PH热榜 | 2026-02-18

@peter_albert2 Same, but it does what they say it does. Haven't had any errors in months. The price is worth it, doesn't mean its not expensive because it is haha. Offloading some code tasks to n8n to stabilize token hemorrhaging.

@anvisha_pai  will definitely check this out! Have been looking for something like this!

@anvisha_pai Flattened outputs are where design gen falls apart. Moda building a WebGPU canvas you can actually edit, plus an agent that understands layout, typography, and color, feels like a better path than prompt-only generators or template libraries. If the brand kit is treated as hard constraints, you get fast iteration without the export and rebuild loop. That's what keeps brand from drifting.

@anvisha_pai I'm curious how the "taste" layer works. Is it a custom fine tuned LLM or more of a rules based engine for the layout?

@huisong_li thanks!

@soubangi_rajkhowa thank you! a lot of early users have been using Moda for personalized HR letters, all hands decks, etc so I totally agree.

The next frontier is gathering more context and helping you with more than just design. Stay tuned!

@curiouskitty this question really gets at core questions we’re asking ourselves as well. First-off, we’ve already had early partners that are currently using Moda for client-facing decks! But anyone who has ever built an editor knows there are always more bugs, feature requests, and file types to be compatible with.

The team has put a lot of

time into testing and developing PPTX, GSlides, and Keynote both for import and export and our compatibility continues to get better. We also export PDF and import SVG, which hits some Figma use cases. Try it out yourself and tell us which features you need before it’s ready for your needs!

@curiouskitty what @steven_frieson said! Hundreds of users tried Moda during our closed beta

• One Series B's startups entire GTM team replaced contractors w Moda

• Multiple companies printed 6 ft+ posters for recruiting events

• Someone ran their first account-based marketing campaign

• A company already migrated off Canva

You can see more stories in the testimonials section on our website :)

@alina_petrova3 Yes you can! We were so tired of image gen getting so close but not close enough , and a new prompt being inefficient and cumbersome, so we made an agent that makes 100% editable designs. If you want to make a small copy tweak go for it! if you want to move around layers you can. Added bonus: the agent’s edits can be just as fine-grained as yours. No more starting over from zero with every prompt.

@alina_petrova3 As Steven said everything is editable in the canvas. You can resize shapes, edit text, change colors, drag images into the canvas from your desktop, etc. Everything you'd expect from a full-featured editor!

@pidsinee_einarsdottir the total reroll is so painful! That was one of the first problems we wanted to solve. Excited to hear what you think.

@pidsinee_einarsdottir yes, that's 100% the pain point we wanted to solve. Nano Banana and other image models are great but they give you a raster output that's not editable unless you re-prompt, and then you rarely get exactly what you want when you do that!

@b_nick can't wait for you to try it!

@b_nick thanks! very much looking for feedback on the slide import/export flows, so keep us posted!

@carissajansen thank you!! 🙌 So happy Moda’s been helping with customer-facing decks — what kind of presentations are you making most often?

@carissajansen Thank you Carissa!

@tanoy27 we've lived the pain 🫡

@tanoy27 Let us know what you think once you've had a chance to use it!

@alanka Thank you! Glad to hear you like it!

@alanka Greatly appreciated!

@rohithreddy Yes exactly - we built this for ourselves, the pain point we've faced so many times.

@maxkolysh Yes, exactly!

@alina_anitei Glad the demo video helped out. There are plenty more to come as we keep shipping new features.

@alina_anitei thanks so much! 🙌 if anything in the video or UI felt confusing / could be clearer, we’d love to hear it — even small notes help a ton.

@anvisha_pai  @ahmetardal Yes, absolutely. We use Moda for this ourselves internally - mock up UI in there, and then right now we export to PNG and just share the image with Claude Code. That is obviously not perfect, so we're working on an MCP that would let Claude (or any agent) just observe the design directly so it can make pixel-perfect designs. This is something WE want to use ourselves so it's definitely high on the feature priority list!

@ahmetardal a LOT of people have been asking us for this

@thekidsprinkles Thank you so much!

@ben_frank3 We haven't gotten to .odp yet, but we do have .pptx. Are you looking for import, export, or both?

@christo_tsvetanov That sounds like a great use case. It's definitely made to be beginner-friendly. There's a tutorial video on the homepage after you sign up before you create any projects.

You have two main options for your case though. You can create a brand kit by uploading your existing materials if you're going to keep going back to the same set for inspiration or you can easily drop your files into the chat as references for the agent.

@tonya_snezh No integration with Figma yet, but we've definitely imagined some use cases where it'd be helpful. What did you have in mind?

@tonya_snezh A couple other folks have requested a Figma integration so we'll likely prioritize it soon! Let us know what you're looking for (import, export, something else?)

@danimirror Hey Dani, quick thought .. SEO worked because Google had stable ranking mechanics. LLMs change rapidly and are less transparent. How does GEO avoid becoming a moving target that resets every model update?

@danimirror The way the tool is more customer-focused makes it stand out. Congrats on the launch!

Hey @olya_vasilevskaya !
thanks for the comment :D

We don't have access to OpenAI user's request (nobody does except them). We replicate the users' questions by asking daily and with different prompts to the AI engines (in this case to ChatGPT), so we reach a statistically true.

@gulipad Thank you! We're excited!

We’ve had a couple of really cool ones, but the one that caught our attention the most is the case of INDYA, a SaaS for nutritionists that grew from position 11 to position 2 across several prompts in less than a week by implementing the insights provided by Omnia.

In fact, you can find the full case study here

Hey @rogarmu8 !
Yes, it is! It will be here soon!

Hey kitty! @curiouskitty 

The best KPI's a team can define are:

• Share of Voice

• Visibility

• Mentions

The timeframe to see impact depends of the industry and the company, but we are seeing high impact in days, like with this case

Thanks! @rebeccahpearson 

We love it too :)

hi @tereza_hurtova !

Very good one :)

Omnia is a action-first tool, we are focusing on recommendations by insights and content (in the next release) to be very accurate to the content we offer to the customers.

Also, we're an agentic platform, we will launch soon the MCP too, so you'll be able to apply all the intelligence residing within Omnia to your day a day workflow.

Thanks for asking!

Hey@miguel_beltran_sanchez !

It has changed a lot!

We're redesigning a lot of features, and releasing others almost every week!

Sure @german_merlo1 !

thanks Jorge!!

Thanks for the comment@andrzej_pacholik !

Yes, we think so, all in with Omnia :D !

Thank you a lot! @benln 

@valeriia_kuna 

Eventually, we will give the user both options :) . At the end o the day, there will be always a human in the loop.

Hey @vasilbo !

Our platform is built around what brands can actually do to improve their presence, and we provide specific, actionable recommendations tailored to their brand based on what they’re monitoring. Also you have a 14 day trial, so you can check it out :)

hey @kay_siegert ! Thanks a lot for trying it out and for the great question 🙌

The prompts in Omnia Trends are trend estimations, not exact copies of a single user query. They reflect what’s most searched/asked across AI tools.

Since only platforms like OpenAI, Perplexity or Google have full access to raw data, our Trends data are estimated using:

• A dataset of real customer prompts (via a Chrome extension)

• SEO data

• Social listening across platforms

So something like “best microphone for beginners in 2026” reflects how users often phrase timely, future-oriented queries.

While volumes and difficulty scores aren’t perfect, they’re highly accurate and very useful for deciding what to optimize for.

Hey@juan_ros , thanks for asking!

We monitor at least daily. Sometimes if the monitoring data doesn’t converge (because LLMs are non deterministic by nature) we start monitoring more frequently until the data converges (basically until we find the statistical truth).

@pruscior congrats!

@pruscior What happens to downstream edits if you regenerate an extended or connected clip in Flixier after you've already trimmed the timeline? Generate from any frame is powerful, but it only stays usable if the prompt, reference frame, and model version stay attached to each clip. A quick compare between takes would make iteration feel safe.

@pruscior Love the “AI inside the timeline” framing. Curious how Flixier handles narrative continuity across multiple generated clips, does the model retain story context between timeline segments, or is each generation still somewhat independent?

@levent_askan Thank you, and we are excited to get this in front of the PH community. Bringing AI video generation directly into the timeline felt like the right next step.

@osman_kocs that’s great to hear!

That fear of ending up with a folder full of almost-useful clips is exactly what we were trying to remove. We really wanted experimentation to feel safe, knowing you can actually turn it into a finished video.

Thank you @cam_strive , for putting this into words so clearly.

@cretueusebiu really glad to hear that!

@otodidakt_20 thanks, really appreciate the support.

@razvan_girmacea1 Thanks a lot for supporting us since the early days!!!

@neilverma Exactly, no restarts, no jumping between tools. Everything stays in the timeline where you're already working. Thanks, glad it landed!

@rohithreddy Thank you Rohith! That means a lot. Visibility into what the agent is doing was the #1 thing I was missing when I started building with OpenClaw, so I built it. Glad it resonates. Would love to hear how you use it with your setup!

@saezbaldo Great question, Damian. Tracing chained agent calls is exactly where most logging tools fall short.

ClawMetry tracks every tool call, sub-agent spawn, and session handoff with full context. So when agent A calls agent B which queries a database and picks the wrong follow-up, you can trace the entire chain: what each agent saw, what it decided, and where it went wrong.

The cron management dashboard also helps here. You can see scheduled tasks, their run history, and drill into individual executions. No more guessing which step broke.

We're actively building deeper tracing for multi-agent workflows. Would love your feedback if you try it out: pip install clawmetry and you're up in 30 seconds.

Wow and it's also straightforward to use. I really like it

@busmark_w_nika Yes you could export videos in 1:1, 9:16 or 16:9 aspect ratios.

@rohithreddy Thanks

@curiouskitty This was a user-demanded feature that was just waiting to be executed.

As for the constraints and trade-offs, to be honest, we didn't think much about them.

@angell019  @neilverma Thank you.

@wyattchang11 what was the hardest part of building Empirical for web?

@wyattchang11  In your opinion, which statistical models you’re using for 30-year risk forecasting, are these based on established frameworks like pooled cohort equations or something proprietary? The predictive angle is very compelling !

@curiouskitty we have a few features that take the results from just "numbers on a screen" to actionable change. When results are back people receive an AI generated report for each biomarker. We also have an in-app tool that uses your biomakers to create nutrition recommendations. This can even prompt you when you enter a restaurant on nutrition choices that would be best for you. Our medical team is also available to answer any questions and our year-long members can access doctor check-ins to review the results, get their questions answered, obtain prescriptions if necessary, and ensure we support each individual roadmap toward a long and healthy life!

@valeriia_kuna our platform can be used across many countries - people can upload their lab results, pair their wearable devices, and take steps toward improving their health outcomes. In the US we are also able to provide medical care and our year-long program includes doctor check-ins to ensure our members get a deep dive into their results and a customized action plan. We would definitely want to expand this component at some point so that it is available to people worldwide!

@garrytan  @ehudhal When I'm in a tight MVP sprint, the agent starts freelancing, new spacing, new radii, slightly different blues. Design Rails shipping a design-context folder with agent-instructions.md and W3C design-tokens.json is the kind of handoff that actually works. Do you also generate CSS variables or a Tailwind theme from the tokens so the app and landing page don't drift apart? That last mile is where brand kits usually fall apart.

@t1m0slav Thanks for giving it a try, and for the feedback!

The logos are a fun problem :) If you gave a thumbs down to those that weren't helpful that'll help us look into the specifics and tune it up further. Or, if you're open to sending over the thread ID (URL) over to hello@designrails.com that would be much appreciated.

In many of the runs, we've been blown away by the creativity the AI achieves (looking at you, nano banana) in capturing concepts, and reflecting the desired style. It fails at times, and there's more work to do! What often works well is going beyond the single-shot, asking for (free) variations and steering it with revision guidance to hone-in on the concept you have in mind.

Hi @sam_alexander1 Design Rails is the name of the product. Company is Chordio. https://www.ycombinator.com/companies/chordio

@rohithreddy Thanks, Rohith!

@curiouskitty- that's pretty much the process. You provide the partial brand (e.g. hex values, typography, visual concepts) as an input to give an initial direction, and you get a more complete brand system. If we define "stable system" as "the UI I'm building doesn't look generic and the AI consistently follows the brand guidelines" - once you're happy with the brand and add the exported package as design context, the LLM picks it up immediately and does a really good job following it.

@vasilbo thanks for checking it out and for the feedback. 🙌

@neilverma thanks Neil!

@matthias_stralman 100%. Our app Subspace has a Rust backend. I’ve never written a single line of Rust before we started building it. LLM’s are a universal translater and the language is system architecture!

SPECTRE uses a manual /handoff that auto inserts a status report into the next session - complete continuity until you run /forget. And the learning Skill documents features that get auto loaded when relevant. It’s really simple but it works really well.

@vouchy oh man all the time. I actually experimented with the first version of our app Subspace and built two versions - one with Spectre (an early version) and one without. The without version looked better, but once you got down one layer it was a complete untangle-able mess. The Spectre version is what exists now.

@stellify_software I appreciate it Matt!

@alpsu_dilbilir Thanks a lot, really appreciate it!!

Yeah, I spent quite some time on the UI. You can actually customize pretty much everything and create completely different presets and styles.

The only real limit is your imagination :)

Glad you like it!

@kkyshko Thank you, really appreciate it !!

It updates using a Shortcuts automation. You basically set it once, and it regenerates the wallpaper automatically at whatever time/frequency you choose.

There’s a simple step-by-step tutorial inside the app that walks you through it, takes less than a minute to set up.

@shaunscdmx What happens when the person learns it was an AI they were talking to and not you? I think some people would be very annoyed that they were lied to.

@rohithreddy Thank you

@vrun Love the “diagram as code + AI” combo! Where do you see OpenFlowKit’s biggest edge versus tools like Excalidraw or Mermaid Live, aesthetics, AI context awareness, or workflow flexibility? Also, when you say Flowpilot understands diagram context, does it reason about structure (nodes, dependencies, hierarchy), or mostly operate at a visual/layout level?

@martin_hernandez4 

Thank you so much for the incredibly kind words! Really glad you’re enjoying the UI and the open-source aspect, it’s the core of why we built this.

Regarding your questions:

1. AI Providers: Currently it’s BYOK for Gemini only, but we built the service layer to be modular. OpenAI and Claude support are the top priorities for our next update. The experience will be identical, just pop in your key and go.

2. Dark Mode: 100% on the roadmap! We're planning to ship 'Dark mode' soon. You can expect this and more model support by early next week.

3. P2P Collaboration: One big thing we're planning is P2P Multiplayer via WebRTC. This will allow two people on the same Wi-Fi to brainstorm on one canvas with zero database or cloud infra required, just browser-to-browser sync.

4. UX & Core: We're also working on a more persistent Flowpilot AI sidebar for better back-and-forth refinement, along with improved node components, auto-layouts, and higher export fidelity.

Thanks again for the support! 🚀

@wilco_kruijer1 

Thanks so much for the kind words on the design! We just pushed a few updates specifically to address your experience with Flowpilot (AI mode).

1. Model Selector is LIVE: We just added a model selector in Settings > Brand Kit > Flowpilot. You can now choose between 2.5 Flash, 2.5 Pro, and even the new Gemini 3 Pro/Flash (Preview) for much better reasoning on designs. We’re also working on adding OpenAI and Claude models to give users more flexibility across top AI providers.

2. Optimized Layouts: We’ve switched our auto-layout engine to a dedicated Tree algorithm (mrtree) for AI results. This produces significantly cleaner, more professional hierarchies than before.

3. Refinement is supported: You can actually keep chatting with the AI to tweak the result (e.g., 'add a retry loop' or 'change the theme to blue'). However, we realized the current UI (closing the modal after generation) makes this feature feel hidden. We're working on a persistent sidebar chat so you can refine side-by-side with the canvas.

If you’re open to sharing the prompt that gave you a less-than-optimal result, I’d love to test it against our new layout and model updates!
Please use this form: CLICK HERE

Really appreciate the candid input, it helps us move way faster. 🚀

@baseline_studio On the skills side, what does a Baseline Core skill look like under the hood, is ux-design just markdown in skills/, or can a skill run code too? The npx @baseline-studio/cli init scaffold that drops skills/, context/, frameworks/, and AGENTS.md feels like the right path for tool-agnostic use across Claude Code, Cursor, Codex, and Copilot. The trust builder for teams is drift control, version skills plus core/voice.md, stamp outputs with the versions used, and ship a tiny eval set so a tweak doesn't silently change PRDs or sprint plans.

@baseline_studio Trent, how flexible is the configuration regarding what exactly gets fed into the context? For example, if I'm making a skill for DB refactoring, can I feed it only the database schema and migrations, ignoring the marketing docs? I want to keep the focus narrow so the model doesn't get distracted or start hallucinating

@isthispalash This is incredible feedback. The fact that you've gone from experimenting with it to adapting it to your own workflow is exactly what the system is designed for. The context layer is what makes that possible. Once it knows your business, every skill gets more specific to how you work. The "impact of a team of ten" framing is something I think about constantly, that's the real unlock here. Not replacing people, but giving a solo founder or small team the structured output that used to require dedicated roles across product, design, marketing, and strategy. Really appreciate you sharing this!

@cogotemartinez Each skill has a manifest.yaml that declares exactly which context, frameworks, and reference files it needs. Reference files are nested inside each skill's folder, so the UX design skill has its own set of 6 references while research synthesis has a different set. The agent reads the manifest first and only loads what that specific skill requires. Not all 34 at once. Agents[dot]md is what connects everything. It routes tasks to the right skill, tells the agent where to find the manifest, and defines the execution protocol. The workflow orchestration framework also handles context management across sessions. For larger tasks it scopes work into multiple sessions and generates a plan so context stays clean and output quality doesn't degrade. That said, managing context within longer sessions isn't always consistent yet and it's something I'm actively looking for better solutions to.

@dan__cleary Thank you! I haven't used OpenClaw yet, but Baseline is modular by design. The skills, frameworks, and context files are all standalone markdown, so in theory they should be portable into other systems but I'd love to explore that more!

As for how I curated the skills, they come from my experience across the full product development spectrum. Each one maps to a phase of pre-engineering product work that I've done hands-on and refined across client engagements.

@gabe Haha you’re way too quick 😆 225 was insane! I just barely hit 279ms 😂 That Reflex Rooster is turning me into a chicken ninja 🐔💥

You should def share it with your friends and see who’s the real cluck-master 😂✨ Try it again at

https://www.reflexrooster.com/ — upcoming updates are gonna be even better (leaderboard + more things 😎🔥). Come beat me again!

@gabe just added a leaderboard and some cool new features at https://www.reflexrooster.com/ 😎🔥 Come try again and climb the ranks!

@guibassa oh wild, I didn't know that but that does feel like it could be gamed! That's a good find.

@otodidakt_20 Thanks so much! Really happy you loved it 🎉😊 Means a lot! If you have any feature ideas or feedback, I’d love to hear them 💬✨

@mhoussam9 Last quarter I wired up a Clay table to enrich about 2k contacts and blew through credits in two days. The markup on enrichment calls versus hitting the APIs directly was painful. Self-hosted and open source changes that math completely... you bring your own keys, you control per-call cost, and the workflow logic lives in your repo instead of a proprietary canvas. A library of pre-built enrichment recipes mapping to common Clay table setups would make migration dead simple.

@mhoussam9 but first, congrats on the launch!

@andresolarte Why not :)

Tbh, I don't know why we picked it, just needed a logo for the moment and it stuck 😅

@muhamad_jawdat_akoum Stepping away mid-run is when the permission prompt hits and Claude Code just waits. yottoCode moving the loop into Telegram, plus emoji reactions as low-friction acks and voice notes for intent, feels like the right direction. If reactions can map to approve, pause, cancel, does it also keep a tight allowlist plus redaction so paths and secrets don't leak? That combo would make phone-first runs feel safe.

@muhamad_jawdat_akoum Love the remote dev workflow idea! The emoji reactions for token efficiency is clever.

Question: How do you handle security with Telegram integration? Curious about the architecture since I'm new to this space.

Also just launched my first app today (FireBlazer) - fellow first-time maker! 🚀

Upvoted yottoCode!

@_anasansari we use the official Anthropic Claude Agent SDK. no workarounds.

@_anasansari you can use yottoCode for free!

@piroune_balachandran Thanks for the question and minds!

At this moment Konta used only for deployments to automate manual steps. If a deployment fails, the stack update is considered unsuccessful. So you don’t end up with a half-switched infrastructure state. Konta tracks which services changed between commits and redeploys only changed.

I want to implement health-check feature after deployment. Because now it only verifies that containers are in a running state. With this feature Konta could decide that new vps state is broken and deployment was failed.

@vlad1323 Letting an agent touch prod APIs is where it gets scary, a prompt injection can turn into real side effects. Cencurity as a security gateway makes sense if it can fail closed on unsafe tool calls, force supervised mode for state changes, and keep an audit trail tied to user identity. Does it support per-tool, per-action scopes plus redaction at the gateway? An OpenTelemetry export to your SIEM would make adoption way smoother.