PH热榜 | 2026-05-13

@terrence_wang1 with the explosion of so many pure software related tools, I sometimes forget how great a physical product can be. This is great, the use case is so clear. I think a lot of people will be updating their amazon wish lists. Congratulations on the launch!

@terrence_wang1 This is really interesting 👀 turning real conversations into useful notes and actions is actually a big problem, especially for founders and small business owners.

I like the idea of reducing mental load and not losing track of things people say in meetings and calls. That’s where most tools still don’t fully help.

Curious how it works in real, fast conversations where things move quickly.
Wishing you a great launch 🚀

@terrence_wang1 How does Memoket Gem handle on-device processing vs cloud for conversation capture to ensure privacy during sensitive client convos, and what end-to-end encryption does it use for summaries sent to tools like Notion/Slack?

@alice_lee0917  @genedai Thank you! And honestly, that's exactly the journey that led us to the wristband form factor. We looked at pendants, clips, cards, phone apps, and kept coming back to the same conclusion: if it's not something you already naturally wear, you'll eventually stop using it. A wristband just disappears into your day. You put it on in the morning and forget it's there until you need it. Really hope the Gem is the one that sticks for you. 🙏

@alice_lee0917  @genedai Thanks so much, Gene! Really appreciate the support. That’s exactly the problem we’re trying to solve, something that fits naturally into your day without feeling like extra gear to manage. Excited to hear what you think once you try Memoket Gem!

@alice_lee0917  @genedai Exactly! We wanted to build something that fits naturally into the routine without the user feeling he/she has to bring or wear something else. And using the phone is really not convenient especially when you are driving or on the go. Hope the Gem can really help people like us :)

@jocky Thank you so much for your support. Happy to hear that the app is not disappointing!

@jocky Thanks so much for your support! It's great you have tried and liked the App! Hope you are using more features from the app and you can share with us your feedback or input anytime. We have a very active Discord and Slack community where early supporters give us very constructive ideas on a daily basis. Again thanks a lot for supporting Memoket!

@jocky Really appreciate this, and thanks for testing it early! Shipping the app first was important to us because we wanted people to experience the core AI memory workflow before the hardware arrives-capturing conversations, extracting summaries and action items, and making them useful right away. Gem builds on that foundation by making the capture side even more effortless and always available. Thanks again for the support!

@alexia_li Thanks Alexia! Founders were definitely one of the core use cases we had in mind, especially those days packed with meetings, calls, and quick conversations in between. We’ve had beta users wearing it as part of their daily routine, and the strongest feedback so far is that it works best when it simply fades into the background and captures context as the day happens. Really appreciate the thoughtful question! 🙏

@alexia_li Hi Alexia, we are just opening for beta users but the founding team have been trying it (the prototype) daily for over a month now. I myself wears it with my Apple Watch and it's really comfortable and convenient. Hope you guys find it useful like we do.

@alexia_li 100%! Around here we joke that the only way to make us take the Gem off is by force. Even showers feel like we're missing context 😂

@ann_y1 Thanks, Ann! That’s exactly the kind of use case we had in mind. Customer calls and casual user chats often contain the most useful details, but they’re easy to lose if you write notes later. Gem is meant to help you capture those moments intentionally and review the key takeaways when you need them.

@ann_y1 The "write notes later and forget half of it" struggle is so real. That's exactly what got us building this in the first place. If you end up trying the app, we'd love to hear how it fits into your workflow. Thanks for checking us out!

@abhay_agarwal5 Great point, we agree that intentional capture vs. always-on passive recording is a core trust design choice.

Gem is designed around intentional capture: recordings are first stored locally on the device, then transferred to the app via Bluetooth or Wi‑Fi and stored on the phone. When users choose to generate transcripts, summaries, or action items, the relevant data is uploaded to AWS for server-side processing, with encryption in transit and storage.

We’re currently a consumer-facing product in beta, and we know professional / enterprise use cases come with a much higher bar around security, compliance, and data governance. Those requirements can vary a lot by company, and we’d be happy to keep learning from teams like yours as we evolve.

@abhay_agarwal5 Really thoughtful question, Abhay. And like Terrence said, intentional capture is a core principle for us, not just a feature toggle. We made that choice early because we believe the product only works if people trust it.

On the enterprise side, we hear you. We're learning a lot from conversations like this about what professional teams actually need. If you're working in a setting with specific compliance requirements, we'd love to hear more. That kind of input shapes what we build next.

@othman_katim Great question, Othman. It’s a bit of both: Gem uses our own algorithm to identify what seems important, useful, or actionable in a conversation, and users can also guide it over time. For example, if a client mentions a follow-up, a timeline change, or a key decision, Gem is designed to surface that as a key point or insight rather than treating the whole conversation equally. The goal isn’t to capture everything, it’s to help the moments that matter flow into your workflow with less effort.

@victor_maximo_arredondo_solis Haha appreciate the trust 🙏 The Anker DNA runs deep in this team.

To your question, your billionaire ideas and BBQ grocery list are safe with us. Recordings live in your account, accessible only by you, and you control who sees what.

A few more things baked in by design:

1. Gem is not always-on. Recording starts with a deliberate press, so you're the one deciding what gets captured (and what stays in your head until the billionaire idea is fully baked 😄).

2. There's a clear red LED on the side of Gem that lights up whenever it's recording, visible to everyone in the room. No hidden recordings, no surprises.

We follow local consent laws as the baseline and encourage users to inform participants. Privacy by design, not by policy.

So yes, feel free to plot your empire and your weekend cookout. Memoket will remember the context, but only for you 😄

@zhengyang_hou Thanks Zhengyang! That means a lot. We’ve heard this from quite a few busy users, the magic isn’t just capturing conversations, but helping turn them into clear next steps before things slip through the cracks. Really appreciate the support!

@zhengyang_hou The juggle is real! And yeah, conversation-to-task has been a game changer for us internally too. So much gets decided in a quick call that never makes it into a to-do list. Now it just does. Thanks for the support! 🙏

@zhengyang_hou Thank you for your support! Oh yeah, you can sync the tasks directly to your Google Calendar and Apple. Reminders, which is actually one of the favorite features from our early supports in the community. You've truly discovered a hidden "gem" ;)

@wenjun_shi That's exactly how we feel too! Once you stop re-explaining everything to your AI tools, you realize how much time and mental energy was going into that gap. Thanks so much for the support 🙏

@wenjun_shi Thank you Wenjun! Yes the gap between what you discuss and what your tool knows is exactly what we want to bridge with the Memoket Gem! We hope it will be helping more people like us.

@wenjun_shi Thanks so much, Wenjun! You put it really well, that gap is exactly what we’re focused on. So much valuable context is created in the moment, and our goal is to make it flow naturally into the tools and workflows people already use. Really appreciate the support! 🙏

@campritchard Thanks Cam, really appreciate this! That’s exactly the part we’re most excited about too.

Recording is only the starting point. The real value is helping people connect context across conversations over time, so useful details don’t get lost between meetings, chats, and ideas.

Thanks for the support!

@campritchard You nailed it, Cam. Recording is table stakes at this point. The real magic is when you can look back and see how a project or relationship evolved across multiple conversations without having to piece it together from memory. That's the part we're obsessing over. Thanks for the support!

@bruce_warren Thanks for the great question!

The honest answer: AI tools are only as smart as the context you feed them. Right now, that context lives mostly in our heads or in scattered notes. Every time founders, SMB owners, knowledge workers, or anyone wanting to use ChatGPT, Claude, or any AI tool to help with real work (make a document, help develop a product launch plan, etc), they have to retype, retell, and re-explain what's actually happening in their business and lives. The strategy call from yesterday. The client objection from last week. The hallway decision that changed the roadmap. None of that is in your AI's memory by default.

So the question isn't really "what should AI remember." It's "what context do you keep having to feed AI from scratch?"

For us, the answer is: the conversations that actually drive decisions. Client calls. Customer feedback. Hiring interviews. Quick coffee chats with co-founders. Investor follow-ups. These are the moments where the real signal lives, and they're the moments AI is currently blind to.

That's what Memoket Gem is built to remember. Not everything you ever say. Just the conversations that matter to running your work, captured intentionally with a single press and connected over time, so your AI tools finally have the full picture instead of just the version you can type out from memory.

@worklab Thanks for the interest and sorry about the shipping limitation! We're working on expanding to more regions but India isn't available just yet.

Here's what we can do: shoot us an email at hello@memoket.ai and we'll figure something out together. We want to make it work for people who are genuinely excited about the product.

In the meantime, you can already start using the Memoket app to capture and process your conversations:

• iOS: https://www.producthunt.com/r/KXBVBSSOJLZEFX?ref=producthunt

• Android: https://www.producthunt.com/r/P3MX6RBZJUUHBF?ref=producthunt

@em_ebert Thank you for your support today!

@em_ebert Thanks for supporting Eric!

@em_ebert Thank you very much for the support!

@hmadhsan Thanks so much, really appreciate the kind words! 🙏 Two great questions, so let me take them one at a time: On privacy and consent: This is the question we take most seriously. A few things baked into how Memoket works: First, Recording starts with a deliberate press. You decide what's worth capturing, not the device. That intentionality matters, because it puts the user (not an algorithm) in control of every recording session. Second, a clear red LED on the side of Gem lights up whenever it's recording, visible to everyone in the room. There's no hiding it. We deliberately designed it so that anyone around can see when Memoket is active, which makes the consent conversation easier and more honest. Third, we follow local consent laws as the baseline, and we strongly encourage users to inform participants when recording. Your data is yours, recordings live in your account, accessible only by you, with permission-based controls. We know the wearable AI category has had some justified scrutiny on privacy. We want Memoket to be the example of how to do this right. Privacy by design, not by policy. On long-term memory and context retention vs. other AI wearables: Most AI wearables in the market do one thing: capture audio, generate a transcript, maybe a summary, and stop there. Memory ends at the meeting boundary. Memoket Gem is built differently. After capture, the device compiles and connects information across every conversation you've had — over days, weeks, months. So: A week of customer calls becomes a single brief on what your team should fix. Three weeks of candidate interviews become a side-by-side hiring decision. A month of partnership negotiations becomes a clear timeline of what was promised, by whom, and when. We call this the context layer — your conversations stop being scattered transcripts and start being living, structured context that you (and your AI tools like ChatGPT, Slack, Notion) can actually use. That's the real difference. We're not a note-taker. We're the bridge between your real world and your AI. Thanks again for being here on launch day 🙏

@xfei Thank you so much for your support! Glad you liked the Memoket App. And yes, the Memoket Gem will make a lot of difference! I've been wearing it everyday myself. We'll ship out the beta prototype as soon as possible :)

@xfei That notebook has served you well. Time to let it rest in peace 📓🪦 Seriously though, construction is the perfect use case. Back-to-back site visits, client calls, supplier negotiations... that's a LOT of context to keep in your head. Memoket's got your back now. Thanks for the support!

@xfei This is such a great real-world example! Construction is exactly the kind of environment where details move fast and context gets scattered across calls, site visits, and follow-ups. Really appreciate you sharing how you’ve been using the app already. Can’t wait for you to try Gem in an even more hands-free workflow!

@libin_yao Thank you! SMB owners are exactly who we had in mind when building this. When you're wearing every hat and running from one conversation to the next, the details that slip through the cracks are the ones that cost you. Appreciate the support!

@libin_yao Thanks a lot Tina. Indeed a lot of our early supporters are SMB owners and founders, having to have a lot of meetings on the go and making lots of decisions between conversations as well. We truly hope Memoket Gem will help build the context for them so that they can focus on what matters and have piece of mind at the end of a busy day. Thanks again for your support!

@henry_kwon1 Good news, Henry! We actually designed a strap specifically for this. It holds the Gem right underneath your Apple Watch, so you wear both on the same wrist. No need to choose. We heard this concern early on and made sure Apple Watch users wouldn't have to compromise

@jesse_harper1 Thanks Jesse! Love that you're already doing this with your Apple Watch. The fact that you built a workaround tells us the need is real. We designed the Gem specifically for this, so everything from the hardware to the app is built around capturing and making sense of conversations. Would love to hear how it stacks up for a heavy user like you. That's exactly the kind of feedback that helps us get better.m

@imad_elkhafi Hi there thanks a lot for your question! Firstly, the user will decide when Memoket Gem starts and stops recording by pressing on the button on the Gem. Among our early beta users, a typical use case would be the user pressing the button to start recording when a conversation or meeting starts, and ends it when the meeting ends. If you do want the Memoket Gem to stay on, the transcript and summaries will provide clear outlines of each discussion topic so you know what's going on throughout the recording. Deleting a certain recording is very easy to operate on the Memoket App. We are still discussing on the "pause" feature and will keep our users updated before mass production takes place.

Thanks again for asking these great questions and for supporting Memoket!

@jinhao_bai2 Hi Jinhao. Thanks for supporting Memoket! Currently the Memoket App is free to download and use. We would like to get early feedback from real users so we can improve the app user experience and features before we launch. Please feel free to download the app from the Apple Store or from Google Play. When the Memoket Gem is launched, it will be used alongside the app. We are still finalizing the commercial plans and will keep our users updated via our official newsletter. Please feel free to check our website and sign up to our newsletter here: memoket.ai.

@chloesamaha hi Chloe, the Memoket Gem is designed to record 20 hours continuously. In fact we are working towards 24 hours as we speak :) If you are not a heavy user and just record 3-4 hours a day, it should last you a whole week. Hope this answers your question and let us know if you need any other information. Thanks for supporting Memoket!

@zrimko Great question, Zrimko. First, quick clarification: the Gem isn't passive capture. You actively press it when you want to record, so you're always in control of when it's on.

On the disclosure side, the Gem has a visible red LED that lights up whenever it's recording. It's intentionally obvious so everyone around you can see it's active. We wanted to make transparency a hardware feature, not just a setting buried in the app.

Appreciate you raising this, it's one of the most important parts of getting this right.

@phoenixhu You just described our exact target user! Supplier calls, customer chats, quick team check-ins... that's where the important stuff happens and where it usually gets lost.

The Good news: today's your day. We opened 50 Founding Member spots for our Product Hunt launch. You get a Gem for free, just cover $5 for shipping. Spots are going fast though, so grab one while they're still open! You can sign up here: https://memoket.ai/pages/product-hunt

@andika_fadhilah Thanks so much! Really appreciate the energy

Sounds like exactly the kind of workflow Memoket is built for: marketing tech, lots of IRL syncs, constant context switching between desk work and conversations. That's our people.

On the noise question, this is one of the things we've spent the most engineering time on. Memoket Gem has dual premium microphones with built-in noise reduction, designed specifically to pick up voices clearly in real-world environments.

We're actually putting this to the test right now. Over the past two weeks we've been using Memoket Gem at a few trade shows and events ourselves, and the results have been really encouraging. Still plenty to improve, but at this stage pickup is reliable and accurate even in pretty noisy spaces. Our beta testers have been giving us consistent feedback that the captures hold up in busy cafés and in meetings with 10+ people.

The combo of two sensitive directional mics plus our audio processing layer has been the difference. It's not magic (really loud spaces will always be a challenge) but for a typical café catch-up or hallway conversation, the captures come through clean enough that summaries and transcripts are accurate and realiable.

@qiwap Thank you! You captured the idea really well. We think the real value goes beyond a single meeting summary, it’s about building a memory layer across conversations so teams can understand how a project, customer relationship, or decision evolved over time. That context is especially hard to reconstruct later, and exactly where we hope Gem can help founders, sales teams, and PMs stay aligned. Really appreciate the thoughtful comment!

@qiwap Spot on, QIQI! That cross-conversation context is honestly what gets us most excited internally too. A single summary is useful, but when you can trace how a decision or a client relationship evolved across weeks of conversations, that's where it clicks. Glad that stood out to you.

@linglistack For sure! That phone call interruption is SO relatable😅. We've all been there, mid-recording and then boom, someone calls and kills it. That's exactly why we went with a dedicated wearable. It does one job and nothing interrupts it.

@linglistack Yes, exactly! Making it feel natural was a big part of the design goal. Pulling out a phone can add friction or change the flow of a conversation, while a wearable can quietly stay with you and help capture useful context in the moment. Thanks so much for sharing this!

@emma_zhang5 Totally hear you! Wrist real estate is real. Gem is designed to co-wear with Apple Watch, so you don’t have to choose between them. It’s lightweight and meant to sit comfortably alongside your existing watch. That said, we’re also thinking about more flexible form factors over time, and a modular or strap-based design is a really interesting direction.

@janicelewis00 Thanks Janice! Indeed as a marketer I've used Memoket a lot for consumer research. It can give me a summary of what my users have told me across a whole week's interview sessions. Super convenient :)

@janicelewis00 Exactly! That weekly roll-up use case is one we’re really excited about. A lot of valuable context gets spread across research calls, team meetings, and reviews, and Gem can help turn that into key points, patterns, and action items instead of making you rebuild everything manually at the end of the week. Saving those 2 hours is exactly the kind of outcome we’re aiming for. 🙌

@janicelewis00 2 hours saved every week is 100+ hours a year, Janice. That's a lot of Fridays back! Thanks for the support!

@davitausberlin Thanks Davit! 🙏 Great question and actually it's something we thought about a lot from day one. We don't have plans to add fitness tracking ourselves, mostly because we wanted to stay laser-focused on doing one thing really well: capturing context from conversations. That said, you don't actually have to choose between two wearables. We designed Memoket Gem with a custom Apple Watch co-wear band, so you can keep your Apple Watch (or any fitness tracker you already love) on the same wrist, and Memoket sits comfortably alongside it. One wrist, both jobs. The bigger philosophy: we don't want to replace what you already wear. We adapt to your life, not the other way around. More co-wear options for other wearables are on our roadmap 👀

@davitausberlin Hey Davit! Like Terrence said, "one wrist, both jobs" is the philosophy. And honestly, we'd rather be the best at capturing conversations than a mediocre fitness tracker. Your Apple Watch already does that part way better than we ever could. We just want to sit next to it and handle the part it can't

@latitude  @heycesr Congrats. How does Latitude break down token costs across subagents and tool calls; e.g., does it attribute input/output separately per turn, and can it export traces for deeper analysis like custom dashboards?

@nathan_tran2 both are supported!

@robert_douglass Thanks, taking a look asap

@olivia_bennett7 It adds no overhead at all. Telemetry data is sent asynchronously and adds no latency to runtime sessions.

@xavier_hernandez2 Both

@michael_vavilov We only support Claude Code at the moment. But if you want to hit your limits less, I suggest you to try Claude Code. It's not that different from Cowork and it can do the same things without spending as many tokens.

@tham_yikfoong what happens if local machine runs out of storage for logs/memory? does it have a way to auto-prune old events?

@tham_yikfoong How does CraftBot ensure data privacy and state isolation when importing GitHub repos into Living UI; e.g., sandboxing mechanisms or LLM guardrails to prevent the agent from accidentally mutating sensitive codebases during iterative evolutions?

@nelson_milla Notion is a part of our existing external integrations. You can easily set it up using OAuth or creating your own notion app.

You can ask the agent to create a living UI connected to your Notion account and fetch data in real-time in order to keep the living UI and notion continuously in sync.

@busmark_w_nika Thanks Nika!

@tham_yikfoong  @vikramp7470 Craftbot has a list of external tools and we're continuously adding more. These external tools are provided to the agent with a simple interface. They allow proactive use and can be easily integrated into the Living UIs without any extra work. We are continuously adding more external tools but other than these, yes the agent can integrate any API needed (you might need to set up API keys though).

@rupeshpandith It is actually something we have been considering to improve. Currently, the main agent create Living UI with a creator skill. However, having multiple sub-agents can save more tokens and stuff. Please look forward to our future release! This is the direction we are heading.

@artstavenka1 Glad you like it! Yes, we also prune it periodically, removing and merging memory depending on the context.

@lakshminath_dondeti Let me answer both of your questions here. The UI change when you ask CraftBot to modify them, and they are device responsive so it adapt well on the CraftBot browser interface whether it is opened in web or mobile. It can be both dashboards/apps, since dashboard is kinda the subset of apps.

@ytubviral Thanks for the comment! There are multiple ways to do this. Living UI can have their logic to run task on schedules, or CraftBot can setup a proactive task to do so! Since we are at here conversation, what dashboard or tasks are you looking to built with Living UI?

Fair push. Being honest: the roadmap still lists Proactive Behaviour as Pending, so today the answer is mostly the "with approval" step in the agent loop; closer to reactive-with-a-confirm than real calibration.

Where I think it lands, given the current architecture: the Skill Manager and Action Router are the right place for this, not a global confidence knob. The axis that matters is reversibility, not certainty; a Skill that drafts into Gmail and one that sends are different decisions regardless of how sure the model is, so the threshold should live in the skill manifest. Override-as-signal fits naturally on top of the Memory Manager (it already consolidates events at midnight), but probably as a gradient edit-before-send is soft, delete-and-rewrite is hard, disabling the skill is structural rather than one counter feeding back into a single threshold.

Still early, and open to where you've seen this work.

@ben_holding What's the API spec like for syncing with tools like HubSpot or Zapier, and how does the AI receptionist handle multi-channel handoffs?

@charlotte_reed1 Great question. Transparency is the foundation: every call, message, and action the AI takes is logged with full history, so you can see exactly what it did and correct behavior in one place. On top of that, you can define escalation pathways on every channel (voice, SMS, email, chat) so the AI hands off to a human based on rules you set, whether that's a specific keyword, a stuck conversation, or a high-value lead. The whole thing runs inside a defined scope (who your business is, what it can and can't say) instead of being a general-purpose agent freelancing on your behalf. Additionally, compliance is built into the platform, not bolted on. Consent, opt-outs, rate limits, and channel-specific regulations are all handled at the system level so a lot of "wrong moves" get blocked before they happen. No AI is perfect, but scoped + supervised + observable beats "general agent doing whatever."

@hudson_blake A lot. Before anything goes out, you define the AI's persona, knowledge base, and scope (what it can and can't say, when to escalate, what's off limits). You can run test calls and messages to yourself to see exactly how it'll behave with real customers. For outbound campaigns, you set the audience, content, timing, and rate limits up front and review the whole thing before it launches. On email specifically, the AI runs in draft mode where every reply lands in your queue for approval. And in any individual conversation, on any channel, you can flip the AI off at any moment to take over yourself. You can dial human-in-the-loop up or down per channel depending on how high-stakes the conversation is.

@cody_spencer Honest answer: most don't start by trusting it 24/7, but get there fairly quickly. They start with the AI handling overflow (after-hours calls, drafts only, missed leads) and expand its autonomy as it earns trust. The comparison that matters isn't "AI vs. perfect human," it's "AI vs. the current reality of missed calls and follow-ups that never happen." With toggles, drafts, full logs, and escalation rules per channel, trust gets built fast. Thousands of businesses run on Frontdesk today and we handle tens of millions of customer interactions a month, so empirically the answer is yes, with the right setup.

@dylan_hayes2 It's a replacement for most of our customers. Frontdesk is built for businesses stitching 5 tools together and bleeding money on integrations that never quite work, so the whole point is one system where the website, CRM, chatbot, lead forms, and AI receptionist share data and trigger each other natively (you can't really get that by bolting AI on top of 5 different vendors). Customers on our native CRM tend to get the most out of the system because that's where "smart variables" really shine: CRM fields the AI updates automatically based on each conversation, like preferred call time, service interest, or appointment status, so your records stay fresh without anyone manually entering anything. That said, if you've already got a serious investment in something like HubSpot or Salesforce, we play nice and integrate. The unlock is the AI front office layer, the underlying CRM can live elsewhere if you need it to.

@caleb_hunter_guahip Fully agreed, and honestly this is where a unified system gets exciting. The AI doesn't just stay reliable at scale, it compounds. Every interaction (AI-handled or human-handled) sharpens how it responds next time. Full customer context and history are always there, pulled from a CRM that updates in real time across every channel. You're essentially teaching your AI COO your business as it runs, with zero data stitching across vendors. A stack that is bolted together has a natural plateau in how reliably it can handle customers, but a unified system's ability compounds over time.

@adam_nathan1 Interesting product but I think the positioning is currently smaller than the actual problem being solved.

“AI marketing tool” is rapidly becoming a commodity category.

The real value here isn’t content generation.
It’s helping businesses avoid invisibility, inconsistency, and falling behind larger competitors without hiring an entire marketing team.

That’s a far more urgent and economically painful problem.

Positioning around output explains the product.
Positioning around consequence creates demand.

Saw a couple other high-leverage positioning gaps while going through the launch that could materially affect differentiation and perceived necessity.

@adam_nathan1 Love to see the evolution. Here's to a great launch!

@adam_nathan1 How does it technically handle noisy GA/social data (e.g., attribution gaps) to prioritize content pillars without over-optimizing on vanity metrics like likes?

@alan_cassinelli truth!

@hilmi_bou +1

@hilmi_bou Love to hear this! We're so glad Blaze is helping you!

@hilmi_bou awesome to hear this is getting time back for you!

@leo_llhomme Blaze is already helping thousands of small businesses show up consistently online, and it’s been incredible to watch that happen. Looking forward to seeing how many more we can help.

@leo_llhomme Great to hear this feedback, Leo! Thanks for sharing!

@leo_llhomme thank you so much for your support! So glad Blaze has been able to help you make amazing ads that grow your business

@sandra_baldwin1 So happy to hear this, Sandra! Really proud to have worked on this. We’re already seeing thousands of small business owners use Blaze to stay consistent with their marketing — and we’re just getting started. Excited to get this in front of more people here.

@sandra_baldwin1 really appreciate you sharing this — means a lot.

Glad to hear the ad creation is working well for you, and that the team was helpful getting everything dialed in. That’s exactly the experience we’re aiming for.

@liam_devine1 really appreciate this.

That’s exactly what we’re aiming for: not adding more work, actually saving you time.

Glad it’s been working for you.

@0xaron Blaze can write and autopost for you on X/Twitter accounts. And it can learn from your X/Twitter analytics. But no Discord (yet!)

@0xaron Not Discord, but it can do Twitter!

@ellen_reynolds1 Sounds amazing. Thanks for the real-world review, Ellen!

@ellen_reynolds1 thats incredible!

@blaze_eddie very true, our customers are the best!

@calvin_collins thanks for all your work on this Calvin, we really nailed it.

@susanne_shi_ this is incredibly thoughtful — thank you for taking the time to share all of this.

Really appreciate you sticking with Blaze across multiple workspaces and clients, and calling out both what’s working and where things still need improvement. The language + tone feedback especially is something we’re continuing to invest in.

Also glad to hear the simple design updates helped bring things back under control there.

@susanne_shi_ We appreciate you! Your feedback has been really helpful for the improvements we have made and will make soon!

@ytubviral Hey Javiar, great questions. In order to create great content, we first start with your business' goals, then choose a marketing strategy, and then within that strategy is a campaign around a specific theme. Before the content is created, we apply your brand voice and add any context from your brand kit that the AI needs to know. All of this ensures that you don't get a generic AI output.

The Learning Loop plays a part here too, after a few weeks it starts to learn what content resonates with your audience, and automatically adjusts the campaigns and topics based on performance.

As for bilingual support, right now you have to choose your output to be in one language. But I'll add that as a feature request for us to look into.

I encourage to to try 7-day free trial to check it out yourself!

@boyuan_deng1 we've solved this with 2.0. the AI does a first pass based off your website, and then you can go into the Brand Kit and update it to exactly what you want.

@kyle_seerins really appreciate this. Glad autopilot has been helpful, and great to hear branding is getting closer to where you want it. We’re continuing to push hard on reducing that review lift. Thanks for sticking with us!

@ethan_thompson3 it will go channel by channel before turning off everything - i think looking at the data some social media channel will work, but probably not all of them!

@halev thanks so much!

@campritchard totally! There are so many great small businesses that just need the help of consistently putting themselves out there.

@madalina_barbu thanks for checking it out! what caught your eye the most?

@owen_shaw2 thanks! Only connected platforms right now. We are building other surface areas to collect data (like leads) so the Learning Loop will be able to have that too

@imogen_wallace Every piece of content is customized to that business' brand kit (source materials, voice, assets, and preferences), but furthermore, each business can add or select multiple strategies in Blaze (and customize those too) so no two businesses will ever get the same content.

@abod_rehman You can give it source materials to train on, and then the Learning Loop will also adapt your content based on what is performing well

@paul_sangle_ferriere1 thanks!

@whateverneveranywhere 

Do agents see a unified shape

Yes, the agents see a unified schema/unified API response, which can be used for read/write operations.

@samz78 This is actually a really strong insight especially the idea that you're not just tracking tool usage but how agents choose composition vs raw calls.

Instrumenting everything via BossHogg-style telemetry is exactly what you need if you’re trying to understand real agent behavior instead of guessing it.

The most interesting part is that feedback loop: when agents prefer decomposed calls vs a single MCP fan-out, that’s basically telling you where abstraction layers are actually worth it and where they’re just adding friction.

@philip_kubinski thanks for the validation!

@philip_kubinski 

If I could have data in one place back then, it'd save my hours and a lot of frustration.

Yes, that is the pain point we're trying to solve with AI agents. We already have our unified API, and adding the MCP on top of it, you can easily combine data from different data sources, and then take that to create combined reports, take actions, etc.

What used to take hours can be done in less than <20 minutes.

@nevo_david thank you so much mate!

@nevo_david thanks Nevo!

@tomschouteden All the best!!

@natalia_iankovych Ahh, if we had launched earlier, maybe you could have used it. But now you do.

How did you approach it? And how did you solve this problem?

@jaidevxb you can already limit what permissions your MCP will have in the Apideck's dashboard.

@abhay_agarwal5 

The data normalization layer is where unified API platforms either win or quietly accumulate debt.

Yes, I agree, and you can test our unified APIs to get the idea of how we normalize data. You can already find the schemas and guides on our docs.

Hi Ihor, mutating data is the scary part. That's why we added scoped permissions before the agent can run anything destructive.

@sarah_andrabi Congrats on the launch!

@charlotte_reed1 good question! I’ve been using Liminary for a month in my consulting work and what I’ve noticed is that the more context you give it, the sharper it gets. For example, it recently linked a specific question a client asked in a meeting to a comment from a strategy audit I did weeks ago.

The technical reason it doesn't just pull "keyword matches" is that it uses more than just text similarity. It treats your access patterns and recency as signals too. If you're looking for something new that's related to an older note, the system treats that as a sign that the older note is still "alive" in your thinking. It also lets you dismiss or mark things as outdated, so the retrieval actually learns from what you find useful versus what’s just noise.

Basically, it’s designed to prioritize your current thinking over just anything that looks similar on paper.

@charlotte_reed1 +1 to Kevin. The other thing I'd add is more about how we approach the design problem than the technical one. Precision in practice is genuinely hard to measure, there's no clean metric for useful vs technically related, and it's also dependent on each user's individual preferences. So we lean on a couple of things instead.

First, we let you tell the system when something isn't useful, dismiss it, mark it as not relevant, and that feedback shapes what gets surfaced next time. The system learns your standard for "useful" as you use it.

Second, we deliberately surface less rather than more. It's easier to build something that throws every "related" note at you, but that's how you get the noise problem. Restraint is a design choice.

Still imperfect, and we're tuning it constantly. The bar we hold ourselves to is that a surfaced note that doesn't earn its place is worse than no note at all.

@hudson_blake That's a fair concern and honestly one we think about a lot. The research we did with consultants actually surfaced something counterintuitive: as AI models get better, hallucinations become harder to spot, not easier. When errors were frequent, people checked everything. Now that output quality is generally higher, the temptation to trust it goes up, but the stakes haven't changed. A single wrong stat in a client deliverable is still a professional liability.

The way we've approached this is to make verification the default, rather than an afterthought. Everything Liminary surfaces is tied back to a source you actually saved. So if something looks slightly off, you're one click away from the original document.

We're not asking you to trust the AI. We're trying to make it fast and easy to check it. The goal is to collapse what our users call the "reconciliation loop" - that painful cycle of generating output, hunting down sources, and verifying every line before it goes anywhere near a client.

@hudson_blake +1 to everything Kevin said. The other piece I'd add: Liminary is a content-first platform, not chat-first like most AI tools. That distinction matters a lot here. We don't respond from an LLM's general knowledge, we respond from the content you've already saved. And what you save required your judgment and expertise in the first place.

So everything downstream of that, what gets surfaced, what gets synthesized, what gets cited, is grounded in a corpus you already vetted. The AI isn't introducing new claims for you to trust or distrust. It's pulling from sources you chose, and showing you exactly which one. That's also why citations are a core part of the product, not just a polish item.

@jack_sullivan5 It's a real tension. The design challenge is relevance and timing vs volume.

What we're building is closer to a research assistant who speaks up when something is relevant to what you're working on right now. We notice that decision making tends to be faster when context is fully sourced and prioritized from sources you've deliberately saved.

If something comes up that isn't helpful, that's useful signal too - Liminary will learn from your feedback. The goal is a system that gets sharper over time.

@cody_spencer In reality, it gets messy. It's just how knowledge work typically happens -nobody has time to file things perfectly in the moment.

What we found talking to consultants is that the maintenance burden itself is what kills most knowledge systems. People start with good intentions and a clean folder structure, then three weeks into a busy engagement it's already out of date and they've stopped trusting it.

Liminary is built around that reality. It reduces the overhead of maintaining a knowledge system over time. You save things as you go, and the system does the organizational work in the background. The knowledge set emerges from your actual workflow rather than requiring a separate one to keep it alive.

@dylan_hayes2 Yes, you've identified what we think is actually the harder design problem. Retrieval is largely solved. Knowing what's relevant to the moment - and what isn't - is where the real work is.

It connects back to something Sarah said early on: people don't always know how to describe what they want, but it's almost always inferable from the context they're operating in. That's the principle Liminary is built around.

And when it doesn't have a confident answer, it says so rather than filling the gap with something plausible-sounding.

@matthew_barclay Thank you Matthew, this means a lot. The "context mess" framing really resonates, it's the exact problem that got me to start building this. Hope Liminary holds up to that promise as you actually use it, and please tell me when it doesn't.

@caleb_hunter_guahip It's a genuinely hard technical problem, and one the team has thought carefully about. When you save sources into Liminary, they don't just sit in a file store waiting to be retrieved. The system runs an extraction process immediately, building an understanding of the content that includes the relationships between sources - where things corroborate each other, and where they contradict.

So if you saved a client interview from six months ago and a more recent one where the same person has changed their view, Liminary doesn't flatten those into a single answer and present whichever ranks highest. It surfaces both, with enough context for you to see where the tension sits.

@michelle_dawson_silbernagel Thank you for sharing how you use Liminary! Wonderful to hear how it fits into your workflow. I’m especially glad the weekly summary is helping you spot those trends and contradictions - I agree this is where the insights can be surprising!

@michelle_dawson_silbernagel Michelle!! You've been with us through versions of this product that barely worked, and the fact that you're here saying this on launch day means more than I can put into a PH comment. Thank you!

@easton_grant Always-on implies a constant feed, which would absolutely add noise. What we're building at Liminary is closer to ambient context.

Our goal is to minimize the cognitive overhead - focused, high-quality context when you need it, not a constant stream to distract you. The quality of the context that is surfaced is where Liminary can prove its value.

@easton_grant Always-on is definitely a challenge, and we spent a lot of time balancing the utility vs distraction question. Like Kevin said, the goal isn't to be on for the sake of being on, but to be available in-context when you actually need it.

A good example: we built controls so users decide when to run fact-check or gap detection while writing, instead of firing those pre-emptively. Surfacing them uninvited can break flow, even when the insight is useful.

Still something we're learning and want to tune per user, because everyone's threshold for "helpful nudge" vs "get out of my way" is different.

@easton_grant Hi Easton, It's a big design challenge to present more information in a way that's not distracting, yet there when you need it. When our goal is providing value instead of asking for engagement, we're at an advantage. We're constantly simplifying language, creating clear information hierarchies and tweaking our model's instructions so the user can choose what's helpful to them.

@campritchard Cam, you summarized this better than most of our own marketing copy honestly. "Guessing your context half the time" is exactly the failure mode we're trying to design out. Thanks for the support, and congrats on Station!

@lakshminath_dondeti Great question, especially given how recent Dreams is. The short version: they live at different layers and solve different problems.

Dreams is infrastructure for developers building agents on Claude. It's a scheduled process that reviews an agent's past sessions, extracts patterns, and curates memory so the agent gets better at its task over time. The user is the agent, in a sense.

Liminary is the user-facing analogue, but for you. The system continuously builds a memory of how you actually work, what you save, revisit, ignore, connect, ask about, and uses that to shape what gets surfaced when you're thinking through something. That memory isn't tied to one agent either, it's shared context that every agent inside Liminary draws on, so personalization compounds across the whole system. The user is you, not an agent serving you.

And you're right that this layer should be LLM-agnostic. We're built that way intentionally. Liminary uses multiple model providers under the hood (Claude, GPT, Gemini, Nova, other open source models) for different tasks, but the memory and retrieval layer is ours and sits independent of any one of them. Concretely, if you want to chat with ChatGPT, Claude, or Gemini about your content, you can. All three model families are available to chat with inside the product. The thesis is that your knowledge layer shouldn't be locked to whichever model company you happen to use today, because the model you use will change, and your knowledge should compound across all of them.

@whateverneveranywhere Good question. Short version: neither pure embeddings nor a reranker on top of them. The architecture is built to avoid that failure mode upstream rather than patch it downstream.

Two pieces. At ingest we run an extraction process that builds structured understanding of each source, so retrieval isn't operating on raw chunks. At query time, the answering layer is built to consume that structured understanding, not a top-k pile of snippets ranked against the question.

So you never get the "longest snippet wins" failure because nothing in the system is choosing between similarly-embedded snippets and hoping the right one floats up.

Thanks@thamibenjelloun! And yeah you nailed it. Storage for the sake of saving isn't the point. Finding at the right moment and finding the most relevant thinking is the problem we're trying to solve.

We're carrying over a lot of lessons from working on retrieval at Dropbox. Couple big ones we leverage are that recency is a strong signal, but so is access. When you look for something new that's related to an older note, that tells the system the older note is still alive in your thinking, even if you haven't touched it in months. It earns its way back up.

Also as users curate collections we treat those as living, not append-only, so pruning and regrouping is part of the workflow, not a chore bolted on top. Updates supersede instead of piling up, so when you rewrite a note the new version is what gets retrieved and the old framing doesn't keep haunting you. And you stay in the loop. When Liminary surfaces something, you can dismiss it, edit it, or mark it as outdated, and that feedback shapes what shows up next time.

Honestly hygiene is a hard, ongoing problem, and I'd rather make curation lightweight and continuous than pretend the system can fully self-clean.

@ytubviral Agent coordination with regards to memory is genuinely a fun problem to solve. Years of building search and retrieval at Dropbox taught us a lot about how this breaks down at scale, and we built Liminary's architecture and memory systems with those lessons in mind.

On the conflict question specifically: at ingest, an extraction process builds structured understanding of each source, including where things corroborate and where they contradict. So if you saved a client interview six months ago and a more recent one where the same person changed their view, Liminary doesn't flatten them into a single answer. Both surface, with the tension visible and citations back to each source.

One thing that helps on the coordination side: the memory layer is shared across every agent inside Liminary, not partitioned per agent. So there's one source of truth underneath, not many that need to negotiate.

Hi @natalia_iankovych  Liminary doesn't watch your ChatGPT activity in the background, nothing gets captured unless you choose to save it. So you stay in full control of what goes in.

That said, ChatGPT chats are very much something you can save via our browser extension. Anything you'd want to keep, a useful answer, a back-and-forth that helped you think something through, a research thread, you can capture it into Liminary so it's there later when you need it.

And it's not just ChatGPT or local files. You can save web pages while browsing, upload from your computer, pull in from Google Drive, and record Google Meet meetings. The idea is to capture the things you actually use to think and work, wherever they live, in one place that surfaces them back to you when relevant. Additionally if you want to chat with ChatGPT, Claude, or Gemini about your content you can. Liminary has all three model families available to chat with in product.

@jim_jeffers that's a really good question. Every answer Liminary generates is tied back to the specific saved sources it used. Not a generic list of related notes, but the actual sources that influenced what got synthesized, with one click back to the original. So you can always see which context won per se, not just trust that the right one did.

On the ranking question, recency is a strong signal but not the only one. Access patterns matter too. If you've been pulling on an older note recently, the system treats it as still live in your thinking. And updates supersede, so when you rewrite or revise a note, the new version is what gets retrieved. A random older note only outranks current policy if you've kept engaging with the old one and let the new one go stale, which is usually a signal worth surfacing anyway.

The other layer is that Liminary builds a memory of what you're working on right now and your preferences over time, so retrieval gets tailored to you specifically. The same library of sources can produce different answers for different users, because what's most relevant depends on the work you're in the middle of.

We currently don't yet have a way for users to explicitly mark a source as the authoritative version, but that's definitely food for thought for us. Right now it's inferred from signals rather than declared, which works well in practice but isn't as legible as it could be.

@jim_jeffers It's a difficult problem for sure. As you suggest, recency and staleness signals help here; not just when you saved a source, but when did you last reference it, and does it include date information in the content itself. The other very useful bit is that Liminary remembers your previous working sessions. So if you've said once that a piece of information is out of date, Liminary can use that in future sessions.

@sinchana_v We have strict scoping of sources by collection -- Liminary won't use sources from one Collection when working in another, and using one collection per client is a common pattern among our users. It's not strict host-based isolation, if your contracts require specific technical measures, but it will prevent leakage.

@sinchana_v +1 to Tom's answer on the product-level scoping, that's how engagement-level isolation shows up in practice for our consultant users.

Adding the infrastructure side since you asked. Processing isn't local, it runs in our cloud, but the architecture is built around isolation and encryption from the ground up. We run on AWS with tenant data scoped per account, end-to-end encryption in transit and at rest, and KMS for key management. So while it's not on-device, the substrate is enterprise-grade rather than shared general-purpose infrastructure.

On your bigger point, engagement-level isolation mattering more than document-level permissions, I think you're right, and that's why we built the unit of organization to be collections rather than tags or folders. Collections are the boundary, and Liminary respects it across retrieval, synthesis, and surfacing. Permissions on individual documents would be much weaker because the retrieval layer would still cross boundaries when answering questions.

Curious, what kind of client work are you running into this with? Some industries push harder on this than others, and I'd love to understand where you're seeing the friction.

@mark_g4 Multi-project/session management for Claude Code honestly feels like a real pain point now that people are running agents across multiple repos all day.

The auto account switching + marketplace layer is especially interesting feels like like a wrapper and more like an operating layer for AI coding workflows.

@mark_g4 Have you noticed users caring more about the multi session management side, or the Marketplace /community aspect of Claudy so far ?

@conduit_design That's a fair point! The real issue is the gap between Pro and Max — Pro's limit runs out too fast for serious coding sessions, but Max is overkill (and expensive) for most solo devs. A lot of users end up with 2-3 Pro accounts instead of 1 Max because it's cheaper and gives more total usage. Claudy just makes that workflow seamless instead of logging in/out manually.

@raydotsh yeah this is getting insane

@raydotsh Haha right? 😅 It wasn't really planned — but when Pro runs out mid-session and Max costs 10x more, devs just naturally ended up with multiple Pro accounts. Claudy didn't create that workflow, it just made it less painful. Hopefully Anthropic ships a mid-tier plan someday and makes this whole thing unnecessary!

@abhay_agarwal5 Thanks Abhay, really glad the mental model resonates! To answer your question — the labeling happens entirely locally using git trailers. When Claudy creates a draft commit, it adds a Claudy-Draft: true trailer to the commit message. No API call, no AI classification — it's just a git convention. This means draft commits are regular git commits that you can inspect, cherry-pick, or manipulate with standard git tools. When you're ready to finalize, Claudy squashes all draft commits into one clean commit so your history stays tidy. Everything stays local and offline-capable.

@qifengzheng you can config it on the project setting

@panphilov Not in any simple way. You could try to use Apple Shortcuts or Apple Script for that, but it won't work perfect, as Apple Notes has a closed format.

@qifengzheng Then try SideNotes.

It can display the same window on each display. However, only one window at a time.

@ganesh_kumar_t good question! Pipali can interact with your filesystem, run terminal commands and control your chrome browser pretty seamlessly. But for desktop apps without MCP tooling or terminal controls, it doesn't have computer use yet (we need better safety, control and seamless handoff mechanisms first)!

Hey @sailikhith thanks for the thoughtful question! Glad you like the Skills and Routines features in Pipali.

The conflict issue is real and can cause conflict at both the data and the UX layer. Pipali asks you when it wants control of an app, so you can choose when to hand over control to a shared app you're working on. It wouldn't just override your work without you knowing about it. You can also give it access to an isolated instance of the app (e.g Chrome with separate profile) or use a collaborative app (like Google Docs) to work with it in parallel.

Is there a specific workload you're wondering about where this cause problems for you?

@oliver_hayes1 Great question.
We provide you with a trend feed, you decide what trend is relevant you can ride on.

Layerproof job is trying to match your marketing angle with that trend to create the final post.  So you are the one who filters which trend works and which does not.

Essentially, you are the noise filter tbh.

We don't have it yet, but you give me a great idea, I will try to test it out, maybe indexing current trends with vector embedding and allow LayerProof to search for the top relevant trend in a period of time for your social post angle

@oliver_hayes1 my honest take is experts know what's relevant better than any algorithm does. So we let you scan the feed and pick the trend yourself, then do the work of turning it into a post that fits your angle. Matching is on us, filtering stays with you.

@oliver_hayes1 
Right now we're keeping it simple by show the current trends, then you pick the one that fits your work and we adapt from there. Eventually we want smarter matching (indexing trends) against your draft so the system can surface what's most relevant, but that hasn't shipped yet.
The simple version has already been more useful than we expected!

@miles_anthony2 mostly are solo creators right now, because that's where most of our users are, experts publishing under their own name.

we haven't built much for teams managing multiple brand accounts yet, mostly because we don't have the right partner on that side to design it well.

curious if you're asking because you know someone running that setup. if you do, we'd love an intro :wink:

@naomi_florence1 honestly, we're no different.

our team struggled with the same thing that I couldn't even cleanly pitch investors on whether we were a text tool, a design tool, or a distribution tool, while most successful tools pick one and go deep on it.

what we've done instead is build the infrastructure to spin up and launch a new niche product in days, not weeks. so instead of guessing which problem to solve, we ship smaller products fast and let the market tell us where to go deeper. Layerproof Matte is one of those. more coming, and the signal we get from each shapes where we double down.

@khanh_phan8 
If you want just a minor fix, you can now just select that specific part and edit it, the rest of your design stays exactly how it is. You just need to tell Layerproof want you want to change!
One thing I think you will like about Matte 2.0: you can generate different creative directions in 1 click. That way you alway have multiple options or varieties to choose if you want to switch things up:)

@akshay_deep_kanda 
Hi Akshay! Let me help you with that.

With your project opened, when you click on 'Add Variant' or start a 'New Post', a menu will pop up. Look at the 'Type of Images' at the top of the menu and select 'Carousel post', from here you can also choose how many slides you want and pick your aspect ratio.
I've attached a screenshot, hope that helps!

@kanng300 yes we do (it's literally on the landing page and the featured image, lol). there's an option to generate carousel posts but you'll want to manually check the title and content on each slide to get the best out of it though.

don't fully rely on the model, i'd recommend a quick read-through before posting.

@kanng300 
It absolutely does support generating carousels!
Layerproof focus on both visual flow and storytelling logic to make sure the result is ready to post. Give it a try, i'd love to know what you think about it

@oliver_hayes1 Hey Oliver. Thanks for the great question!

Accuracy in transcription and note-taking actually surprised me, and I think it's on par with cloud models, especially for English.
For other languages like Portuguese, some of my users said they were pretty impressed with the results. Cloud models were still slightly better, but they still prefer this option because the privacy and cost benefits outweigh this small drop in quality. Hope it helps :)

Hey @noah_bennett5!

The app supports 26 languages, including English and all the major European languages such as Spanish, German, Portuguese, French, Italian, and Russian, among many others.

Generally, there are no problems with accents. For example, it easily understands Brazilian, Portuguese, or strong English accents, and it handles conversations mixing languages without issue. However, currently, the app does not support transcribing meetings in Asian or African languages.

@lakshminath_dondeti Hey!

The application does not join the meeting. Instead, it runs locally and provides you with a transcript and generates meeting notes.
No video is recorded; only the audio is stored. You should ask for permission from the other meeting participants before recording.

Hey@whateverneveranywhere! Thank you for the nice words. The app actually records both the microphone and the system audio, so it captures all the sounds coming in and out of your MacBook, providing you with a complete picture of the entire meeting.

@qifengzheng the transcription is done completely offline on your MacBook. You just download the app first and then download a 500MB model that does the transcription. This way you generate your transcript locally without sending it anywhere else.

If you then want to generate the meeting notes, you can also download a local LLM that runs on MLX

@charlotte_reed1 A few things going on under the hood.

Every read sees a consistent snapshot, so a single query never trips over a half-applied change.

All edges carry first_seen and last_seen timestamps internally. Nothing is overwritten. When something flips (domain switches MX, IP moves to a new ASN), the old edge gets its last_seen stamped and a new one is appended.

Writers (BGP, DNS, all threat feeds, etc.) are pushing continuously. Reads work against a coherent slice. We traded strict ACID for immutable history plus fast appends. The payoff: an agent can ask "what was true when the incident started" rather than only "what is true right now."

Soroush built the fantastic math behind engine updates. Happy to introduce if you want to go deeper on the consistency model.

@hudson_blake  Single lookups sit at sub-millisecond p99 and we've load-tested at 120K queries per second.

For chains specifically, here's the thing: a lot of what looks like a chain in REST world collapses into a single Cypher traversal in our graph. "Find all domains sharing this name server, then their WHOIS owners, then any with threat-feed hits in the last 30 days" is one query for us, not three round trips. So the latency budget you'd normally spend chaining mostly disappears.

Please hammer it. MCP endpoint is free for launch week and we want to see real agent traffic. If anything looks weird, ping me here or DM Soroush.

@dylan_hayes2 Honestly it's huge.

You're not wiring up 20 named REST tools, so your agent doesn't burn turns figuring out which endpoint to call next. Cypher is one endpoint and your LLM already speaks it. And you're not teaching the model what a Domain or an IP or an ASN or an MX record is, because it already knows. Context savings on both the tooling surface and the vocabulary.

In Claude (or any LLM for that matter) that means a five-hop investigation that would normally need 10+ REST calls plus a wall of intermediate JSON becomes one Cypher query and a clean result set. The model spends its context budget reasoning instead of bookkeeping.

Try it on a real investigation. You'll feel the difference within the first pivot.

@caleb_hunter_guahip  A few things show-up in roughly the same order every time.

Temporal joins with multiple axes are first. "Domains using this NS on March 15, that also had abuse reports between March 1 and 30, and a registrar change in that window." Our graph was built for this. Every edge carries first_seen and last_seen internally, and the agent tooling pushes the time predicate to the front of the traversal so the planner has the smallest possible candidate set. Multi-axis time joins stay fast even on tight date ranges.

Next is unbounded fan-out. "Show me everything connected to this IP within 3 hops" looks fine on a clean test domain. On a shared CDN edge or a popular nameserver it returns millions of edges and dies. We built the engine ourselves so we have cover almost all cases that detect and do some very smart handling of huge fan-outs, but I wouldn't call it solved yet. That path needs more real SOC traffic before I trust it under everything.

The one we hear about most from analysts is fusion with internal data: alerts, EDR telemetry, etc. That stuff isn't and should not be in our graph. A few teams and SOCs are running us now and every one wires it up differently, which is the honest reason there's no single answer. We're not going to ingest your internal logs, so the join has to happen client-side, and that's where workflows get unpredictable.

@easton_grant Exactly the challenge we wanted to tackle when building this and that's our magic sauce. My co-founder Soroush has two PhDs in mathematics and has spent his adult life studying information dissipation in large networks. We ingest in real time, sure, but the harder bit is we also push the update through to every affected node the moment it lands. If an ASN gets hijacked on BGP, every domain served by it is flagged at the same instant. Not on the next crawl. Not at end-of-day. Right then.

Real-time ingestion is the easier half. Knowing which downstream nodes are now suspect because one upstream signal flipped is what's actually hard and we have it!

@oliver_hayes1 I think this only feels transactional if the tip is positioned as payment.

@oliver_hayes1 The framing matters more than the mechanic. Tip prompt only appears after the ticket is solved, never during. It's positioned as a thank-you, not a fee. Fully optional, no follow-ups. If we ever saw it shift CSAT down, we'd reconsider. So far it's gone the other way.

@miles_anthony2 Great question. Variance is real. Some ticket types (refunds, account saves) naturally generate bigger tips than password resets. Some customer segments tip more than others (US > EU, enterprise > SMB). We can't fully eliminate that and we don't pretend to.

What makes it workable in practice is that tips are supplemental, not core comp. Base salary still does the heavy lifting on fairness. We're adding upside, not redistributing the foundation.

Managers see tip distribution per queue and per agent. If certain queues are consistently low-tip, they can rotate assignments or use it as a staffing signal.

And the honest counterfactual: today, agents get zero financial recognition regardless of which queue they're in. Even a low-tip queue is a better earnings situation than the status quo. Variance beats nothing.
Let me know what your thoughts are on this.

@noah_bennett5 The whole point of TipTap is that it auto-scales and stays out of the way. Install once at the org level and it works the same whether you've got 50 agents or 5,000. No per-agent setup, no manual ops burden. Nothing changes in your payroll, HR, or finance systems. Tips flow customer → Stripe → agent directly. The company is never in the money flow.

We're also finalizing SOC2 compliance so we can work with enterprises as well.

If you're at a large CX org and want to take a look, happy to chat.

@naomi_florence1 This is the right concern to raise. There are two scenarios we considered:

If tickets are auto-assigned: there's no cherry-pick surface. Agents work what the routing rules send them.

If the team uses a looser pull-queue setup: cherry-picking is already a pre-existing problem, with or without tips. Today, agents skim the easy tickets and leave the complex ones, the ones that actually need their best work, to sit in the queue. With TipTap, the incentive flips: the harder, more complex tickets are the ones that drive bigger tips, because that's where customers feel real gratitude. So if cherry-picking happens, it shifts toward the tickets nobody wanted before. That's a net positive.

Either way, it's manager-visible behavior. Queue distribution and SLA metrics surface it in the helpdesk. We don't think tips amplify the issue. Please let me know what you think.

@priya_kushwaha1  Thanks so much Priya, great question! Payouts are fully automated. Agents receive their earnings biweekly, deposited directly into their bank accounts via Stripe. Zero work for your HR team.

On the tax side, Stripe provides all the necessary documentation so agents can handle their own filing seamlessly. And worth noting, in many countries like the US, tips are actually tax-free, which is a nice bonus for agents.

@busmark_w_nika Amazing!! glad to hear it's interesting. Let me know what you think and if you have any feedback 🙌

@qifengzheng Detection uses jerk (sample-to-sample acceleration delta), so typing and slow wobble stay low, a real tap is a sharp spike. Not perfect though, cafe-table edge cases happen.

There's a sensitivity slider in Settings if it misfires for you. Anything I really don't want firing by accident I map to 3 or 4 taps. 2-tap is for opening Claude right now.

Hey!@mquiroz- The observation that Apple Silicon Macs have a built-in accelerometer that basically nothing uses is the kind of thing most developers notice and file away. You noticed it and shipped a product in a few days. The result is genuinely useful — 2, 3, and 4-tap rhythms as a natural risk gradient for different types of actions, with no kernel extensions and no network calls, just the hardware you already own doing something it was never asked to do before. For solo founders who live on their MacBook and want instant access to Claude or their most-used shortcuts without breaking flow, this belongs in the stack quietly. Added tapcut to SoftRankings under the pre-seed productivity tools for that reason.

@gretl This is actually a super useful every dev has had that what’s running on 3000? moment 😄

Love the idea of treating ports like a named, versioned system with shared config across CLI + team onboarding. The auto-detect + gr.toml workflow especially feels like it could reduce a lot of onboarding friction in real projects.

@cradeai 

This actually solves one of the most annoying friction points with AI workflows right now.

Feels especially strong for:
-debugging
-spreadsheet work
-workflow automation
-research tasks

Would be interesting to see creator-led “real workflow” demos around those use cases.

@hunter_powabase 

hi, will your brand be expanding into international markets? We can help you with that.