PH热榜 | 2026-04-13

Hey PH!

We launched Accent Conversion for meetings back in March. People loved it, but we kept hearing the same request: "Can I use this on YouTube?"

So we went and looked. Pick any popular lecture by a non-native English speaker and scroll to the comments. It's always the same: "Can't understand what he's saying." "Had to watch at 0.75x." "Captions are completely wrong."

Millions of views. Brilliant speakers. Huge chunk of the audience rewinding or just leaving.

That felt wrong. The content is right there. The only thing in the way is accent clarity. So we packaged our on-device AI into a free Chrome extension.

Open a YouTube video, toggle it on, speech gets clearer. That's it.

Ask me anything, I'm here all day.

Great job Krisp team! This is a life-saver 🤌

btw, what accents does it support?

Most tools focus on captions, but real-time accent normalization in audio addresses a different problem! Curious how it handles heavily accented technical content like coding tutorials or medical lectures where vocabulary matters as much as clarity. Congrats on the launch, team Krisp!

So freaking useful! Do you see accent conversion becoming as standard as captions or playback speed on video platforms or maybe youtube will add this feature itself??

We built it for builders like us 🎉 Everyone I know who watches Vibe coding tutorials needs to try this.

Looks Neat! Can we take it beyond meeting rooms ? Can we use this for voice over for my content, with natural emotion ?

We are Curious how natutal the converted voice sounds across different accents. does it preserve the speaker’s tone or lean more neutral?

Just tried and it's amazing! Can't believe Youtube couldn't come up with this feature all these years!

Just installed it. Tried it on an NPTEL lecture and before/after is impressive

intresting approach, but I hope it doesn’t distort the speaker’s tone too much. That natural flow is still important for learning.

Cool !! I usually spend half of my time on technical tutorials trying to decipher specific terms that get lost in translation or heavy accents.

Does this handle real time processing well enough ? there is no weird audio lag with the video ? if the sync is tight , this is very useful .

This feels like when captions first became standard on YouTube. Once you have it you wonder how you watched without it

Does it affect the voice (changes it, or maybe makes it robotic)?

Heheh! Need this. As a non-native english speaker we constantly deal with it and I'm sure many founders gonna feel the same Asti! Wish you all the best here

I can finally understand my CS lectures

Now I can go back to my resolution to learn Python, since I will actually understand the content. :)

You know, 40 seconds of video and everything is clear. The main feature is clear!)

love the idea. very useful for native language speakers. what's the accuracy of these models, the cost of a wrong accent can be significant in some use cases like critical meetings.

I have struggled in meetings, where people don't have good microphones or are sitting away from the laptop or a lot of background noise, If Krisp can support that as well it will be awesome.

But that will be real time.

Finally. I spend half my day on YouTube watching technical documentation and deep-dives, and it's a constant struggle when the auto-captions can't parse technical jargon because of a thick accent. I usually end up wasting time rewinding or just giving up on the video entirely.

Seeing this as an on-device Chrome extension is interesting from a performance standpoint. I'm curious about the browser overhead—have you guys noticed any significant impact on CPU or RAM usage during longer 30+ minute lectures?

This is a genuine friction point for the global dev community. Great to see a practical use case for on-device AI that isn't just another chatbot. Good luck with the launch!

How is this different from auto-captions? Is it actually changing the audio?

This is one of those apps you only appreciate once you’ve tried it in a noisy environment. Curious how it performs with more complex background noise like cafés or street traffic.

This is a huge deal for educational content. I teach an Excel for Financial Modelling course on Udemy (https://www.udemy.com/course/excel-for-financial-modelling/) and a big chunk of my students are non-native English speakers working in finance globally. Accent barriers in video-based learning are real, and on-device AI that solves this without requiring the creator to re-record is brilliant. Curious if you're seeing higher retention rates on videos where accent conversion is active?

This is what I can genuinely useful, practical leverage of AI, very nice. I have one question and one suggestion.

• Question: does it work well also when people are talking on another or only for solo speakers?

• Suggestion: Make also a funny/gimicky version where everyone can switch their voices to anything they like - I always wanted to sound like British royalty :)

YouTube has had speed control since forever but accent was always the missing piece. I've rewound the same sentence four times trying to catch a word. One toggle sounds right - this doesn't need to be complicated.

Luma Agents is a creative agent platform that plans, generates, and iterates across video, image, and audio within a single shared workflow.

I'm hunting this because the gap it's addressing is structural, not just a feature gap most AI creative tools are isolated generators, not pipelines.

The problem: Creative teams at agencies and studios are stuck stitching outputs together across tools. Every handoff is a restart. Context gets lost. Scale means adding headcount.

The solution: Luma Agents embeds context across every stage of a project concept to delivery. Agents see what you see, carry that context through video, image, and audio generation, and iterate without you re-explaining from scratch.

What you can do with it:

🎬 Run a full brand campaign with cohesive visuals and variants across formats

📦 Generate e-commerce product shots lifestyle, hero, on-model in one workflow

📱 Produce short-form video ads with hooks, captions, and platform-specific framing

🌍 Localize videos with natural voiceovers and synced visuals across languages

Who it's for: Creative directors, brand teams, and agency producers who are already using AI tools but spending too much time wrangling outputs between them. Also relevant for solo creators who want to produce at team-level volume.

Time will tell whether teams are going to use this to replace specific tools in the stack or layering it on top of what they already have.

P.S. I hunt the latest and greatest launches in tech, SaaS and AI, follow to be notified → @rohanrecommends

The "shared context end-to-end" angle is what makes this genuinely interesting to me.

Most AI creative tools treat each output as a fresh start — image, then video, then copy, all disconnected. The result is creative that looks like a ransom note: technically competent, visually inconsistent.

Building ad-vertly.ai, we obsessed over this same problem from the advertising side. A campaign should have a through-line: same brand voice, same visual DNA, same audience understanding — whether you're running a static banner or a 15-second video. The moment you break context, the creative stops feeling like a brand and starts feeling like a vendor.

The market that'll love this first: performance marketing teams at agencies where speed-to-creative is the bottleneck. The question is whether the iteration loop is tight enough to replace the current "generate, export, feedback, regenerate" cycle that kills time.

Really excited to see where this goes. Congrats on the launch!

Is this release from this company? https://luma.com/

Works best for product teams with repeatable assets, consistent brief and clear output. I am not sure this works for service-based business. The pipeline thinking is solid either way.

How much control do teams have over each step when the agent starts refining outputs on its own?

The "shared context" piece is what kills every creative workflow I've seen. You finish the video, send it to copy, they produce something totally different in tone, then back to square one. Curious whether the context travels across handoffs between team members or just within a single session?

mostly layering first - teams dont pull out existing tools until the new one handles edge cases reliably. curious what a full replacement workflow looks like.

Really interesting direction. Feels like we’re moving from tools to actual creative collaborators. The shared context across assets is a big deal.

Would u consider giving a trial plan to test this. I find only paid subsciptions

quick question, does it flag both regulatory tracks or pick the stricter one?

Congratulations @naomie_halioua on your PH launch,
Is Cleo built more for brands already selling globally, or can a founder use it before* launching to decide which markets to enter first based on compliance complexity?

Hey Product Hunt! 👋

I'm Naomie, co-founder & CDO of Cleo Labs. I'm a Polytechnique engineer and AI researcher. My co-founder Anaëlle is a Sorbonne-trained lawyer who spent years running legal transformation at Havas Group. She knows the regulations. I build the AI that reads them.

Why we built Cleo:

If you sell a physical product internationally — a bike helmet, a lipstick, a washing machine — you're dealing with a regulatory nightmare that no one talks about. Materials restrictions, labeling rules, certifications, customs codes... and every single one changes from one country to the next. A single connected bike helmet triggers over 100 regulations. Multiply that by 106 markets.

95% of RegTech serves financial services. For physical goods, compliance teams still rely on consultants, spreadsheets, and prayer. We thought that was insane.

What Cleo does:

You type your website URL. Our multi-agent AI pipeline (we call her MARIA) identifies every product category you sell, maps every applicable regulation across 106 countries, and delivers a structured compliance map, with human-in-the-loop validation on every output. This isn't a chatbot giving you a guess. It's AI precision, verified by legal expertise.

We genuinely believe product compliance is the next massive RegTech vertical, and by 2030, every product sold in Europe will need a Digital Product Passport. The wave is here.

Would love your honest feedback. What would make this more useful for your team? 🙏

Still feels a bit surreal to finally share this 🚀

We’ve been working on MARIA for a while now, and seeing it live - scanning regulatory requirements across +100 countries in minutes - is something we genuinely needed ourselves. What used to take weeks of back-and-forth with legal/compliance teams can now happen almost instantly.

From day one, we knew this couldn’t be just “another AI tool.” Compliance is too critical. That’s why we built a human-in-the-loop verification layer - combining speed with the level of reliability this space demands.

Product compliance is a massive blocker for any brand going global, and we’ve felt that pain firsthand. Really proud of what the team has built, and this is just the beginning.

Big thanks to everyone who supported us along the way 🙏

compliance matrices across 50+ jurisdictions is one of the nastier unsolved problems in hardware. real issue worth automating.

This is a big shift in how compliance work gets done. Mapping regulations across so many countries in minutes feels like a real unlock for global teams.

will it be only useful for electronics ?

Congrats the launch! Cleo helps every physical product seller in global. Are you planning to make a function that shows regulation passed products? I thought example products make regulations be understood easier.

Anyone who's dealt with cross-border compliance for physical products knows the pain: one SKU can trigger 100+ regulations across materials, certifications, labeling, and customs — and they change constantly.

What stands out here is the multi-agent AI pipeline combined with legal expert verification. That hybrid approach (AI speed + human accuracy) is exactly what this space needs.

Really impressive scope with 106 countries covered.

Well done team! @Cleo Labs

AI has made the world a lot more convenient, but the standards set by governments around the world vary so much that when you try to do something outside of your computer, it's a completely different world (nothing has changed since the days before AI), so this is exactly the kind of feature I wanted! Thank you.

Cross-border compliance is one of those silent deal-killers in international transactions. In my work structuring renewable energy deals across multiple jurisdictions, regulatory mapping is always the most time-consuming part of due diligence. I build project finance and valuation templates for cross-border deals on Eloquens (https://www.eloquens.com/channel/samir-asadov-cfa) and the regulatory assumptions are always the hardest to standardize. Love that Cleo is automating this layer — would be curious to see how it handles regulatory divergence between EU and APAC markets.

11% active rate sounds about right. I have a similar pile of skills I keep adding "just in case" and never trigger. The dedup feature is the real value here - I definitely have overlapping skills that quietly fight each other for context space. Going to run this today

I've been dumping so many custom tools into my Claude Code config lately that half my context window is probably just unused skill descriptions. I am really curious if this parses the local Claude logs to calculate usage stats or if I need to run it alongside a proxy. Hooking this up to an automated weekly cleanup script would be a massive time saver.

solid util. though if you're surprised by which skills you actually use, the taxonomy was probably wrong to begin with.

Solid idea. My skills folder gets crazy real fast.

Man, I love this and need it. Really interesting idea.

finally. something everyone needed but no one had the courage to ask for :)

I have a folder of nothing but .md files CLAUDE.md, MEMORY.md, various agent instructions and opening any of them in Finder has been genuinely painful for two years. Installing this right now!

Hi @jollife !
App works great, I was looking for exactly this - I hate looking at md source ;).

bold positioning. most markdown apps lean into the 'it's just text' thing - this is the counter-argument.

What a great outcome from our first coding evening in years 🎉 You’ve added a lot of nice features to our initial minimal prototype – Well done!

Good idea in theory. I’ve seen a lot of markdown preview tools, but most fall apart once you hit larger repos or mixed content. Curious how this compares.

Sounded like my cup of team, so I tried to install with brew and
==> Fetching downloads for: johannesnagl/tap/showmd

✘ Cask showmd (1.0.1)

Error: Download failed on Cask 'showmd' with message: Download failed: https://github.com/johannesnagl/showmd/releases/download/v1.0.1/showmd-1.0.1.zip

curl: (56) The requested URL returned error: 504

Hey Product Hunt 👋

We built ContextPool because we kept hitting the same wall: every time started a new Claude Code or Cursor session, my agent had zero memory of what we'd already figured out together. Same bugs re-discovered. Same architectural decisions re-explained. Same gotchas re-learned.

It felt like working with a brilliant colleague who gets amnesia every morning.

So we built a persistent memory layer specifically for AI coding agents. Here's how it works:

1. Install with one curl command (30 seconds, single binary, no dependencies)
2. Run `cxp init` - it scans your past sessions and extracts engineering insights using an LLM
3. Your agent automatically loads relevant context via MCP at session start

What it remembers isn't conversation summaries - it's actionable engineering knowledge:
→ Bugs & root causes ("tokio panics on block_on in async context")
→ Fixes & solutions ("Use #[tokio::main] instead of manual Runtime::new()")
→ Design decisions ("Chose libsql over rusqlite for Turso compatibility")
→ Gotchas ("macOS keychain blocks in MCP subprocess context")

It works with Claude Code (zero config), Cursor, Windsurf, and Kiro. Local-first and privacy-first - raw transcripts never leave your machine, only extracted insights sync when you opt in.

The team memory feature is what we are most excited about: push insights to a shared pool, and everyone on the team pulls the collective knowledge. Your teammate debugged something last week? Your agent already knows.

Free and open source for local use. $7.99/mo for team sync.

We'd love to hear: what's the most frustrating thing you keep re-explaining to your AI coding agent? And if you try it - what insights does it extract from your sessions?

GitHub: https://github.com/syv-labs/cxp

Really cool! Btw how does Contextpool handles codebase evolution like when old decisions become invalid? Also how are you structuring extracted insights, are these embeddings, structured schemas,or something hybrid? And is all of it stored locally?

Vibe-coder here. I maintain a claude.md file and update it manually at the end of every session. It's manual, but it works. For a solo builder (no team) what does ContextPool give me that a well-maintained claude.md doesn't?

The brilliant colleague with amnesia framing is exactly how it feels, you spend half the session rebuilding context instead of actually building. The team memory angle is where this gets really interesting though.

Does it handle conflicts when two teamates have solved the same problem in completely different ways or does it just load both and let the agent decide?

This is solving a real problem. I've been building a full SaaS in Claude Code for the past year: 13 AI agents, FastAPI backend, Next.js frontend — and the context loss between sessions is genuinely the biggest friction point.

The thing I keep re-explaining: project architecture decisions. Why certain agents are split the way they are, why the credit system works a specific way, which database tables relate to what. Every new session I'm pasting the same CLAUDE.md context block to get the agent back up to speed.

Curious about one thing, how does it handle multi-stack projects? My repo has TypeScript frontend and Python backend with very different patterns and gotchas in each. Does it extract insights per-language/per-directory, or is it all one pool?

Going to try this today.

I don’t quite understand how you handle control and cleanup of memory from bad, incorrect, or outdated solutions. I often reset the context on purpose so the agent forgets everything and we can start from a clean slate — otherwise past mistakes can compound into even worse decisions over time. I’m really curious how this is managed in your approach.

What I've used so far that works very well for me is the compound part of Compound Engineering. The problem I see to CE is that it's per repo, ContextPool looks amazing since all my repos can share these eng learnings!

Great work!

This solves a genuine pain point. I run a small agency and every time I spin up a Claude Code session on a client project, I spend the first 10 minutes re-explaining the stack, the deployment quirks, and why we made certain architectural choices. The idea of capturing that as structured, searchable memory rather than just dumping everything into CLAUDE.md is a much cleaner approach. Curious about one thing: for the team sync at $7.99/mo, is there a way to scope shared memory per project or repo? In an agency setting, you definitely don't want client A's context leaking into client B's sessions.

Interesting concept, but “exhaustive scanning” sounds expensive at scale. Curious how it performs with large document sets in real production use.

Built something similar for a different layer persistent memory across business workflows, not just coding sessions. The "docs graveyard" concern from the comments is real. What helped us was making memory write-on-use, not write-on-save. If an agent references a piece of context during a task, that context gets reinforced. If nothing ever pulls it, it decays. Curious how you handle relevance scoring when the pool grows past a few thousand entries.

This is really cool. Does the agent have persistent memory on only your work or also the work your team is working on?

This is super interesting! I Wish I had discovered it earlier lol.
Anyways 6x increase sounds strong but what was the baseline and what kind of companies saw the biggest lift? Also have you seen any cases where AI actually hurts conversion maybe due to wrong tone or over automation etc

I've tried dozens of these tools, and Clarm is hands down the best

We had a lead sit on our docs site for 40 minutes reading everything and then just close the tab. Nothing. If something had just said "hey, want a quick demo?" at the right moment, that would've been a customer! +1 upvote!

This is a strong idea, especially for dev tools where questions repeat a lot. Curious how you keep answers accurate when docs or repos change quickly.

Nice product, can it work with Crisp?

What's the turnaround time to get this live and working? Does it require a hands-on detailed setup?

Interesting approach to inbound qualification. In financial services, the gap between "someone visited our site" and "this is a qualified deal opportunity" is massive. We built ModeLoop (https://modeloop.net/?i=1) to help finance teams with modeling and deal structuring, and lead qualification has always been the bottleneck — especially distinguishing serious institutional buyers from casual browsers. The AI intent detection across multiple channels (web, Slack, Discord) is a smart move. Does Clarm support custom qualification criteria specific to B2B financial services use cases?

Congrats on the launch! Chat interfaces always win!

Hi everyone!

VoxCPM2 is the next-generation open-source audio model from the @MiniCPM family, and it perfectly continues their signature trait of incredible "capability density" — packing all of these features into a model that is only 2B parameters!

Despite its highly compact size, the feature set it brings to the table is quite rare for an open-source release:

• Voice Design: Instead of hunting for the perfect reference audio to clone, you can just prompt the model directly (e.g., (A young woman, gentle and sweet voice) Hello world.). It generates a completely novel voice on the fly.

• Native 48kHz Output: It has a built-in super-resolution VAE, meaning no external upsamplers are needed to get studio-quality audio.

• Controllable Voice Cloning: You can clone a voice from a short clip, but still steer the emotion, pacing, and style using text prompts.

• Production-Ready: It hits an RTF of ~0.13 for real-time streaming and is fully open-source under the Apache-2.0 license.

It is incredibly refreshing to see this level of controllable, high-fidelity audio hit the open-source ecosystem in such a lightweight package.

Try it out here!

Voice design from text prompts instead of hunting for a reference clip is the thing I didn't know I needed. "A tired middle-aged man reading terms of service" and it just... makes that? 2B parameters for this is wild. Will try it locally today.

2B params delivering 48kHz + voice design + cloning is impressive capability density. As someone building an audio/video editing tool that relies on audio analysis for precise segment boundaries, I appreciate how much source quality matters.

Curious: how does VoxCPM2 handle multilingual switching within a single utterance — e.g. Japanese with embedded English terms?

Really nice idea, especially the real-time interference visualization. How do you handle more complex layouts like multi-floor homes or offices?

This is actually really slick.

I’ve run a decent number of site surveys, and having something lightweight like this in the browser is a big win compared to traditional tools.

How accurate have you found the attenuation modeling across different materials?

Congrats on the launch! The product concept looks solid.

Quick performance alert: I just ran a real-device test from Beijing, and the landing page takes 30s+ to become interactive. It seems like a major bottleneck with the API or font loading in this region.

I’ve got the full recording with console logs ready. Would you like me to send you the link to help your team debug?

Going to test it! Looks great!

Ok this is one of those tools I didn't know I needed until now. The realistic wall material attenuation is what makes this actually useful vs just guessing where to put access points. Does the 3-stage optimizer account for expected device density? Like a conference room with 20 people on video calls vs a hallway nobody sits in?

Really handy!! Have users discovered any issues in their setup that they wouldn’t have noticed otherwise??

I have this robo vacuum at my home and I also really don't understand why in some areas it doesn't get the Wi-Fi access. Having this loaded onto the system would really help me plan out those and also how to position my Wi-Fi. Having it open source is ready to add on. Thank you for this, Sean

Hey 👋
We’re Ashish and Pratik, founders of SigmaMind AI.

After watching developers jump between dashboards, docs tabs, and their IDE just to configure a single voice agent - we knew the problem wasn't the technology. It was the workflow.

So we built SigmaMind MCP server.

Open Cursor, Claude Code, or VS Code. Type what you want:

"Build a customer support voice agent. GPT-4o. ElevenLabs, calm British female. Agent speaks first. Extract sentiment and escalation flags after every call."

That exact spec deploys. No dashboard opened. No context switching.

Here's what you're actually controlling from that one prompt:
→ LLM Model - GPT-4o, Claude, Gemini, or your own
→ Voice & TTS - pick the exact voice experience
→ Conversation Flow - who speaks first, how it behaves
→ Welcome Message - define the opening line
→ Background Audio - optional, on-brand
→ Post-Call Insights - sentiment, intent, escalation

Every layer configurable. All from natural language.

And telephony is built in - buy numbers or bring your own, assign to agents instantly, run real calls not simulations.

Under the hood:
→ Sub-800ms latency
→ IVR and phone trees navigation
→ Built-in VAD (Voice activity detection)
→ Noise cancellation for noisy background environments
→ Model-agnostic (Deepgram, GPT, ElevenLabs, or your own stack)
→ Multimodal (voice, chat, email - one agent brain)
→ Parallel tool calling for real-world actions

Set up in under 5 minutes: https://docs.sigmamind.ai/mcp/se...

i can direct claude code to an existing open source github repo and build voice agent for me with VAD / sub 800ms latency etc ? I understand this may consume more tokens. How is this MCP diff ?

MCP for voice agents makes sense. wiring voice into an agent stack always gets messy - nice that this abstracts it.

This is so cool. Running a VOICE AI company this is really good stuff. Congrats on the launch

So proud of the team for getting the SigmaMind MCP Server live today!

I’ve seen how much effort went into making sure this wasn't just another 'cool tool' but a production-grade orchestration layer.

My favorite part? Being able to create and manage a Voice AI agent, all from a simple prompt, without leaving Cursor. It feels like magic every time.

We’re all hanging out here today to answer questions and get your feedback.

Hi Product Hunt! 👋 I’m Prince, the creator of Open Comet.

Most AI agents today feel 'trapped' behind a chat box, forcing you to copy-paste data back and forth. I built Open Comet to bridge that gap—an agent that actually lives where you work.

My goal was to combine the power of autonomous research with a strict privacy-first architecture. Whether you're a researcher needing deep link exploration or a developer automating repetitive web tasks, Open Comet is designed to stay out of your data and in your flow.

Excited to hear your feedback and answer any questions!

Lovely! Can this browse my logged in pages like linkedin ? And take actions on my behalf. Some UX flows especially linkedin is sub-optimal and this can help save me lot time. Thanks.

love that you went with local history storage. too many AI tools are black boxes with your data. curious about the STORM research loops - does it actually follow citation trails and cross-reference sources, or is it more like iterative searching?

Is this availabe on Linux ?

Autonomous AI research agents are where things are headed for anyone doing serious market analysis. I built PolyMind (https://polyminds.netlify.app/) to track large trades on prediction markets using AI-powered alerts, and the hardest part was always the research layer — synthesizing signals from dozens of sources into actionable insight. An autonomous browser agent that can handle deep research tasks across the web would be a game-changer for financial due diligence workflows. How does Open Comet handle source reliability and conflicting information across pages?

every autonomous browser agent I've tried hits the same wall eventually - persistent auth sessions and bot detection. curious where this one breaks.

Tried a few browser agents before and the scariest moment is always when it hovers over a "Submit Order" button. Does this one ask before doing something irreversible or do you just trust it and hope?

really appreciate the zero-data architecture and local history storage. most ai tools are just black boxes with your data so keeping it in the browser is a big plus for privacy. definitely going to give the chrome extension a spin today

Does this work only on Perplexity's Comet browser? Is there an affiliation with them as Open Comet's branding, name and logo is almost 100% on par with Perplexity's - most specifically, their Comet product.

autonomous is doing a lot of work here - the distinction that matters for real adoption is whether it blocks before acting or just does things and logs it.

I'm in Germany right now and this is such an issue here. Really hopeful that we can continue to digitize the bottom of the iceberg!

Hey Product Hunt 👋 I’m Aida, founder & CEO of Legitify.

Somewhere in the world, a deal is stuck right now…because someone is waiting on a stamp. A physical stamp. In 2026. Notarization and apostille are still shockingly manual - print it, sign it, scan it, find an office, wait days, repeat. It’s slow, expensive, and weirdly offline for something that blocks global business.

So we fixed it.

👉 Notarization + apostille, fully remote
👉 EU & UK notary infrastructure
👉 Valid across 50+ jurisdictions
👉 From request to certified docs in minutes/hours
👉 No unnecessary travel. No printers. No appointments. No “please come back tomorrow.”

If you work in legal, compliance, ops or anything cross-border, you already know: this stuff quietly kills momentum. We built Legitify so “waiting on notarization” stops being a sentence.

Already used by teams at leading enterprises and - and scaling fast 🚀

Hot take: paperwork shouldn’t be harder than the deal it enables.

I am here - AMA 🙌

Especially interested in the most ridiculous cross-border bottlenecks you’ve seen 👇

Nice initiative? Which countries do you support ? How do you keep yourself in sync with local policy changes ?

Harsh is first class shipper! This looks amazing!

Can you add this program to Windows, or is it basically the same for Mac OS?

Hey Product Hunt! I'm Paul-Antoine, co-founder of Vekta.

We're thrilled to share Vekta, an AI-powered training and coaching platform built to transform endurance sport.

Endurance sport has never been more data-rich. But data without the right system behind it just creates noise. Vekta connects planning, analysis, and insight into one integrated performance system so coaches and athletes can focus on what actually matters: performance.

What makes Vekta different?

• Moving beyond FTP: Vekta goes beyond FTP with Critical Power and W' modelling, giving the most advanced and accurate measure of an athlete's current ability and true performance.

• Built for athletes and coaches: Performance is a partnership. Vekta connects coach and athlete inside the same performance model, aligning planning, execution, and feedback in one shared environment.

• Human expertise meets AI: AI surfaces patterns, detects change, and highlights what matters. Coaches and athletes apply judgment. Technology amplifies decision-making. It does not replace it.

• Powering the pro peloton: Trusted by WorldTour teams including Lidl-Trek, FDJ United-SUEZ, and Decathlon CMA CGM. Chris Froome, 4x Tour de France winner, joined as Chief Innovation Officer.

• Structure meets personalisation: Sessions, events, recovery data, and feedback in one coherent system. Build sessions in seconds. Every detail is connected to the bigger picture.
  
  Whether you're a dedicated athlete, a coach, or a WorldTour rider, Vekta gives you a precise, evolving understanding of how you train, adapt, and perform.
  
  We'd love your thoughts, questions, and feedback!
  
  Paul-Antoine Co-founder & CEO, Vekta

Super cool idea! Love that this connects with my Oura ring. Was training for a half marathon last year and became obsessed with stats.

This is cool. Turning slides into a JSON-driven system is a fresh take. Curious how much control users have over styling vs the agent deciding structure.

Do you think this is better than cluely?

Building an AI tool that actually stays out of the way during live interviews is a tough design challenge. I do a lot of user discovery calls and always struggle with breaking eye contact to check my notes. I am really curious if you are capturing the live audio through a virtual microphone driver or just hooking into system output.

Hey PH! 👋 I'm Mit, solo dev behind GhostDesk. Built this after bombing a few meetings where I knew the answer but froze under pressure. GhostDesk sits invisibly on your screen — captures your voice, reads the screen using ocr, and feeds you AI responses in real time, all without appearing in recordings or screen shares. V2 is a full rebuild: faster transcription, smarter AI routing, and a stealth engine that actually works. Would love brutal feedback. AMA!

i wonder how pricing scales with usage. That’s usually the deciding factor for me in the long run.

The mention of AI agents alongside chat is interesting real time infra is evolving beyond human messaging now.

scaling chat for livestreams is the final boss of dev. handle millions of concurrent users without the message order getting cooked... if ably chat actually solves this, it's a game changer. @Ably Realtime @faye_mcclenahan1

Realtime infra is one of those things where correctness > features every time.
At your scale, I’m more curious about message ordering guarantees under partition than the API surface itself.