PH热榜 | 2026-04-03

Hi Product Hunt! I'm Ning, founder of ZooClaw.

Back in February, I was playing around with OpenClaw and built an AI companion agent — just for fun. I shared it with my team.

What happened next really surprised me.

My HR lead — zero technical background — started playing with it and somehow turned her own expertise into a career planning agent. 33 iterations in one afternoon. It's now live on ZooClaw for anyone to use.

Another colleague built a social media agent. A post it created went viral overnight.

People didn't just use the agent — with the right tool, they started creating their own.

That's when it clicked: AI is incredibly powerful — but it needs the right people to guide it. Everyone has expertise that could help thousands of others — they just never had a way to turn it into something that scales.

So we built ZooClaw — a platform where your expertise becomes an AI specialist that works for you, and for others.

🦊 One entry point, multiple specialists — Fox for marketing, Owl for office tasks, Beaver for data analysis. The right agent picks up the right task automatically.

⚡ Proactive, not reactive — Your morning starts with results already waiting for you. Scheduled tasks, monitoring, follow-ups — handled while you sleep.

🔧 Zero setup, zero token anxiety — No API keys, no deployment. Best models first, open-source fallback when needed.

💬 Voice-first — Talk to your agents like you'd talk to a colleague. No prompts to craft, no UI to learn.

The era of the one-person company is here. But even a one-person company deserves a full team. That's what ZooClaw is — your team.

We're still early. I'd love to hear — what expertise do you have that you wish could work for you around the clock?

We're here all day. Your zoo is waiting 🚀

Interesting! But if multiple agents can handle the same task (e.g., marketing or research) how does zooclaw decide which specialist is actually the best fit in real time?

Really liked the story, especially the part where your HR lead kept iterating and actually built something usable, that feels pretty engaging. wonder what part do non tech users usually get stuck on when they try to build their first agent?

Congrats on the product launch! I'd love to have Fox beside me and handle routine marketing issues. But how do you manage to consolidate enterprise-level context that is embedded in various systems and files, across multiple functions and apartments?

Interesting angle.
Feels like the market is moving from AI as assistant → AI as operator.
Curious how much of this is real repeatable execution vs strong launch storytelling

This is really cool. Can a specialist hand off part of a task to another one mid-conversation?

the "no token anxiety, no setup" angle is genuinely clever positioning. most people who'd benefit from a multi-agent setup are scared off by the infrastructure overhead, and removing that friction to get to an immediately useful team of specialists is the right instinct.

the tricky part will be routing quality on ambiguous or cross-domain requests. a single entry point works cleanly when tasks are discrete, but "help me prepare a business case for this new feature based on our usage data" spans writing, analysis, and product thinking at once. getting the routing to coordinate across agents or correctly decompose the task is where these systems tend to fall apart, and the failure mode isn't obvious to debug.

Congrats for the launch!

But whats the difference from Openclaw?

the "no token anxiety" line hits hard. constantly monitoring usage across different APIs is such a productivity killer. how does the fallback to open-source models work when the main ones are overloaded? does it maintain quality or just keep things moving?

congrats on the launch, the proactive scheduling angle is genuinely different from most agent tools i've seen.

one thing i'm curious about though. "zero token anxiety" sounds great as a user but someone's eating that cost. is there a usage ceiling on the free tier, or are you subsidizing compute to grow and then switching to a credit model later? asking because i've watched a few AI tools launch with generous free tiers and then hit a wall when the unit economics catch up.

not trying to be cynical, honestly excited about what you're building.

"The era of the one-person company" resonates hard. Built Krafl-IO solo and the biggest challenge isn't the code, it's wearing every hat simultaneously. The idea of specialized agents handling different domains is compelling. We use a similar approach but narrower- 3 agents that each own one step of LinkedIn post generation. Curious how you handle agent handoffs when tasks cross domains.

Looks cool, is this built on openclaw?

a really cool zoo

Congrats on the launch, Ning!

You mentioned a colleague built a social media agent and a post went viral overnight. Would you mind sharing the skill?

Google's Gemma 4 looks like a serious leap forward in open AI models.

An open model family built for advanced reasoning and agentic workflows, it solves a key problem: getting frontier-level intelligence without massive compute costs or closed ecosystems.

Stands out for its intelligence-per-parameter — outperforming models up to 20x larger while running efficiently on phones, laptops, and desktops.

Key Features:

• Advanced reasoning – Strong multi-step planning, math, and instruction-following

• Agentic workflows – Native function calling, structured JSON output, and system instructions

• Multimodal capabilities – Supports images, video, and audio inputs

• Long context window – Up to 256K tokens for handling large documents and codebases

• Code generation – High-quality offline coding and local AI assistants

• 140+ languages – Built for global, multilingual applications

• Hardware efficiency – Runs across mobile devices, laptops, and GPUs

It’s open (Apache 2.0), meaning developers get full control, flexibility, and the ability to run and fine-tune locally or in the cloud.

Start experimenting with Gemma 4 now in @Google AI Studio 2.0 or download the model weights from:

1. Ollama

2. Kaggle

3. LM Studio

4. Docker

5. Hugging Face

Who's it for? developers, startups, and enterprises building AI agents, coding assistants, multimodal apps, or privacy-first solutions.

Whether you're building global applications in 140+ languages or local-first AI code assistants, Gemma 4 is built to be your foundation.

Read more here:

• https://deepmind.google/models/gemma/gemma-4/

• https://blog.google/innovation-and-ai/technology/developers-tools/gemma-4/

• https://opensource.googleblog.com/2026/03/gemma-4-expanding-the-gemmaverse-with-apache-20.html

Curious how it performs in real world coding tasks compared to larger closed models, especially for niche stacks.

Just posted about this on X today. Apache 2.0, runs on your own hardware, 256K context window. The fact that you can run this locally on a laptop and still get serious reasoning is wild. I'm curious how the Flutter/Dart code generation compares to the bigger closed models since that's most of what I write these days.

Congrats on the launch! What design choice had the biggest impact on getting this level of performance while keeping compute requirements so low?

This will make amazing local experiences for app creators, cant wait to test this in my App, been usung gemma3:4B with excelent results, so this is excelent news....Thank you Google

The agentic workflow angle is the interesting part for me. Most open models get benchmarked on reasoning and coding, but the harder question for production use is how they handle multi-step tasks where the model needs to recover from partial failures.

Running Claude Code agents in parallel - local inference becomes appealing but reliability in long workflows is still the blocker. Anyone tested Gemma 4 on tasks with 10+ tool calls?

curious about the "low compute overhead" claim - are you seeing meaningful performance gains over Llama models in the same parameter range? we're always evaluating new models for healthcare applications where inference speed matters a lot.

Cursor 3 is a unified workspace for building software with agents — faster, cleaner, and built from scratch around how engineers actually work with AI today.

• All agents in one place: local and cloud agents in a single sidebar, including ones kicked off from mobile, web, Slack, GitHub, and Linear

• Parallel agents: run multiple agents across different repos simultaneously, with demos and screenshots to verify cloud agent output

• Local ↔ cloud handoff: move agent sessions between environments instantly — push to cloud to keep running while offline, or pull local to test and iterate

• Diffs and PRs: review changes, stage, commit, and manage PRs from a cleaner diffs view

• Integrated browser: agents can open, navigate, and prompt against local websites directly

• Cursor Marketplace: browse and install plugins that extend agents with MCPs, skills, subagents, and more

• Full IDE depth: view files and go-to-definition with full LSP support, anytime

Parallel local/cloud agents in a unified workspace is the right direction - context switching between Claude Code, a cloud runner, and your local terminal is where a lot of time disappears right now. The MCP piece is interesting too. How does Cursor 3 handle agent conflicts when two parallel agents try to touch the same file or resource at the same time?

I have been having a great time with Cursor IDE. But there seems to be a lot of updates in recent times that seem so hard to catch up with. I have also been seen some breaking in external integration

Curious how this handles context across multiple agents running in parallel — does it maintain shared memory between them?

Congrats on your launch, Cursor team.

This developer tools race is insane.

You pick a tool, dive into it, and use it until their competitor releases a new version. I'm jumping between Cursor, Claude Code, and Codex.

Wow! Can't chase all these updates. But I'll get it anyway. All the best team!

curious about the MCP integration piece - are you supporting custom protocol handlers or just the standard anthropic ones? we've been building some healthcare-specific MCPs and the tooling around that ecosystem still feels pretty early

Parallel local/cloud agents in one workspace is the right direction - context switching between Claude Code, a cloud runner, and local terminal is where a lot of time disappears right now. How does Cursor 3 handle agent conflicts when two parallel agents try to touch the same file at the same time?

most ai coding tools still assume you're working on one thing at a time in one place. you kick off a task, wait for it, then start the next. if you want to run something in the cloud, that's a different context, and if you need to review the diff or open a pr, you're back in the browser or a separate terminal tab. the agent does work, but the workflow is still fragmented.

what cursor 3 is doing differently is treating the workspace as the unit, not the editor. being able to run parallel agents across repos, hand off sessions between local and cloud without losing context, and review prs from the same place you kicked off the task, that collapses a real chunk of the context-switching overhead that currently just gets accepted as the cost of working with ai.

Parallel agents is what I've been waiting for. Right now I'm bouncing between Claude Code in terminal and Cursor for UI stuff and the context switching kills the flow. If this actually lets me run local and cloud agents side by side in one workspace, that changes how I build. How's the MCP setup? Plug and play or does it need a lot of config?

Hey Product Hunt,

I’m Jonah, co‑founder of VoiceOS.

We started building VoiceOS around one frustration: the gap between deciding to do something and actually getting it done on your computer.

You think: “Reply to that Slack message.” That’s half a second.

Then you cmd‑tab, find the channel, scroll to the right thread, type, re-read for tone, hit send. It’s 30–60 seconds of mechanical overhead for a half‑second intention.

VoiceOS closes that gap.

Huge shoutout to our Japanese community! We’ve invested in native Japanese localization across the UI and writing behavior because you’ve been incredibly supportive from day one.

VoiceOS includes a free 14‑day Pro trial. No credit card required.

What integrations should we build next?

Super cool, will definitely try!! But how have you balanced speed and accuracy, especially when a wrong action (like sending a message) has higher stakes than wrong text?

voice productivity tools have mostly tried to make typing faster, which misses where the time actually goes. the real friction in knowledge work is the navigation overhead: app-hopping, finding the right thread, clicking into the right field. voiceos is targeting the context-switching cost, not the keystroke cost, which is a meaningfully different problem to solve. the confirmation step before each action is a smart call too. it sidesteps the 'what if it does the wrong thing' concern that's quietly killed most voice automation products in the past.

I've been using VoiceOS for about a month+ now and it's INSANE. I've been able to navigate different apps, talk to my LLMs and go through my workflow a lot faster than before. It's like giving @Aqua Voice apps and tools. I've even been able to use it to talk/send messages to @Poke.com.

I was lucky to speak with the founders and their vision is definitely promising. I'm looking forward to seeing how voice can become the main interface of my computer.

For those looking to try, there are some kinks, but the team is super responsive in their Discord and patch out any bugs super fast.

Congrats @jonahdaian and @kai_brokering on the launch! PS - more integrations please!!!!

This is awesome. Like WisprFlow (which i love) on steroids

(typed by voice OS) I've been using VoiceOS for a few months now, and it has helped my work tremendously. It makes tasks easier and more effective, and I can't really reply to emails without it lol

Hi my son who has Dyslexia and relies exclusively on speech to text, now uses VoiceOS exclusively. This is as big of a product endorsement anyone possibly could make.

This is really cool, I loved the intro ad, a lot of effort went in to this! Correct me if I'm wrong but this is like open claw with voice functionality right? just with no messy setup part and one more thing what extra features we get above open claw?

Happy launch day!

This looks really smooth. Curious about the workflows - how deep does the customization go?

Can users build their own multi-step chains or is it more about the pre-built integrations?

This is interesting concept and congratulations to you.

Is this available to download for Windows?

Cool launch. One thing I keep bumping into in the voice space — there's a real split between voice→action (do this thing) and voice→data (fill in 15 fields). For actions the confirmation step makes total sense. But for structured input like CRMs or intake forms, confirming every field kind of brings back the friction. Are you guys thinking about that side of things or staying focused on workflow automation?

Love the mission of cutting down on mechanical overhead, but how are you ensuring VoiceOS handles nuanced commands in dynamic conversations, like coordinating on Discord or Slack?

Looks cool! Can I also link this to Claude Code and use it for my coding tasks? Or does it need to have a separate integration for that?

Visual communication has always been the bottleneck nobody talks about.

You do the research. You synthesize it with AI. Then you paste it into Canva, fight with layouts, pick fonts, give up, and ship something mediocre. The insight was great. The visual killed it. 😮‍💨

NotebookLM just quietly patched that gap. 🔧

Custom Infographic Styles gives you 10 visual presets plus a fully custom mode to transform any source material into a publication-ready infographic. One click. Source-grounded, not hallucinated. ✅

The styles are genuinely varied:

🌸 Kawaii for approachable, shareable content

📰 Editorial for thought leadership pieces

🗂️ Bento grid for modular, social-first layouts

🧊 3D clay for visual pop without the Blender learning curve

✏️ Sketch, anime, storyboard, bricks, professional rounding out the rest

The deeper unlock here is the custom mode. 🎨

You describe the aesthetic you want and NotebookLM builds around it. Color codes, tone, layout philosophy. This is where brand consistency starts becoming possible inside a research tool.

Who this is for right now:

🔬 Researchers who need to make dense material accessible

🏫 Educators turning syllabi or notes into visual explainers

🚀 Founders building decks or one-pagers from raw research

📣 Content teams who live in NotebookLM and hate the Canva commute

The honest limitation: you can't manually drag or edit elements. It's prompt-driven. If you need pixel-perfect control, you still go to Figma. But for "good enough to publish in 30 seconds," this clears that bar easily.

The trajectory here is what I find most interesting.

NotebookLM started as a research tool. Then it added audio overviews, slide decks, mind maps. Now styled infographics. It's quietly becoming a full production layer on top of your knowledge. The gap between "understanding something" and "communicating it visually" is collapsing.

Curious what use cases people here are most excited about. Are you using this for personal knowledge work, team communication, or client-facing outputs? 💬

Hey PH! 👋 @markmdev and I made Straude.

The NYTimes recently covered the "tokenmaxxing" trend sweeping across Silicon Valley — the idea that AI token consumption is becoming the new status metric for AI-native builders. Companies like Meta, OpenAI, and Shopify are tracking it on internal dashboards. Jensen Huang proposed giving engineers a token budget as high as 50% of their salary.

At the same time, I found the experience of agentic coding gets pretty isolating. You spin up dozens of agents and spend thousands on tokens, but have no way to track, compare, or share the journey. And since there’s no “town hall” where AI builders gather, you can’t easily learn from how the best builders are pushing these tools to their limits.

Straude fixes that:

• One command `npx straude` logs your Claude Code or Codex sessions: tokens consumed, spend, models used. Support for other agents coming soon.

• Compete on a global leaderboard, build your streak, and get inspired by what the top 1% of AI builders are pushing daily.

• Privacy-first: only aggregate stats leave your machine, no code or session history.

• Zero install, zero config.

How it works:

1. Sign up at straude.com

2. Run `npx straude` at the end of a Claude Code / Codex session

3. Your stats sync to the feed and leaderboard automatically

It's completely free. The project is open source on GitHub.

Strava made lonely sport (running) social and fun. Straude does the same for agentic engineering. You wouldn't train for a marathon without logging your miles. Why would you burn through tokens without tracking on Straude?

We'd love for you to sign up, log your first session, and share Straude with the friend you know who burns the most tokens 🔥

Would love your feedback — what features would you want to see next?

Awesome timing with this one. The leaderboard is a fun idea for the agent builders out there. Do you have plans to add team-based leaderboards or just keep it individual?

The end-of-month billing surprise is too real. I've been running Claude Code in parallel on a few projects and had zero idea what I was spending until the invoice hit - it's kind of absurd we don't have native visibility for this.

Strava framing makes sense here. Is the leaderboard purely by USD spent or is there any weighting for output quality? Curious what "winning" looks like on this.

This is the most 2026 product I've seen all week and I mean that as a compliment. I'm genuinely curious what my monthly token spend looks like across projects because right now I have zero visibility on that. Does it break down by project or is it one global number?

So on this leaderboard what gets more weightage I mean does the top ranker has most amount of tokens spent or most amount of output from AI? I mean what is the ranking criteria?

Hi everyone!

Qwen3.6-Plus is a massive upgrade over 3.5, especially with its emergent agentic coding capabilities. In real-world scenarios—from frontend web development to complex, repo-level tasks—it can autonomously break down the problem, plan a path, test, and iterate until the job is done.

Built on native multimodal data, it can look at UI screenshots, design drafts, or natural language descriptions to generate frontend pages, complete code, and modify interactions. It truly closes the loop from "understanding the UI" to "generating code" and "invoking tools to modify it," making multimodal models genuinely practical for everyday dev workflows.

It not only handles complex code management and cross-domain long-term planning for pros, but also drastically lowers the coding barrier for everyone else.

Qwen3.6-Plus is now generally available through the official API. You can seamlessly integrate it with @OpenClaw, @Claude Code, @Kilo Code, @Cline, @opencode and Qwen Code.

Bonus: Sign in with Qwen Code OAuth to enjoy 1,000 free calls per day!

Love this update from OpenAI 👏

What it is: ChatGPT is now available inside Apple CarPlay, enabling voice-first AI conversations while driving.

Problem → Solution: Typing or interacting with apps while driving is unsafe. This brings a hands-free, voice-native AI assistant directly to your car.

What stands out: Built specifically as a voice-first experience for driving, not just a mirrored app.

Features & benefits:

• Start or resume chats via voice

• Continue conversations or projects on the go

• Automatic voice mode option

• Works globally across all ChatGPT plans

Who it’s for: Drivers, commuters, and anyone who wants AI assistance safely while on the road

Use cases:

• Ask questions hands-free

• Continue work conversations

• Brainstorm ideas while driving

Limitations:

• Manual Launch: There is no "wake word" support; the app must be opened manually.

• No Visuals: The interface is intentionally barebones, focusing solely on voice communication.

• No Vehicle Control: ChatGPT cannot control car functions (climate, audio, navigation) or phone settings.

Clean, practical, and safety-first execution. Big step for ambient AI 👌

Love this!

My old Mercedes has a terrible infotainment system, so I have to rely on my phone for maps, music, and podcasts. It’s really dangerous while driving.

ChatGPT for CarPlay looks like the perfect solution! Thanks for sharing, Rohan!

I see, its like finding diamond in mines of coal! But if everyone can spin up ai startups instantly, what becomes the real scarce advantage - distribution, ideas or something else?

Cool! It’s time to get started on my side project 👊🏻

Congrats on the launch! The Kevin Rose example is a clever proof point, but he’s not starting from scratch - he has an audience and name recognition doing most of the heavy lifting. What does the timeline to a first paying customer actually look like for someone with no existing distribution?

Happy launch day! The $1 for $50 ad credit is a clever way to get started. How does Otto decide which ad channels to use for that first batch of tests?

The question in the comments about what becomes scarce is the right one. I think it's judgment - when to kill an idea, when to pivot the positioning, when a channel isn't working. Otto can execute fast but someone still needs to make those calls.

Curious how you handle that - does it surface decision points to the founder, or mostly runs until metrics show failure?

Really love how fast it is for people to go from "someone should build a company doing x - to execution with these tools."

I built most of what's under the hood here. The thing that excites me most isn't the landing page generation or the ad setup - those are cool but honestly table stakes at this point.

It's the autonomous loop. Otto doesn't just execute a checklist and stop. It watches what's working, adjusts, and keeps going. That's a fundamentally new behavior – and it's fascinating to se how resourceful Otto can be, which as long been a characteristic of entrepreneurialism that we've studied.

Also - if you're a dev, check out the agent skill on GitHub (prehype/audos-agent-skill). You can plug this into Claude Code, Cursor, or any agent that supports skills. Would love to hear what you build with it.

Good idea, it could be cool, something like that label maker pistol with a camera integrated into this device and make those stamps physical :)

Something like this (but less 00s and more 2020) :D

That's great one! i like the idea. its some nostalgic memories while creating, thanks

You made me miss my old stamp collection, oh the memories!

Exciting to see a CMS building so directly on top of Astro! Love that the importance of extensibility and plugins is being thought about from Day 1.

Congrats on the launch! The architecture here is really well thought out. I’m excited to watch this ecosystem grow and build on it!

This may be the most under-appreciated launch of all times. It's about time for a lightweight WordPress alternative.

I run my main website on Ghost but I'm open to switch.

And how awesome is the name?! 🫸🫷

I've been looking for a WordPress replacement for some of the older websites I manage for a while. All of my latest commercial clients' websites run on Astro, so I'm already familiar with how it works. I'm looking forward to seeing where EmDash CMS is heading, and I'll consider migrating to it in the future once it's a bit more mature and stable.

Hi PH! After years of building design tools, one thing really stuck with me: most people are not trying to become designers. They don’t want endless controls or a “powerful” platform full of options. They just want to get the job done without fighting the tool.

That’s the idea behind GeneratePPT.

I kept seeing the same pattern in my previous products too. People were not asking for more complexity. They wanted less friction. Less time wasted tweaking things that should have been simple.

And presentations are one of the clearest examples of this. You open a slide builder to make something quickly, and somehow end up spending way too much time aligning elements or digging through options, when all you really needed was one good layout that looks clean and professional.

GeneratePPT is my response to that.

It’s not a creative suite. It’s a tool built to help you go from idea to polished deck fast, especially when the content matters more than playing designer.

It won’t be for everyone. If you want a million knobs and sliders, you’ll probably hate it! But if you want to make a solid presentation and move on with your day, that’s exactly who I built it for.

And personally, that matters a lot to me. Building tools like this has given me something I value more than money: time. That’s why I care so much about building products that save other people’s time too.

So we give the AI details and get the PPT's generated right? what extra do I get by using generate ppt instead doing it from general LLM's?

Instant slides that don't look like an AI made them is harder than it sounds. How much control do you have over the layout after generation?

I love the anti-bloat idea, but how do you handle more complex content needs like animations or custom branding?

Creating a good PPT is always a pain point. Can I manage more than 1 brand under single subscription, or is it 1 account = 1 context?

most ai slide generators fall apart at the visual layer. the content might be structured well enough, but the output ends up looking like every other ai-generated deck, which undermines the whole point if you're trying to make something that reads as professional.

generateppt's 'anti bloat' framing suggests they've approached this differently, by constraining the design surface instead of expanding it. pre-decided layouts and a curated slide library means the tool makes the visual choices that most people waste time on anyway. the trade-off (no drag-anywhere editing) is exactly the right one if the goal is a clean, fast output rather than a sandbox.

If the platform can genuinely manage the context of user-dumped materials well, I think this scratches the perfect spot I got what I want to put in a PPT, but it looks bad when I do it myself. I've been working as a freelance marketer for couple months, and just dumping all the contents to AI makes horrible PPT.

One issue I found using it - when uploading a file, it says it will only accept PDF, DOCX, TXT, I actually can upload another forms of file. Although it seems to be not processed, it gets messy. Maybe you could just block a file if type is inappropriate.

Also, as a marketer, most of my reference materials are in PPTX, JPG, or PNG. It'd be useful if these types are also supported.

Is this one version also 18+? :D For sure, asking before public embarrassment.

My cat jumped onto the table, the laptop farted and now the cat is traumatized for life. Are u covering my vet therapy bills?😭🤣

Few days back only I upvoted slapmac as the idea was fun and asked for windows version! Thanks Man!

This is so unnecessary… I love it 😂

I knew it was coming after I saw slapMac, lol!

It's SETI at Home but for inference! 🤩

the auto-configured p2p setup is the clever bit. most self-hosted inference solutions require you to manually manage which node is running what model, handle routing yourself, and accept that you'll be ssh-ing into machines whenever something needs updating. auto-configuring the mesh and exposing a standard openai-compatible endpoint means your existing agent tooling just works without a custom client.

where this gets hard is reliability under real agent workflows. spare capacity is honest framing, but spare capacity is also the most volatile kind. when an agent mid-task makes a follow-up call and the node has dropped or a worker left the mesh, the retry behavior matters a lot. handling partial failures gracefully without surfacing errors to whatever client is consuming the api is a non-trivial coordination problem, especially as the mesh grows beyond a few trusted machines.

Hey PH! I'm Jake Stein, co-founder and CEO of Common Paper.

We built Gerri because of a pattern we kept hearing from our users. Contracts were slowing deals down, and when we dug into why, the answer was almost always the same: redlines.

What surprised us was how repetitive the problem was. Most companies were seeing an 80/20 situation where a handful of the same issues came up on contract after contract. They wanted to loop in their attorneys for the really thorny legal issues, but not to answer the same questions over and over again. Also, many of the contract questions aren't legal at all, but instead need to be answered by a different member of the team.

What was missing was a way to decide once, and then have the system handle it automatically from that point forward. That's the core of how Gerri works: you build a playbook that reflects how your company actually negotiates, and Gerri's AI applies it to every contract that comes in. Accepts what you'd accept. Pushes back on what you'd push back on. And escalates the things that genuinely need a human, routing each issue to the right person (legal, sales, finance, etc.), then generating a clean file that's either ready to sign or send back to the other side.

The result is that 90% of contracts get reviewed in under 3 minutes, often before anyone on your team has realized that the contract arrived.

Happy to answer any questions about how the playbook works, how teams are using it, or anything else. And if you want to see it in action, there's a short demo linked above.

First 3 contracts are free at getgerri.ai

I'd love to hear how you're handling redlines today and what you wish existed

does a company need a documented playbook before gerri can do anything, or does it help you build one from scratch?

based on the description, it sounds like the playbook comes first, since gerri "applies" it rather than creates it. that actually makes sense for later-stage teams where contract positions are already established but stuck in someone's head. the real unlock isn't the 3-minute review time, it's forcing cross-functional alignment on what's actually acceptable upfront, which is a conversation most companies avoid until a deal is already in trouble. :)

Congrats on the launch! Let’s go!

This could be a huge unlock for deal flow. The biggest hurdle I see is getting the non-legal teams to trust the automation. What does the onboarding look like to get a company's playbook into the system and build that initial confidence with the sales and finance folks?

Hey Product Hunt! 👋

Turn It Gen Z translates your boring text into Gen Z slang instantly -
pick a vibe, hit convert, share to X or anywhere you want.

I built this because I kept seeing marketers and brands try to sound
young and failing spectacularly. The slang was wrong, the energy was
off, it was giving very "fellow kids." So I made an AI that actually
gets it.

There are 7 modes: brainrot, sigma, TikTok, Twitter, corporate, soft,
and an "explain" mode that translates AND glosses every term so you
actually learn the slang. Intensity slider goes from barely-there to
maximum chaos.

3 free translations, no signup. Rizz Pass unlocks unlimited + slang
invention mode (it generates slang that doesn't exist yet - it's unhinged).

Use PRODUCTHUNT80 for 80% off on Rizz Pass.

Would love to know: which mode do you try first? And what words are you
already using that are apparently ancient? 😭

Needing AI to understand some slang feels like the "How do u do fellow kids" meme, haha. Seems like Im officialy old rn)

This is actually pretty fun 😂

Can see this being useful for social posts.

Interesting idea!

As a middle-aged person, I sometimes really struggle to understand Gen Z slang, especially when I'm working on products aimed at that group. It honestly feels like I need a translator.

This is interesting! also the dictionary is quite helpful for someone who wants to understand meaning of these words which they must have heard.

does natural language search actually handle descriptions like 'fintech operator who's exited once and is now investing in b2b saas' differently from keyword search, or is the 'describe your ideal lead' interface just a more polished way to run the same structured filters?

from what i can read, the differentiation is in the enrichment layer. 1b+ profiles with web data layered on top of linkedin means the matching isn't limited to what's in someone's headline or job titles. if the enrichment actually captures context that structured fields miss, like conference speaking history, writing, or other signals, then natural language queries start to mean something qualitatively different. for use cases where the right person is hard to find by title alone, that's where this earns its pitch.

Quite useful! But if lead discovery becomes trivial with tools like this, do you think the real bottleneck shifts to outreach quality?

This is going to be really helpful for HR's at our company, just a quick question how do we make a bulk search and bulk export?

the physical reaction angle is the most interesting bit. almost all developer tooling assumes your only feedback channel is the screen. having something in your physical space that reacts when claude runs or fails creates an ambient signal you can notice with peripheral attention instead of actively watching a terminal. it's a different mode of awareness entirely.

the 'opinionated' framing is doing more work than it looks. classic rubber duck debugging is you explaining the problem to something that doesn't respond. adding opinions, especially on your variable names and decisions, shifts it closer to a passive code reviewer than a debugging aid. that's a different value proposition and probably a different use case.

The physical reaction part makes it unique, I was wondering does it have specific physical reactions when you ignore it's suggestions.

@Slide2Video @tiga_liang can you add a video here in Product Hunt showing your slide to video working?

Interesting — this solves the creation side, while we're solving the extraction side. At NexClip AI, educators import long lecture recordings and pull out topic-based clips. Slides → video and video → clips feel like natural complements for the education workflow.

So this will add relevant voiceover on the existing slides right or we have something above that?

I need to try ASR and it's perfect for me. Thanks Zac for hunting it! Feel I gonna love it

I run Whisper in prod for a voice input thing — accents and background noise break it constantly. If this actually handles noisy multilingual audio better, that alone is worth switching. $0.36/hr is solid too. Gonna try it this week.

Multilingual ASR is a hard problem — especially for noisy audio. We deal with this at NexClip AI too, where accurate timestamps on every word are critical for topic-based video editing. Curious how MAI-Transcribe-1 handles word-level timestamp accuracy across languages?

Congrats on the launch! 👏

Also launching today — curious, what worked best for you to get your first users?