concurrent agents? That feels like it could get expensive fast.

Can it constantly track a list of my competitors and keep giving me updates on the pages that are launching everyday?

Congrats on launching! The vision first approach is what sets this apart. Most browser automation breaks the second the UI changes, but agents that actually see the screen the way a human does should be way more resilient to that. How well does it handle sites that are heavy on dynamic content or have login walls? Does it manage sessions on its own?

The "sight-driven" approach is the right bet. APIs break every time the UI changes, but vision-based agents adapt the same way humans do. We're working on something similar for desktop automation (not just browser) and the reliability difference between DOM scraping and screen vision is night and day.

How does Magine handle sites with heavy dynamic content like infinite scroll or lazy-loaded elements? That's usually where vision agents struggle.

Congrats on shipping this, the vision-based approach vs DOM scraping is the right bet. One question: once your agents are running scheduled tasks autonomously, how do you get visibility into what they're actually doing at the prompt/response level? We ran into this with local agent stacks and it became a serious blind spot. That's what Veil-Piercer solves, curious if browser agents hit the same wall.

curious about the vision aspect - are these agents actually processing visual elements on pages or just seeing the DOM structure? the idea of AI agents that can navigate sites like humans do is fascinating, especially for automating tasks that require visual context recognition.

@sagar4nfs Super cool. So, if I have a Playwright script that suffers from this DOM hell you speak of, constantly breaking, could your agent analyze the script and recreate it using vision?

How does Magine handle CAPTCHAs and other anti-bot protections while browsing autonomously? Really exciting to see vision based agents in action, nice work!

The State of Vibe Coding 2025 report highlighted a security challenge. @Lovable is addressing it with @Aikido Security. Promising.

Awesome product, @flxg! I spent 5 minutes on the page.

The "Secure everything you build, host, and run" line is a bold promise.

The hero section caught my eye for one reason. You claim to fix vulnerabilities automatically with AI AutoFix. Then below, under the Unified Platform section, you list 16 categories of tools you replace. That's a lot to digest.

Here's the tension.

A developer landing on your page reads one central system. But... the page immediately shows a long list of 16 things you replace. That creates cognitive load. Got it?

And a user might think this is a platform and then be handed a shopping list of features.

The middle section has a strong stat: "Cut false positives by 85%." But it's buried under a HOW IT WORKS block.

That's your strongest differentiator. A security lead needs to see that number before they see how you do it.

I attached a screenshot to show what I mean.

Spotted 3 other things that need to be tightened. Happy to share.

Security built into Lovable is a great move. How deep does it scan?

I have been able to dogfood Insights for a while on my other product RYSE (social workout tracking app). Figured it'd be fun to show what sorts of insights I am getting for free there just by being hooked up to Basedash.

Basedash Insights surfaces really cool usage trends that we never would have thought to investigate.

For example, one insight was: "When teams start using Basedash in Slack, chat activity jumps — and nearly half of it moves there." That gave us a strong signal that our Slack integration is valuable and something we should promote more, since it's a meaningful value-add for users. I realize this insight is a bit "meta," since it's Basedash generating insights about Basedash, but many of our customers have reported similarly useful insights.

Super pumped about this launch. The first time you receive an insight that you didn't even ask for you really appreciate the magic of autonomous analytics. It's so much better than having to remember to check things yourself.

Really interesting direction.

It feels like this goes beyond analytics into something closer to a decision layer.

When the system not only analyzes data but decides what matters, surfaces priorities, and suggests actions, the role of dashboards starts to fade.

Instead of exploring data, teams are being guided toward decisions.

Curious how you think about this long term.

Do you see Basedash evolving as a BI tool, or as infrastructure that actively shapes how companies make decisions?

I was testing it:
https://x.com/BusMark_w_Nika/status/2036180150677897324?s=20

Also they have other features interesting, esp. video

Hi everyone!

With Uni-1, Luma is making a very strong statement about where image models are going. Generation without understanding can only go so far.

By unifying understanding and generation in one architecture, Uni-1 is Luma's first serious step on the path toward unified intelligence. That is what makes this more interesting than a normal image model launch.

A model that can actually understand style, structure, references, and intent before drawing is a very different thing from a model that just predicts attractive pixels. That is the real signal here.

Try for yourself: https://app.lumalabs.ai/

What's the maximum lenght of the video generated?

Hey everyone! 👋

I'm Robin, the dev behind Coddo.

I built this because I was drowning in Claude Code tasks with no visual structure. Branches everywhere, context lost between sessions, no idea what was actually running. I wasn't coding faster, I was just managing mess faster.

So I built a task-first dev environment. Not a better IDE. A different paradigm entirely, your Kanban board is the interface, the code is just the output.

I've been using Coddo to build Coddo itself for the past few months. That was the real test.

The app is free to download on macOS.

Curious: what's the first thing you'd delegate to Claude Code if your whole workflow was organized around tasks?

Congratulations on the launch! Do you have plans to release a Windows/Linux version in the near future?

Congrats on the launch! This is so real. I always think of new tasks while working with AI agents and just forget them by the time I'm done reviewing. Love having the Kanban right there for that.

I've been using Cursor daily and still find myself context-switching between "what am I building" and "which file am I in." congrats on your launch!

the "skills" concept caught my attention. we work on a lot of healthcare integrations where conventions matter (FHIR naming, security patterns, etc). if this can actually learn and apply team conventions automatically, that's huge. most AI tools ignore the stylistic consistency that makes codebases maintainable. what kinds of conventions have you tested it with?

How do handle plan mode inside Coddo?

Hi, Is it for one project at a time or multiple projects can be handeled?

The skills concept is the most interesting part to me. Managing AI agents across a team means everyone ends up with slightly different prompting patterns and code conventions drift fast. If skills can encode team standards and actually get applied consistently, that's a much bigger deal than the kanban UI. Does the skills layer work per-project or is it shared across the team?

What 3D file formats can you import and export from the workspace? This fills a real gap in the design workflow, kudos on launching!

@marcworms Very interesting. But does it create and export 3d models (in formats such as fbx, obj, etc), or is it just generating images using 3d look styling?

Amazing concept! Would love to know more about export formats and how well the models integrate into common workflows like Blender or Fusion 360.

@thad_hughes_imbue it has been great to see you evolve this work on container setup over the last year from just something for us, to something anyone can use

Proud of you @thad_hughes_imbue for the work shipping this! I've learned so much from using Keystone as a benchmark to understand cost, performance, and failure modes of Claude Code vs. Codex vs. Opencode.

The idea of codebases becoming more autonomous is really interesting.

Feels like we’re slowly shifting from writing code to orchestrating systems.

Hey Stas! This is genuinely useful - the kind of tool that makes you wonder why it didn't exist five years ago.

It solves something I didn't realize had a name. I've been calling it "the Loom paradox" - the recording takes 5 minutes to make and 25 minutes to extract meaning from, which means it's faster for the sender and slower for everyone else. The fact that you're turning that into structured tasks with individual clips is a real unlock.

How does it handle the moment where someone backtracks mid-recording? "Actually wait, ignore that last part, the real issue is..." - does it catch the correction or does it dutifully create a task from the thing you were told to ignore?

Turning a screen recording into structured tasks is the kind of friction removal that sounds obvious in hindsight. Does it work with any app or only specific tools?

The "someone please turn this into tickets" problem is painfully real. I've sat through 20 min Loom videos where the actual bug is in the first 30 seconds and the rest is just someone thinking out loud.

For the integration question, Linear would be my pick. Most teams I know that move fast are already on it and the API is clean enough that you could probably push tasks there with minimal friction.

Does it work with live calls too or just recordings?

Very impressed! For teams with sensitive product info, how is data handled? Are the screen recordings and transcriptions stored on your servers, and if so, for how long? Can admins set policies like “auto‑delete after 30 days”? And do you support SSO for enterprise teams?

Love this! Does it work on any type of website?

This sounds interesting. I always wonder with tools like this, how much control do you actually have over what the AI is testing and changing?

i spend too much time checking updates manually, this kind of system would handle that silently in the background.

What's the minimum scheduling interval you support and is there built in retry logic for failed jobs? Great idea bringing cron jobs to AI agents, congrats on shipping!

Hey everyone 👋 Maria here, co-founder of Maritime.

Over the past year, I’ve spent a lot of time working on AI agents through MIT and the OpenClaw ecosystem, and one thing kept coming up again and again:

Building agents is getting easier. Deploying them is still a pain.

A lot of people have something cool running locally, but getting it live in the cloud in a reliable way, without dealing with infra headaches, scaling issues, or paying for a server that mostly sits idle, is still a real blocker.

That’s why we built Maritime.

Maritime is a simple platform for deploying and running AI agents in the cloud. No DevOps, no complicated setup, just get your agent live and keep it running.

Right now, you can:

• deploy OpenClaw, ZeroClaw, or your own custom agents

• run multiple agents at once

• manage everything in one place

• scale without having to think too much about infrastructure

We also wanted to make it genuinely accessible, so pricing starts at $1/month. The whole point was to remove friction for builders who want to experiment, ship, and learn fast.

We’re currently in beta. If this sounds useful, sign up for the waitlist and include a short note on what you want to build. We’re reviewing everyone manually and will reach out to the top users over the next 2 weeks with beta access.

If you’ve ever built an agent and thought, “I don’t want to spend $20 a month just to keep this thing online,” we probably built this for you.

Would really love your feedback: what makes sense, what feels confusing, what’s missing, and what you wish existed. I’ll be around in the comments.

curious about how the pricing changes as request scales. also, what security measures do you have in place for data security?

$1/month is wild. What agents can I run on this?

$1/month for agent hosting is a no-brainer price point. The biggest friction with running AI agents right now isn't building them, it's keeping them alive and accessible somewhere. This solves that cleanly.

How does it handle scaling when an agent suddenly gets a burst of traffic? Like if someone shares your agent link on Twitter and 500 people hit it at once.