PH热榜 | 2026-06-10

@eugenia_ivanova3 Great project! Good luck with the launch, and thanks for the promo code.

This is a great project solving a real problem and offloading all of the different platforms to a single MCP is much more useful for agentic workflows.

@eugenia_ivanova3 most social tools still feel built for humans clicking buttons, while the real shift is toward workflows, agents, and APIs. making publishing + comments + replies accessible in one automation layer is genuinely useful. good luck with the launch 🚀

@heyalizaid Thank you so much!

@sergebulaev As part of the team, I can say — we really do read everything. Your feedback shapes what we build next. Thanks for being here on day one! ❤️

@sergebulaev  good luck with your launch, the feature set looks impressive! how do you guys handle linkedin messaging - any concerns regarding their ToS?

@maria_anosova Thank you so much! Daily posting takes real discipline — we built Publora exactly for people like you who are serious about showing up consistently. Hope it makes your life a little easier. Good luck with your LinkedIn journey! 🚀

@nikogermish Thank you so much!

@nastassia_k We use official LinkedIn APIs only — no browser bots, no session hacks. That's exactly why we built it this way. Official integrations don't trigger suspicion. 😊

@fberrez1 Exactly. Our goal is to remove that headache completely.

Developers and agents just create the content and send it to Publora. Everything else — platform-specific APIs, token refreshes, rate limits, media upload quirks, and API changes — is handled on our side.

@fberrez1 Thank you for the kind words and great question! Ilya covered it perfectly. 😊

@alexander_anastasin Yes! We have a free Starter plan — 15 posts/month, no credit card required. Just sign up and go 😊

@dmitry_zakharov_ai Thank you!

@dmitry_zakharov_ai 😁😁😁

@muravyov Thanks for the question! Yes, that's possible — give it a try! 😊

@jared_salois Claude Code handles it really cleanly. For other agents it mostly comes down to how well they deal with tool schemas — but in practice it holds up well, and we've built a couple of fallbacks so capability isn't a blocker:

• Ready-made skills — you can just install one instead of self-configuring, e.g. our Threads skill: https://www.skills.sh/publora/skills

• Or go straight to the API — feed any agent our API docs and it'll wire itself up reliably. We put a lot of care into keeping the docs clean and clear exactly for this. https://docs.publora.com/getting-started

And we're continuously improving the docs to make the self-set

@nurlyzhann 

The short version: we push all the constraints down to the schema layer, not the prompt — so the model can't really get this wrong.

A few specifics:

• Every closed value set is a z.enum, so a hallucinated argument dies in Zod validation before it ever reaches the API — the LLM gets an immediate, typed error to correct from instead of a silent bad call.

• Opaque platform IDs are only accepted verbatim from a paired list_connections discovery call — the field description literally says "do NOT invent or guess IDs", so there's nothing for the model to make up.

• Cross-posting to X + Bluesky (or all of them) is a single create_post with a platforms[] array. The per-network fan-out happens server-side, so the model never has to reason about platform differences at all — it just says "post this, to these".

That combo is what keeps it from turning into 18-tools-worth of cognitive load: the agent reasons about intent, the schema + server handle correctness. Happy to nerd out more if you're building in this space!

@timur_sekamov Thank you!
Absolutely — that's exactly the workspace flow

1. You (as the owner) purchase the channels you need.

2. Via the API you create a user and get a unique URL for channel connection.

3. Your customer follows that URL to complete OAuth and connect their channel — generating a unique channel ID.

4. Your agent then uses that channel ID + your owner API key to post.

All connected channels draw from your main channel quota, so your customers don't need their own subscriptions.

Guide: https://docs.publora.com/guides/workspace

@danshipit Thank you! Would love to hear your feedback once you try it! 😊

@furkan_kara1 One integration, posting to multiple platforms — give it a try! 😊

@aymnart Yes, it's included! 😊

@aymnart That’s exactly the point! You don’t need an X developer account or API credits. Just connect your account once via OAuth in Publora, and we handle all the API complexity on our side.

@shirokolobov Ha - you just described why we exist 🙂 This isn't a "vibe-code it in a weekend" project. The hard part is the huge bureaucracy of the digital giants, where you sink tons of time, effort and money into official channel approvals. We've done that legwork across every platform so you don't have to.

@keirodev Both Instagram features are fully supported! 😊

@vinitvr  Good questions!

- Use case: both. Agents call it via MCP; scripts hit the REST API directly. There's also a dashboard if you'd rather not write code.

- Scheduled job: no cron on your side — scheduling is built in (queue posts ahead of time), or fire on-demand from your agent/script.

- OAuth: you don't refresh anything. Connect an account once and we rotate/refresh the tokens in the background — you only re-auth if you revoke access.

- Analytics: yes — per-post engagement (likes/comments/views) LinkedIn only.

@vinitvr Serge covered it perfectly! And if you ever get stuck setting things up — we're here. 😊

@michael_vavilov At the moment, we don’t expose proxy/IP geo selection to users. Our infrastructure currently runs from the US (Virginia).

That said, it’s a great point. We know that for TikTok, not only the country but often even the publication location can affect distribution. We’ve also seen creators deliberately optimize for local presence.

We’re definitely thinking about adding support for publishing through residential proxies and allowing users to choose the target country (or even city) in the future.

@dani_mashael Thank you for the feedback! Hope you enjoy Publora 😊

@dani_mashael Thanks, our goal is #1! 😀

@ilya_kolesnikov1 Good point! You connect your account once via OAuth — we handle token storage and refresh automatically on our side. You only re-auth if you revoke access yourself. 😊

Thanks @caneraras! It is!

@scream4ik Thank you! So glad to have you here! 😊

@igor_martinyuk Thanks! 🙏 Honest answer — no, Publora doesn't learn your style. It's a purely technical tool: it publishes and schedules the content you pass to it across your social networks. The taste and style stay on your model's side — Publora is just the layer that ships what your agent creates.

@igor_martinyuk BTW if the "writing in my style" part is what you're after we have another product, Co.Actor, that does exactly that. It's built on top of Publora under the hood, but Co.Actor is the layer that actually learns your voice and drafts posts in your style, then Publora ships them. Kind of the perfect combo for your "AI CMO" idea 😄

@thinkharshh Ah, if you mean scraping/indexing social networks — then no, that's not something we do 🙂

Publora is focused purely on the content side: creating, publishing and handling your posts (and replies/comments) through the official APIs. We're not in the data-harvesting business. If your use case is more on the "post & engage from my agent" side, that's exactly our wheelhouse — happy to dig in!

@soni_karan 🙌

1) Metrics & insights — yes. You get analytics on the posts you publish through Publora: LinkedIn is the deepest right now (impressions, unique reach, reactions, comments, reshares — per-post or aggregated), and the other platforms are a bit lighter for now, with more on the way. Since it's all through the MCP, you can just ask your agent "how did my last post do?" → https://docs.publora.com/guides/linkedin-analytics

2) vs Buffer — honestly, it comes down to who we are: we're a small, fast, AI-first team (~8 people), so we move quickly, ship constantly, and build everything agent-native from the ground up rather than bolting AI on later. That speed is also why we're one of the most affordable / best-value options out there. So you get an AI-first product that evolves fast, at a friendlier price 🚀

@gizem_ozturk Every network has its own quirks and requirements — but we figured them all out, and that's what matters! 😊

@boris_lifatov Exactly our thinking! Thank you! 😊

@nevo_david Good landing pages share a lot of DNA 🙂 The product underneath doesn't — Publora is API + MCP-first for agents. Thanks for checking it out, Nevo.

@tony_dinh2 congrats on the launch, Tony! Looks like TypingMind has come a long way since I last checked it out. Will give it another go.

@tony_dinh2 TypingMind is by far the most advanced Chat UI, nothing comes close!

@tony_dinh2 Super cool - thanks for sharing!

@jon_jonssonExactly this. At some point you're paying for four different AI subscriptions and using maybe 20% of each. Pay per use with your own keys is the only model that actually scales with how you work.

@jon_jonsson This is the real unlock. Subscription stacking quietly becomes a line item nobody owns. The open question is whether teams actually track usage per model, or just feel the bill at month end.

@ann_ng1 thank you!! 🥰

@absurdfounder thank you Vaibhav!

Just want to bring you papercut fix and feature idea for 2026.

1.provider credit balance display under api key management

2.Image paste support on Individual input during multi chat conversation (attachment support exit but paste support is not their)

3.Fix : Install MCP from JSON (TM cloud connector)

4.Fix : edit option for installed MCP

This request present in our feedback forum

Thanks again

@natalia_iankovych hi Natalia! In TypingMind you bring your API keys so it is very flexible! 😄 You can add multiple models to the app and switch between them among your chats.

@nickzaleski a fork has full parent context. A parallel chat only get context from it’s own branch. There is a “Finalize” feature that will use a chosen model to read all branches and merge them into a final answer 😄

@rampatra thank you so much, Ram!

@fabian_maume Open Router gives you access to the models, TypingMind gives you a UI to use it 😄

You can use Open Router models on TypingMind too, it is supported natively in the app!

@rodrigo_rocco thank you so much Rodrigo!

@martin_donadieu 

The most obvious benefit is to have the flexibility to switch between models and to use multiple models in parallel! :D

There is no specialized coding mode, but users can still ask the AI to write code, and TypingMind runs it in a code sandbox (provided natively by the model provider!).

Thank you for visiting our launch, Martin! ❤️

@nevo_david thank you, Nevo! Big fan of what you do with Postiz!

@antifreeze congrats on the launch! Such a needed product right now.

@antifreeze The SSH keys is exactly why this matters, gents move fast and

the blast radius is invisible until it isn't. Congrats on launch. Trying it today

@antifreeze S/O for this launch! Keep up the great work 👏👏

@chrismessina Twenty years and I pay it off by showing you the bacteria on your toothbrush. 😅 Don't be embarrassed about the red -- we build security software and our own reports lit up too. It would be embarrassing if it weren't happening to everyone.

"Non-engineer engineers" are exactly why we made seeing this free: everybody's shipping like maniacs now, and everyone deserves to know what their agents are doing. (And "behind every successful coding session is a backplane" is going on a shirt.)

@curiouskitty This one's fun to answer, because the trick is that there's no trick: the harnesses already write everything down. Claude Code and Codex keep a transcript of every session. Spotlight's CLI watches those transcripts on your behalf, redacts sensitive info from new activity locally, and then sends the redacted version up for pattern analysis and report generation.

So attribution falls out of the session itself: it knows its user, its project, and its tools; org rollups are aggregation, not integration. Spend is computed from the same token counts the provider meters, which is why it tracks invoices closely. It's an estimate by construction, but a well-grounded one, and no OAuth or provider-side hooks are involved.

On tradeoffs: we deliberately chose to start with reading over intercepting. Real-time gating means sitting in the request path, and a proxy adds latency to every call and breaks when harnesses update. Our way, you're minutes behind live but never blocked mid-flight: a genuine trade, and one that buys zero added latency and zero workflow friction.

We'd love for you to take a look. The install is one line, and your first report lands in a few minutes. Let us know what you find! :)

You're spot on @fberrez1 : the session replay is the easy part; the interpretation layer is the product.

Short version: the raw session info is the input, not the report. The bookkeeping (counts, files touched, domains, cost) is computed mechanically, and the analysis on top is held to one rule: every finding has to point at the specific moment in the session it came from. If it can't cite the event, it doesn't ship.

On noise: that's most of what the engineering portion of the report is for. It surfaces retry storms, redundant tool loops, repeated lookups that should have been cached, and it distinguishes failing retries from deliberate re-verification. Those land as "Faster Next Time" items with payoff grounded in the session, like "~60 calls collapsed into one." CI, test, and lint outcomes get flagged when the transcript shows them. And when something isn't observable, the report says so in a blind-spots section instead of guessing. We'd rather show you an empty field than an invented one.

Security is a separate findings stream, severity-ordered, with categories like credential, shell, file, network, production, and subagent. Concretely: a live-looking key written into a tracked .env (with a paste-into-Claude prompt to rotate it), a destructive command against prod with no dry run, a call to a domain you've never used, a subagent reaching outside the project. One detail worth knowing: secrets are redacted on your machine before anything uploads and a second pass is run on the server before we write, so the report can flag the secret class without ever holding the value.

You're right that those are two different use cases. The report carries both on purpose: the security stream and the engineering narrative come from one pass over the same session and give you the full picture. That's the bet we're making.

Run it on your messiest session and tell us what you find, here on our community Slack. :)

@fberrez1 Neil's got the depth covered, so just one top line note: the part that surprised us most is how high-signal the reports turned out to be. Even on our own sessions -- and we live in this thing -- they cut through all the back-and-forth and retries and showed us things that really mattered, starting with credentials we'd leaked and never noticed. We were expecting most reports to have very little useful signal, and that you'd have to wait to see things in aggregate stuff for the really meaningful items to bubble up. Nope! The important stuff is right there from the get go. Would love to hear what you experience and if it matches ours. Thank you!

@anand_thakkar1, that war story could be one of ours: an agent improvising around a missing env var by quietly inlining the key is the kind of move nobody catches in review. One small correction, it's not OS-level instrumentation. Spotlight reads only the session transcripts the harnesses themselves write, nothing else on your machine. But your key insight holds: it's what the tools actually did, not what the agent says it did.

On fixtures vs leakage, we treat those as two different jobs. Redaction is deliberately paranoid: anything secret-shaped gets masked on your machine before upload, fixtures included, because that step shouldn't be guessing intent. The judgment lives in the analysis: where the credential landed, whether it looks live, and what the session was doing at the time, with severity reflecting that context. A dummy key in a test fixture and a live-looking key written into a tracked file are very different findings. Every finding carries its evidence, so on close calls you're the judge, with the receipts in front of you.

We'd rather flag a fixture at low severity than miss a live key. That asymmetry is on purpose.

Thanks for the support! Let's spread the word on LinkedIn, repost this

@daniel_oyetunde Great questions! Windows support is a priority for us now. It's being actively worked on, and we expect to have it out in the very near-term.

@kcpike framing this! S/O to ?makers for building this

@kcpike "Nothing closes the loop back to the human" is the cleanest statement of the problem I've seen yet. :) That's exactly it: review used to be where engineers got better, and agents quietly took that away while speeding everything else up.

@kcpike 1000% Everything is changing under our feet at a crazy rate, so the faster we learn and improve the better we get and the faster we all go! Recursively compounding returns!

Amazing! FYI you can see a sample report here: https://www.backplanes.com/features/session-reports

Curious to have your thoughts about it

@david_marko, the end-of-session log is the certified power-user move, and you've described our origin from the other side: we were hand-rolling the same ritual and got tired of it. :)

If you're open to it, I'd love to see what would happen if you ran your hand-rolled log running for a week alongside Spotlight, and then let us know what your log caught that we missed, or vice versa. That kind of feedback is gold for us, and it sounds like you're exactly the right type of person to give it.

@conduit_design Ha, right?! The wild part: that's the agents' own tab, we just hand you the receipt. It's crazy how quickly token usage accelerates when you're running multiple subagents on an intensive job, and Fable pricing is going to make this even more fun for all of us soon. 😅

On "recursively better," the idea is that it's a loop with you in it. The model never changes; your setup does. Each report turns what happened into concrete and actionable advice: a fix to apply, a CLAUDE.md line to include, a Skill to draft. Your agent loads that richer setup next session and starts more informed than the last one.

Thanks@ianhxu. "Too real" is exactly how it felt on the inside too. :) The answer is both, at different layers. Each session gets its own report, with its own findings and evidence, so you know exactly which session did what, and exactly what to do about it.

Running several harnesses at once just means several reports, and your Claude Code and Codex reports live side by side in Spotlight. But your highest-leverage opportunities are often in the trends and patterns across sessions, and so Spotlight gives you a report of what's important across all of them. And finally, connect your whole team and the view widens even further: patterns and trends across every engineer in an org.

One individual or a team, one harness or seven, Spotlight gives you both the detail and the big picture.

"The "what your agents actually did" angle is great, that read-47-files scare is too real."

@ianhxu Fun fact: This "read-47-files scare" actually is a true story! See background story by @antifreeze here

@luca_capone, "the secret you committed three sessions ago" is almost word-for-word how Spotlight actually started for us: I asked Claude to fix one file, and an API key ended up in a tracked .env that we only caught by accident. So yes, security is very much called out specifically, and it leads the report rather than riding along. Findings arrive severity-ordered in their own stream: secrets landing in files git tracks, prod-touching commands that skipped a dry run, an agent quietly reaching a service you've never used, each with the evidence behind it and a concrete fix.

Code quality and patterns that can help make you more effective with your harness are in there too, so the report always gives you value, even when there are no security-related findings. And since those transcripts are already sitting on your machine, the first report can start with the sessions you've already run. Your "three sessions ago" is still catchable. :)

@luca_capone exactly! Would love for you to give Spotlight a spin and tell us what lands for you, and what we could improve!

@artstavenka1 Thanks, Art! Definitely deliberate, but one (good!) clarification: it's not limited to being post-session. Spotlight reports actually build while you work, just minutes behind your agent, so you're not waiting for a session to end to find out what happened.

What we deliberately avoided is the inline path. A guardrail that interrupts mid-write has to sit between you and the model, with the latency and potential breakage that path implies, and we very much want to help keep you in flow. :)

@louislecat lfg! here you go: backplanes.com

looking forward to your thoughts

@louislecat amazing! excited to hear what you think and what you'd like us to build next!

@van_ho : Thanks, Van! "Come back after a break and have no clue what just happened" is exactly the moment we built this for. If you give it a run, I'd love to hear what your first report catches. :)

I’ve had way too many Claude sessions where I come back after a break and have no clue what just happened in the project.

Spot on! Go to backplanes.com and generate your first report. Curious to have your thoughts about the result 👀

@armand thank you! "Obvious once you see it" is the best kind of compliment. Excited to hear what Spotlight illuminates for you-- and what you wish it showed. That's the stuff we want to build next!

"I’m using Claude Code/Codex every day right now, and the trade off is you don’t really know what got touched, what got skipped, or what weird security debt just got created."

@armand exactly! @Spotlight by Backplanes brings clarity to your sessions. it improves your code and makes you a better engineer.

S/O to @antifreeze @gogogadgetneil @natwar86 and team

@richiebonilla agreed! we think there's a real opportunity here for Spotlight to help build better skills for engineers on the fly based on the content of these reports. We're excited to make engineers who use coding harnesses faster, safer and more cost effective!

@richiebonilla, "a bespoke string of skills run by hand" is exactly the ritual we kept finding, and living ourselves. Replacing yours is about the highest compliment a tool like this can get, so thank you.

Thanks, @farrukh_butt1 "quietly" is the certainly operative word: the touching, breaking, and exposing all happen mid-flow, in the part nobody's watching, while the shipped code gets all the attention. :)

@yrechtman "Rad and kind of terrifying" is the almost universal reaction so far. ;-)

It's the bar tab after a great night: the fun was real, and now there's an itemized record of exactly how.

The difference is this tab works for you. It shows what drove the cost, and the biggest line items usually turn out to be the easiest fixes: the same dead end paid for ten times, an agent left grinding away on the wrong thing.

Curious to hear if your next session's number behaves.

Thanks@malithmcrdev Malith! "Feedback loops" is exactly the right frame, and it's the whole bet: agents sped everything up except the part where the human gets better. Watching the first reports land today and seeing what people fix first has been the best part of launch day so far. :)

@odeth_negapatan1 honest answer: the most eye-opening pattern wasn't actually a risk. When we stitched our own sessions together, the scary stuff was there (an API key in a tracked file started this product), but the bigger surprise was how many good moves never traveled. One of us had solved a problem cleanly, in one session, on one machine, and the rest of us kept burning time on the same thing for weeks. Nobody was hiding anything; there was just no surface for the unlock to move across. That discovery was a genuine unlock for us, and is a core part of why Spotlight focuses so heavily on highlighting effective patterns worth sharing..

What I've found fascinating is almost every first report we've watched land in person contains one thing the person would have sworn didn't happen. That moment is kind of the product. Curious what yours catches, if you give it a run. :)

@retain_dev Gaurav, "filesystem boundary decisions" tells me we've fought some of the same battles. :)

Short answer: the anchor is identity, not inference. The CLI is signed in as you, so every session carries the same account identity no matter which repo or machine it ran on, with the repo, harness, and model riding along as context. Patterns aggregate across the account, so the same retry habit surfaces whether it happened in your API repo on a desktop or a scratch project on a laptop. That's literally how Spotlight started for us: stitching our own sessions together across machines, and being floored by how much we'd missed and how few of our good patterns traveled.

One thing we deliberately don't do is behavioral fingerprinting to guess "same agent" across accounts: identity stays explicit and predictable. And since you mentioned boundaries: every report flags file access outside the project, per session, so drift is visible long before it needs to be policy.

@keirodev Kévin, speaking as the Neil in the story and a career-long security and privacy nerd: agreed, the scare is the pitch. :)

Your "drift is invisible at scale" line is exactly the problem statement, too: one engineer can review one session, but nobody reviews session forty-seven across six teammates.

On "really an audit trail": half yes, and the half matters. It has the rigor of one, every finding cites the exact moment in the session it came from. But it's pointed forward, not backward. An audit trail waits for the auditor; these reports feed your next session: what to fix, what to codify, what's worth sharing. Same evidence discipline, opposite direction.

And good news on file-level visibility: that's not roadmap, it's already in there. The 47 is the headline number, and the report under it lists the files, read vs written, inside or outside the project, with risky paths tagged, env files and credential-shaped paths included. For content-level changes we go finding-first: when a write matters, the finding cites it with evidence, rather than re-rendering diffs git already shows you better.

@kevinzrzgg, you seem to have written our report spec almost exactly. :)

Changed files: all files read and written are in there. Tests and checks: flagged with their outcomes when the session shows them. Failed attempts: called out, including the distinction between deliberate re-verification and flailing retries. "What a human should look at first": that's the top of the report, a one-line verdict with the main outcome, then findings ordered by severity and guidance on what to do for each.

The one we can only claim half credit on is risky assumptions: concrete risky choices surface as findings, and a blind-spots section names what the report couldn't verify, but a dedicated assumptions section is a great idea.

And we're with you on boring: the standing rule inside the report is no invented findings and no padded advice, an empty section beats a manufactured one.

@robskigb thanks for the support! IMHO you'll enjoy using @Spotlight by Backplanes as it doesn't just improve your code, it also makes you a better engineer.

Give it a spin and generate your first report on backplanes.com -- enjoy!

@robskigb Thanks! These agents are getting more and more powerful by the day and, as a wise man once said, "With Great Power Comes Great Responsibility." We're excited to introduce Spotlight as a way to use these agents to their fullest, but also still be safe as you do it! Thanks again!

@robskigb Thanks for your support here! Looking forward to Backplanes being your sheath, or armor, or whatever the appropriate katana-analogy-extension would be. :)

@walkeriwilliams Walker! You believed before there was anything to test, and you've kept the feedback sharp and neverending ever since -- in the best way. Spotlight is better for every round of it. Thank you. Keep it coming!

@benparr thank you! Super curious what you learn from Spotlight and how we can make it better for you!

@ozandag Thanks for comment Ozan. This is a great question!

We don't do code quality analysis yet, but what we do is deep analysis of specific sessions and also all your sessions in aggregate. The report covers a few major areas:

1. Things you really should watch out for:

   • Was a credential sent out inappropriately?

   • Was PII sent to a website that wasn't known?

   • Did the agent access production without knowledge consent?

   • etc

2. The second thing we do is we look for patterns that should be replicated for you and across your teams. There are certain things you do really well replicate in those and other places. Like TDD use.
   

3. The third thing we do is we also find ways to speed you up. If there are certain things that you're doing that could be optimized, both from a token perspective or a speed perspective, we highlight those as well. For instance, I had an instance that auth'ed to the same service 4 times, killing 25 minutes.
   

4. The fourth major thing we do is we actually just talk about how you use your tokens: where do they go, what MCPS/etc are being used.

In addition to that, we also give stats on what code bases you work in the most, CI/CD pass/fails, abandoned work, github activity. And we do this across all sessions for both Codex and Claude Code together.

If you work with somebody else, we team level insights as well.

Code quality is not a bad idea though! We'll add it to our backlog. Thanks so much for the great question Ozan!

@gogogadgetneil  @srinath_murthy we couldn't agree more! We hear this ad nauseam from everyone we talk to, and are so glad to help people meet this need. Would love to learn from you what connects most about the reports, and what work you'd love to see tackled next!

@sergei_nazarov92 Great work Sergei!

I like the interface. It's clean and modern. I have a couple of ideas you can implement.

@sergei_nazarov92 Hey Sergei! Congrats on finally getting Screen Charm onto PH. As a Mac user who values clean, efficient workflows, I know how hard it is to find recording tools that don't feel 'clunky.' Love that you built this to solve your own need for high-quality demos.

@sergei_nazarov92 Congrats Sergei! A year of building this alongside a full time job is no joke, really glad you kept going and finally got it to launch.

If you ever want to get your demos in front of a live audience, check out @SimuLive, it helps keep people engaged through webinars and meetings. Best of luck today! 🙌

@farid_sukurov Thank you Farid! one of my differentiators from the competitors. I want to move more into the area of cool camera recording layouts instead of focusing on more polished screen recording

@a_6 thank you Anna!

@farrukh_butt1 Thank you for the kind words Farrukh! This is our main target to make creation of professional screen recording easier

@rodrigo_rocco I will Rodrigo!

@thinkharshh Hey! Nope, it's not screen studio. Thank you!

@mayarossi Hey! It works only on Mac. No windows version for now

@david_marko Thanks a lot! Yes, currently there is no need for subscription. But I have plans of switching the pricing model to subscription based. Don't know yet when will it happen tbh

@dhiraj_patel5 Hey! Yes, it works exactly like this. We track the cursor movement and event. Then transform them into smooth zoom effects

@german_merlo1 Hey! Thank you! Currently the pricing is a lifetime deal for $79, but there is a promo code for 30% off

@frboulais you're welcome! it's only the pleasure to build such tool, i don't feel tired at all. the only only problem is million of edge cases

@maryna_klokova Hey Marina! It's music to my ears to hear that :) More features are coming soon!

@bojan_sala yeah thank you! I started building it because I could sell lifetime deals and easily validate that I can get users. The price started from $29 when I just launched it. Since then I was steadily increasing it. Probably next year I will switch to subscriptions

@song_kirby That’s exactly why we built iArt to be a fully multi-modal platform. You can feed your raw outlines to the AI in whatever format you have on hand:

- Text & Data: Paste directly into the chat or upload files (.txt, .md, .pdf up to 10MB; .json, .csv up to 2MB).

- Visual Sketches: Upload up to 6 images (.png, .jpg, .jpeg up to 10MB each) if you have raw whiteboard mockups or architecture diagrams.

- Voice Outlines: Upload up to 10 audio clips (.mp3, .wav, .m4a up to 30MB each) if you prefer just talking through the logic.

iArt will parse the underlying logic and automatically map it into a clean, cinematic animated breakdown. Give it a spin on your toughest project—we'd love to see what you create!

@thamibenjelloun yes， you can try the brand function . or strict prompt .

@alstonzhuang Thanks! It’s a great question, and to give you an honest technical answer: our Agent is designed to follow complex, detailed instructions, and in the vast majority of cases, it sticks strictly to the details you provide. In fact, the richer and more specific your prompt is, the better the final cinematic output will be.

However, because multi-scene long prompts can sometimes cause the AI to occasionally skip an edge-case detail, here is an insider pro-tip to make the Agent execute your scene-by-scene script flawlessly:

Structure your input with clear sections, such as:

- Duration (e.g., scene lengths)

- Global Rules (e.g., visual style, tone)

- Visuals + Script (the exact scene-by-scene action)

- Voiceover / Dialogue Lines (you can even specify the pacing here—just remember to select your preferred voice from our list first!)

When formatted this way, the Agent handles the complexity beautifully without trying to simplify your creative vision.
💡 Here’s the shortcut: If you don't want to format all of this manually, you'll love the "Enhance prompt" magic wand button right next to the enter key in our chat bar.

We’d love for you to drop your most detailed script in and push our Agent to its limits!

@yulesenmiao Thanks for the love! 🙌 You’re going to love this: our tool inherently keeps your project status and history intact. No matter when you re-open a project, you'll always see the exact full chat history right where you left off. You can freely experiment without losing your train of thought.

@charlenechen_123 Hi Charlene! This is such a key pain point — generic “AI look” is exactly what we built iArt to avoid 🙌

From your side, the workflow is designed to keep brand consistency front and center. Before you prompt, just head to the Brand panel and select the specific brand profile you’ve already set up — this tells the agent to prioritize your logo, colors, and fonts from the start. When writing your prompt, feel free to ramble or be as messy as you want; the “Enhance Prompt” magic wand right next to the chat bar will refine your request, giving the agent the full context it needs to stay on-brand.

On the technical side, our model is trained to prioritize your selected brand guidelines over generic stylization. It uses layout-preserving motion logic to keep graphics clean and flat, while dynamically weaving in your brand assets instead of applying one-size-fits-all filters. This way, the output feels uniquely yours, not just another AI-generated clip.

@frey_loong Hey Frey! Great question 🙌

Our agent is built to balance creative direction with layout integrity. It interprets abstract prompts by referencing your selected brand profile and existing video style, so it can add kinetic energy without straying from your corporate look or messing up the layout.

If you’re struggling to put your ideas into words, feel free to put the AI to work overtime 😆 and ask it to create several design concepts for you to choose from. This is a fun way to align on design directions and lock the elements you don’t want changed before moving on.
The latest prompt you sent can be interrupted or retracted at any time, so you can tweak your ideas freely. If the result still isn’t quite right, you can always refine it further in chat — the agent will adjust transitions while keeping the original layout intact.

@lennoxbeflying Hey Ziang! Thanks for the thoughtful question 🙌
For voiceover: Turn on the voice feature in setup, pick a voice/accent, and and specify tone or pace in your prompt. The agent generates voiceover synced to your visuals. You can also adjust the script and even the timing of specific lines later in the process (a hard refresh Ctrl + Shift + R may be needed to see updates).
For BGM: You can pick tracks from our built-in music library and adjust the volume after video generation. We’re working on new audio features, so stay tuned 🎬.

@theshumba Hi Melusi! Thanks for loving our positioning.

For short videos, you can set overall pacing, transitions and style directly via text. No need to guide frame by frame — just tell the AI your requirements and it will adjust accordingly.

If you have super specific ideas and want the AI to follow your instructions strictly without creative tweaks, simply state it clearly, and it will fully comply.

Occasionally preview loading issues are caused by cache. Just press Ctrl + Shift + R to hard refresh, and you’ll see the latest updates.

You can also ask the AI about its creation logic anytime in the chat panel. It will explain details from both artistic and technical perspectives. Enjoy using it!

@oliverbuilds Hey Oliver! This comment made our day — thank you so much for noticing what we care about most 🙌
You’re exactly right: most AI video tools force you to rework finished videos. Unlike other tools, you can revise the content repeatedly until you’re fully satisfied before rendering and exporting. Our duplicate project feature also lets users create freely with greater peace of mind. Plus, you won’t need to re-render files over and over again.

This is just the beginning for us, and we’ll keep improving. Thanks again for the support! 🎉

@diana_nadim2 Hi Diana! Thanks a lot for your great questions.

We’d like to view professional motion graphics in a balanced way. Our product is mainly built for general users, especially those who struggle with ideation or lack professional editing skills. Our AI agent helps them easily create decent visual results.

As our founder mentioned, iArt is fully prompt-driven, replacing the traditional manual canvas with simple chat. We also offer "Choose Element" and "Capture Frame" tools for precise and quick editing. That said, the experience can vary a bit depending on your network speed and patience. 🤣
And the system works well for both formal production and rapid prototyping. We actually create all our social media content across platforms with iArt. Beyond regular "+ New Project" feature, our "Batch Render" is quite practical: you can pick a preferred version from the batch to remix or polish further, or ask the AI to generate entirely new takes.

We really hope you can try out our product and share your feedback with us!

@millwiller Hey Will! Thanks so much for the great question 🙌

Both are great tools, maybe just built for different workflows.
CapCut’s AI features are add-ons to its powerful, timeline-based editor — they give you tons of manual control, but still require hands-on tweaks to pacing, graphics, and transitions after the initial generate.

iArt, by contrast, is fully AI-native. We built it for creators who want results fast, without the learning curve. For non-professional users especially, the AI’s ability to interpret your request and deliver a polished video quickly is often more practical than endless manual controls. You just chat with the agent to refine anything you need — no timeline, no keyframes, no tedious edits required.

@daniel_oyetunde  Thx ! yeah , attach with the img , it will transfer for you . may not be losslessly, but normally > 80 % reconstruction.

@luigi_receiptorai  yeah buddy! basically , it's generating inforgraphic/ Vector with motion ~

@galyna_arikh  hey , thanks for asking , It generates Motion Graphics , basically vector graphic with motions.

@natalia_iankovych  thanks! hope it helps, you can also give it a try with voice over turn on ~

@margarita_tsygankova It's actually quite simple , just prompt.
Also you can output clips , and reuse them in cap cut
Since the Agent nowdays may not be smart enough in video editing, it may get stuck on a small detail.

@vinitvr  not sure if it can reconstruct it perfectly , but screenshot is supported , give it a try : )

@tina_chhabra  yeah , definitely !

With vibe coding , we now ship product 20 times faster , be still stucked in AE when we want to marketing it .

@keirodev Thanks !

we got a brand button for it .

it's kind like skill.md , but for brand , we're still on the way to enrich it. : )

@dioiv  Thanks , we support both uploading and reusing ~ may I ask what would you create for batch render : )

@anand_thakkar1  we are still on the way, quite hard actually, we tried a looot of ways to upgrade the agent. I think the key is about letting the agent focus on the brand or style guidiance. component libs also helps !

@fberrez1 Thank you so much for the support! Yeah we made the video with our agent and here's another launch video in different visual style. 😋

@syozz Great question, Ethan! For non-Latin scripts, Chinese is supported with stable rendering.

Regarding Arabic, we want to be precise: we recently audited past projects from our Arabic user base with no major text rendering issues found.

However, we are always pushing for deeper optimization when it comes to script-specific typography (like advanced line-breaking or perfect font pairing). So while the core formatting works reliably, quality might vary on some unique layout edge cases.

We highly recommend running a quick test with your own text! If you spot any layout bugs, please let us know—user examples are exactly what help us prioritize and polish these localized features.

@alina_tyslenok_ Thank you so much for the kind words! That's exactly why we built this—traditional video editing has such a steep learning curve. We wanted to make creating professional content as simple and natural as having a conversation. Hope you enjoy trying it out!

@eexlkuang_se Spot on! As you can see in our Brand Kit setup, we've built specific slots for Logo, Colors, Fonts, and Images.

You can absolutely input your custom hex codes and upload your brand's TTF fonts. The agent takes these inputs as hard constraints to restrict layouts and styling, so you get brand-compliant generations every single time. Give it a spin and let us know how it works for you!

@summer_tao Great question! iArt is primarily powered by dynamic generation, but we also support templates. Even better, you can explore and remix any community-created layouts from our showcase!

We are currently fast-tracking the development of a brand-new batch of premium templates to give you even more design flexibility. Stay tuned for the upcoming updates!

@jiaqichen Appreciate the love! Cheers to a future with fewer keyframe headaches. We'd love to hear your thoughts once you try it out—let us know how it compares to your usual workflow!

@ericries Congrats on the launch, Eric! How early in a startup’s lifecycle should founders start implementing this governance design to proactively resist financial gravity

@david_knox1 thank you! I look forward to launching the UK edition formally in early September.

@thamibenjelloun Thank you.

@stphnmrrs it's confusing because both of these things are true:

1. this problem has been going on a long time (I give examples in the book going back >200 years)

2. this problem has gotten a lot worse recently

I think one of the key mistakes was the shift to "shareholder primacy" a few decades ago. This shift is, fortunately, young enough that it's still quite easily reversible, if we will it.

The recent trend to allow companies to incorporate as "public benefit corps" in states like Delaware is an example of this rebalancing that you describe. We have to deal with the fact that many governments have been captured by the interests of the wealthy and especially of investors, which makes them less likely to want to do this rebalancing. But this, too, can be resisted, if we will it.

@markusb79 Let's get them copies ASAP!

@regina_jaslow Pretty much, although there are the occasional exceptions.

@sandra_sydnor I would say it's far more important, as you're much more at risk of the corruption that this book is designed to prevent.

@rrhoover Well, not really, because now I know how difficult it is, and so it's actually a bit daunting. In all seriousness, it's a bit of a mixed bag. Because I've been successful with my previous books, I have had the immense privilege of being able to afford a team to help me. I had researchers and editors and all kinds of allies.

On the other hand, I felt immense pressure to try and deliver something worthy of being seen by people as the successor to The Lean Startup. As many other artists have attested over the many generations, that pressure can be, in some ways, debilitating.

@natalia_iankovych I hope it's ok!

@natalia_iankovych , the Startup is there, it is just Lean - pun intended
I'd recommend you pack the book with its ante-predecessor, starting with the latter ;-)

Many Indie authors post on PH, some time the full copy, other times extracts. If there is someone legitimate to post his newest book here, it's Eric

@alexryan thank you!

@dani_mashael one of the most important ideas in the book is that "it's always too early until it's too late." I personally think companies, whenever possible, should reserve the option to install mission-protective provisions into the legal documents from as early an age as possible. Notice that investors do this routinely and without needing to explain or defend themselves. Why shouldn't the mission get the same consideration?

@curiouskitty privacy and trust are critical at Napkin Math! First off, you control everything in the app and what is shared or kept private.

All notes about your meal, how you are feeling throughout the day are kept private, chats with our AI, and plans you build are kept private. Food is personal and Napkin Math is safe place whether you are managing a health condition, an eating disorder, or just your plan to cook more.

That said - people are sharing a lot! People share both their beautiful home cooked dinners and their scavenged leftover lunches.

People can share with their friends or globally in our social feed. We designed a simple upvote system to keep it positive when you share globally. And with friends we open up comments that we call "bites" to both praise their food and playfully remind them when they skip lunch (that's me 🙃).

We're looking to add more you can share in the future such as recipes or meals plans. Would love to hear anything else you might like to share about your food!

@alheri_murya Agreed! Food shouldn't make you feel guilty and we want to change that for everyone.

@jynnie Congrats on the launch, Jynnie. The fact that this came out of a real experience with your own gut health makes the product feel different from everything else in this space.

One thing I noticed: the story you just shared in this comment is the sharpest, most emotionally specific copy for Napkin Math that exists anywhere. The ER visit, the list of food triggers, opening the App Store and finding nothing but calorie counters. Someone with a mystery gut issue or unexplained energy crashes would read that and feel completely seen.

The homepage doesn't have any of it. Curious whether that was a deliberate choice to keep things minimal?

(direct response copywriter, which means I can't read a homepage without doing this)

@farrukh_butt1 Thanks! Would love for you to give it a try and let us know what you think!

@thamibenjelloun we find it to be incredibly accurate just from a photo!

But what we see people doing even more is adding notes to their meals like "I only ate half the bagel" or "I scraped off the icing on the cupcake" and we can quickly reanalyze the meal. Including those little notes are really how we get more accurate with the way people eat.

@sasha_mcclendon You can always include written notes with your photo or even after you have submitted your photo. This updates the nutritional analysis and helps people accurate capture what they really ate such as seasoning or sauces.

As for workouts, yes, we integrate with Apple Health and are working on a beta to show more things you track like heart rate, sleep, hormonal heath, and more. For example, I track a plan to have Napkin Math remind me to eat more protein and hydrate after I go for runs!

@marianna_tymchuk Thanks Marianna! That's our goal! We believe food is joyful and social and building a community to do that together.

@luca_capone You get it! Building a habit is hard and a lot of apps just lean on your willpower to keep doing the work. In short we make it fun, social, and have a coach to remind and adapt your plan.

Napkin Math is making food fun with a visual journal that you want to keep building on. Many people join just for the sharable calendar view.

Then the hidden unlock has been friends. We are seeing an increase of partners, families, and friends join the app where they remind their spouse to eat lunch (that's me 🙃) and beg their friends to show them where they got that matcha from.

For people tracking a specific goal we have exactly what you called out - a coach that reaches out to check-in. This helps remind people to stick to their plans but also adapts when you get sick, go on vacation, or only have two choices at the office lunch.

@gabriella_anjani love to hear how our approach resonates with you!

We're building Napkin Math to be a life long companion so it's fun all the time and give you the tools to help manage your conditions day-to-day.

If you try out the app, I would love to hear your feedback!

@arathy_kushalappa1 welcome to the Napkin Math community!

My wife is Indian and it captures her family's recipes quite accurately! And every time you log a meal, Napkin Math remembers it so you can reference it again. So if there are some extra ingredients maybe a photo can't capture - write those down when submitting (or the whole ingredient list if you want) and every time you log it again, it will remember!

You can also add a note after logging a meal to tweak anything like swapping between chicken and tofu or saying you added yogurt this time.

@igorgurovich Great question Igor! We do both!

Napkin Math does a great job of identifying even complicated dished from a single photo and you can add text to explain even more of what is in the dish. For example, a photo of a plate of piroshki (top of mind because I had these recently) could also have text explaining they were filled with mushrooms.

@denitsapenchevavaltchanova Ahh! Love that you like our design!

And we absolutely want to work with healthcare providers - we have spoken to some nutritionists already but are seeking more to learn and work with. If you are interested I would love to connect find some time to chat about how we can help providers and patients with nutrition!

@connortbot thanks!

@malani_willa curious how comfortable it is for longer sessions. Does the display feel natural after wearing it for a few hours?

@malani_willa I like that this is built around a real workflow problem. if developers can interact with agents naturally, this could open up new ways of working

@saba_k Really cool to see physical AI becoming more accessible. The combination of dual arms, long reach, and easier data collection makes this feel like something teams can actually deploy for useful tasks, not just demos. Excited to see what people build with Axol. 🤖🚀

@thamibenjelloun The first few folks we've been talking to happen to focused on kitchens.

@conoremclaughlin Thank you!

re: 3 cameras: You can probably get away with 2 wrist cameras (this is what generalist does) but 1 head camera + 2 wrist cameras is how folks often do training. I like to think of the wrist cameras as a clever way to replace touch since robots don't have it yet. You can also try 1 head camera + torque feedback and see how it goes, our API supports it.

re: We only support linux right now so you can run it from any linux box. That said, the ZED cameras require Nvidia GPUs so the ZED Boxes are the perfect compute module for the application (linux + GPU)

re: We'd love to! As we train Axol on tasks internally, we will open source the weights.

re: The stand is fixed but has locking wheels. Yes you can mount it anywhere you need. The mounting holes are in the datasheet.

agreed on the github link, we'll fix that now!

And YES having models fine tuned that an agent can select between would be amazing. As we get more Axol's into people's hands we'll make this happen.

@oliverbuilds Thank you! The first few folks we've been talking to happen to be focused on kitchens.

@marcelo_vegas Less than an hour! The kit comes everything you need to collect data & deploy a policy.

@amanda_gao2 Thanks! We're able to ramp up production to 100s of Axols a month.

@alina_tyslenok_ Thank you! We're excited to work closely with such teams.

@egret_studio Hi there :) I upvote to support the community spirit of the product. Wish all blenderers all the best✨

@samagra_gune Hi, there is support center on the landing page, thanks.

@novamaker01 Hi Felix, Each addon is tested and reviewed internally before goes live, a seller specifies which version of Blender their extension is compatible with.

Thanks

@dani_mashael Thank you for the support!

@blah_mad For changing state and assign labels, it can be done without approvals. For response, it is better for a manual review and approve.

@retain_dev In SeaTicket, we utilize both keywords and vectors to identify relevant issues, followed by re-ranking using an LLM. This approach ensures that the search results are highly accurate.

@anand_thakkar1 In SeaTicket, we first define a search tool that, upon receiving a query, identifies relevant issues and documents, followed by a re-ranking process using a large language model (LLM). This approach guarantees that the search results are highly accurate.

When presented with an issue, the AI agent will invoke the search tool multiple times with different queries until it believes it has gathered sufficient information to address the issue. Therefore, if the AI model is sufficiently powerful, it can reason about the answer even if one symptom can correspond to multiple potential root causes.

@bryan_williamson3 this could be great for students and researchers. A lot of learning content is already on YouTube, it just isn’t easy to navigate.

@bryan_williamson3 That is exactly the feeling we were chasing, thank you Bryan. 🙏 Over 70% of views on YouTube is algorithm driven. When you search something, actually it doesn't search as we know from Google.

@thamibenjelloun Hi Thami - in short both and more. Step one is context matching: does the transcript contain the answer, not just a matching title. Step two is authority, scored per topic, so "who knows about X" is judged differently than "who knows about Y." We lean on signals of real expertise for that domain, never raw channel size or view count.

@charan_tej_kammara Hey Charan. The creator and marketer side is something we think about a lot. Not built yet, but noted, and your framing (will what I put in my video actually land) is exactly the right question. What else would you want to measure?

@margarita_s88 Hi Margarita. This made our day, thank you :) Actually you can find lots of great videos to watch as well, because we do heavy search/job on YouTube to find what you're looking and we always discover lots of under-the-radar videos for the topic.

@varun1jan great to hear that you liked and find it useful. thank youu 🙏

@luca_capone Hi Luca - this is the exact use case we built the citation for, so it means a lot coming from you. On ranking: we never trust "most viewed." We first check whether the transcript actually contains a working answer to your specific question, then weight by how credible the speaker is on that topic, and for fast-moving things like coding we lean harder on freshness so a year-stale video does not win by default. It is genuinely the hardest part of the product and we tune it constantly. If you try a coding how-to and the ranking feels off, please let me know.

@alina_tyslenok_ Really appreciate it. "Searchable and verifiable" is exactly the bet, especially the verifiable part. If a claim looks too good, one click and you are at the source. What would you point it at first?

@farrukh_butt1 Hey Farrukh, indeed. It's one of our core feature and how we drive more value.

@retain_dev We run the full ROS 2 stack on the robot - Nav2, drivers, and the usual dependency footprint all live in our pre configured appliance, which maintains an authenticated low latency link to the platform. That lets you run full-stack robotics from a web session without installing a Linux robotics environment on every machine. Several sessions per robot share one appliance-side connection; the service distributes telemetry and state to connected clients and routes request/response traffic to the initiating session when required. Bottom line: ROS and time-sensitive compute stay on the robot, control and collaboration stay in the browser, and authentication, routing, and policy are enforced in the cloud.

@avd01 Love the idea of moving beyond the chat box. Having a private workspace where agents can actually connect to tools and run workflows in the background without dealing with Docker or cloud setup sounds like a huge quality-of-life improvement. Curious to see how people end up using this. 🚀

I think this is a great solution because it takes effort to set up your dev environment and the focus on non technical users is great as well because this is a category that is exploring these tools more and more and does require better setup to explore different tools. How do you handle the case if an agent is running a background workflow and then hits an error? Is this surfaced to the end user via some sort of notification system or is it moreso this is just the container and up to the user to handle this?

@avd01 Great idea. Many people experience AI tools stopping working as soon as their computer shuts down. It's great that you've created a solution where everything runs on the server and doesn't require deep technical knowledge. This really makes AI much more accessible to regular users.

@thamibenjelloun We don’t have a dedicated page for these settings. However, you can create a custom skill and assign it to your agent, which will give you this level of control.

@gabriella_anjani Thank you so much. That’s exactly the problem we’re trying to solve with Zingle.

Most learners don’t struggle because they can’t memorize one more word list. They struggle because the word gets separated from the moment where it actually made sense.

@alheri_murya Thank you, that means a lot.

That’s exactly the feeling we wanted Zingle to have. We didn’t want to build another streak-pressure app where the main goal becomes “don’t break the chain.”

For us, vocabulary learning should feel calmer and more connected to real understanding. If you meet a word inside meaningful content, save it with its context, and review it over time with SRS, it has a much better chance of moving into long-term memory.

That’s the direction we’re building toward: context first, memory second, pressure last.

@review2idea Thank you so much! Really appreciate the support and the thoughtful use case.

For Word Review, Zingle uses an SRS review system, similar to Anki. When you save a word, you can review it later with the original context sentence first, so you’re not just memorizing the word in isolation.

So the current flow is: saved word → original context sentence → recall → answer/details → SRS rating. The word details also include definition, sense explanation, examples, synonyms, antonyms, collocations, and pronunciation.

@novamaker01 Yes, that’s exactly the problem we’re trying to solve.

In Word Review, the goal is not to show only the saved word and explanation. You should also keep the original sentence/context connected to that word, so when you review it later, you remember how it was actually used.

So with something like Dutch news articles, you can save a word from the article, and later review it with the meaning, explanation, and the sentence/context it came from. That context is the main difference we care about, because isolated words are easy to forget.

@saba_k Thanks! It’s not a browser plugin right now.

Zingle works as a web app: you paste or add content you care about, then Zingle helps you discover useful words inside that context. You can save words, see their meaning and usage, and review them later with the original context connected.

The idea is to make vocabulary learning feel less like memorizing random word lists and more like learning from real content you actually want to understand.

This is cool trying to recall and make sure to check basic security settings is hard to always keep in mind and I feel like this is very relevant today with the high risk with AI agents. I like that you kept this lightweight and private as well avoiding a heavy MDM solution. How do you handle the case where you may need a system reboot or elevated permissions that could temporarily disrupt a local environment?

@spartan_phuothuynh Great question! Right now, you can absolutely share your canvas with a teammate so they can jump in and pick up right where you left off. It's async for now, but full real-time 'multiplayer' collaboration could be in our next update! Can we get in touch to learn more about your team workflow?

@tracy03 As someone who constantly bugged you for "just one more quick tweak" on campaign assets, Vellum locking in those brand styles is a lifesaver for our whole pipeline ;))

@ngochoang Yes. You can. After grouping nodes in Vellum, you can also @group to reference all items in the group. I also use this feature to tidy up the canvas then the campaign gets too complex ;))

@trang_do19 You can simply drop the reference image, and prompt AI to use the exact subject without edit. Choosing a Pro model also helps improve visual quality. You can see in the example here:

@creativewjordyn 
Building decks on Vellum is such a creative idea!! I also noticed that the AI agent is very good at setting up workflows. I did try to dump my campaign notes and it helps stucture into different nodes and draw connections between them. Super helpful!

@quynh_le5 We don't limit in one base image model but offer a range of models for you to choose. The list includes: Gemini, GPT, FLUX, Seedream, Ideogram, Stable Diffusion, Z-Image
Pick your fav model when you generate!

@marianna_tymchuk thank you, that means a lot coming from someone at Stripo. You are doing this for email, turning painful hand-coded HTML into something editable and reusable, so the living-document idea is right in your wheelhouse (those reusable content modules are something I keep eyeing for decks). One thing I am genuinely curious about: now that AI writes more of the email, do your users still want to drop into the HTML themselves, or has that mostly faded?

@lilamoreau Ha, the slide that picks the worst possible moment to betray you, right as the boss leans in. We have all been there. Glad those days are behind you.

@shivam_goyal10 Thanks, Shivam, this means a lot. Investor decks are about the highest-stakes thing you can put a tool on, so hearing the conversations got richer (not just that the deck looked nicer) is exactly the bar we were aiming for.

A core component of the same engine is what today's launch opens up; the deck builder is now open source, so anyone can start from one prompt and get a real, editable deck. Grateful you have been putting it to work where it counts.

@hiteshjoshi thanks. Getting a doc published on a domain that is exactly what we wanted the MCP server plus hosting to feel like, so good to hear it holds up in real use.

Today's launch is the open-source deck builder, the same one-prompt-to-real-output idea, and it plugs right into that same publish flow. Glad to have you on this one.

@philip_lr Thank you. Two things worth saying out loud for anyone reading: the deck builder is completely open source (MIT), runs standalone with no account, and the output is one self-contained HTML file. And yes, hosting is available via FluidDocs at fluiddocs.ai if you want a shareable link, fully optional, free to start.

So you can run the whole thing yourself and never touch our hosting, or lean on it when you want a link instead of a file.

@starkturtle great question, and honestly, the part I cared most about getting right. deck-import pulls the text, along with the original screenshot of each slide, and then rebuilds it as editable HTML. So charts and embedded media come back pixel-accurate, because it keeps the original image of that slide, while the text and structure around them become editable. Multi-column slides get classified and rebuilt rather than flattened. The honest tradeoff: a chart comes back as the original image, so it looks right, but it is not a live, re-editable chart yet. If you have a deck you can share, drop a link here, and I will convert it and post the result so you can see exactly how it holds up. If it is not shareable, DM me, and I will run it privately.