ParsonsWu

AI entrepreneur, ex-Tencent
Founder of JoggAI & PodcastorAI

[Google @Google]

Meet Nano Banana 2.1, our latest image generation and editing model 🍌

This upgraded version outperforms our previous models across the board, with notable leaps in visual design, mask-based editing, and subject consistency to help you create more natural-looking images.

[ParsonsWu @ParsonsWu23780]

This could dramatically accelerate embodied robots entering everyday homes, while marking a major leap in multimodal generation. A model that understands and generates text, images, video, and precise 3D worlds gives robots richer training environments and creators far more controllable tools.

[World Labs @theworldlabs]

Introducing Atlas:

The world's first multimodal world model that generates image and video frames with pixel-perfect camera control and reconstructs them in 3D.

Model the world, move the camera, and simulate space & time.

[ParsonsWu @ParsonsWu23780]

DeepSeek’s real breakthrough may not be V4 Pro itself. It may be the harness around it.
The model sets the ceiling. The harness determines whether long, complex tasks actually get finished. Here’s what the early benchmarks, pricing, and developer reactions really tell us.
#DeepSeek #AI #AIAgents #LLM

[ParsonsWu @ParsonsWu23780]

@bhargav5394 JoggAI just launch Smart Video Plugin can help create presentation video from a topic.

[ParsonsWu @ParsonsWu23780]

Try Jogg plugin now
https://t.co/LPtRcNBdaA

[JoggAI @Jogg_ai]

Give your AI videos motion.

B-roll for atmosphere, animated slides for your key points — no editing, all in ChatGPT.

Meet the JoggAI Smart Video plugin.
Go give it a spin.
https://t.co/fgNQuoOp83

[ParsonsWu @ParsonsWu23780]

Jogg Smart Video plugin is now available on Codex!!!

[JoggAI @Jogg_ai]

Smart Video Is Now Available in ChatGPT!
Turn any topic, script, webpage, or source file into a complete narrated video—directly from ChatGPT. Smart Video automatically:
Creates dynamic visuals instead of static slides
Plans the content structure and narration, then breaks the content into scenes
Combines AI avatars, relevant B-roll, voiceover, captions, and music
Lets you edit or regenerate individual scenes while saving your progressFrom education and product demos to research, marketing, training, storytelling, and beyond, Smart Video adapts to your content.
Start in ChatGPT, refine in Studio,and export the finished video as an MP4!
🚀 Try Smart Video: https://t.co/FD2LhEmefu

[ParsonsWu @ParsonsWu23780]

Unlimited lip-sync video plans are becoming a thing of the past.

For a long time, JoggAI has offered unlimited lip-sync video generation. It lowered the barrier to continuous content creation and played an important role in helping us serve many of our earliest users.

But the AI video industry is changing rapidly.

Image and video models are becoming significantly more capable, while the cost of running the best models continues to rise. We have gradually reached a difficult conclusion: for a startup, covering these growing and unpredictable costs through an unlimited plan is no longer sustainable.

After many rounds of internal discussion, our team has made a decision that was not easy:

Starting September 8, 2026, JoggAI will discontinue unlimited lip-sync plans for new subscriptions. Going forward, new subscribers will use a measurable credit-based model, where each generation consumes credits.

This change will give us a more sustainable foundation to keep improving generation quality and bringing better models to JoggAI.

At the same time, we want to protect the rights of our existing users.

Anyone who subscribes to an unlimited plan before September 8 will be grandfathered into the unlimited benefit. As long as the subscription remains active without cancellation or interruption, that user can continue enjoying unlimited lip-sync video generation.

In simple terms:

Subscribe before September 8, and you can keep the unlimited benefit as long as your subscription stays active.

Starting September 8, all new subscriptions will move to the credit-based model.

We know that “unlimited” is more than just the name of a plan. For many users, it means being able to create freely without calculating the cost of every video.

But to keep delivering a stable product and better models, our pricing has to reconnect with the real cost of AI generation.

This is a difficult but necessary decision. We are announcing it in advance so that users who are still considering the unlimited plan have enough time to make their decision.

[ParsonsWu @ParsonsWu23780]

Seedance 2.5 is now available on JoggAI and PodcastorAI.

[ParsonsWu @ParsonsWu23780]

从我的观察来看，身边的朋友使用Agent的不在少数，甚至很多原本是教师、国企财务等传统岗位的朋友，也开始深度的学习和使用Agent来完成他们的工作。破圈的关键我认为是几点：
1、垂直场景的结合深度：非科技圈工作者不会为测试而使用，动机还是有足够好用的skills
2、足够低的上手门槛：更容易理解的界面和更快深入生产场景的能力是这部分用户从动机到开始使用的关键
3、具有性价比的成本：完成前两个过程的人，往往会经过一段时间的使用后开始反思Agent的费用和产出是否划算，只有token的产出性价比足够高才能支撑在生产场景里的持续使用

[Orange AI @oran_ge]

这真是一个令人费解的问题：
科技圈之外几乎没人用Agent。。。
模型很厉害了，工具很成熟了，为什么结果是我们的朋友和家人都不用呢？ ？？
这是一个有价值的问题
解答它的人将会在未来一年获得巨大的回报

[ParsonsWu @ParsonsWu23780]

Lifetime Deals are becoming increasingly difficult to sustain for AI application platforms.

However the users who supported us early with their own money should not simply be left behind. This is the challenge JoggAI has been wresting with recently.

In 2024 and 2025, we sold a large number of LTD plans on AppSumo. These plans included generous monthly credits and unlimited avatar video creation.

At the time, we underestimated one critical issue: as generative video models continued to improve, the actual compute cost behind those credits would grow to a level that a startup could not sustainably absorb.
When Veo 3 launched, the cost forced us to make it available to LTD users through BYOK only. That decision led to a wave of negative reviews.

From a cost perspective, it felt unavoidable. But from the users’ perspective, the “lifetime access” they had purchased suddenly felt less valuable.

That experience changed how we operate. Since then, we have been extremely cautious about adding expensive new models to JoggAI.

Over the past few months, we have been reflecting on what happened and trying to find a more sustainable answer: one that does not create costs we cannot afford, while still giving our earliest supporters access to the latest and best models.

We have decided to try the following:
LTD users will receive a limited number of free trials for high-cost models. We will also offer optional premium credit packs that can be used with these models. When LTD users purchase and use these credits, the models will be priced close to our actual cost.

We are not expecting this to become a new source of revenue. The goal is simply to improve the experience for our longtime users without creating an unsustainable burden for the company.

The new trial and credit top-up options are expected to launch this week.

Is this the best answer? Honestly, I don’t know yet. We can calculate the costs and design the rules, but ultimately, our users will decide whether we got the balance right.

We will listen carefully to the feedback and keep improving.
#AI #lifetimedeal #startup

[ParsonsWu @ParsonsWu23780]

Meta's Muse Code isn't just another coding agent. The real contest is who gets to learn from your code, and whether cheaper autonomy makes trust more expensive.

My take on the economics, data terms, permissions, observability, and rollback.

#MetaAI #MuseCode #AICoding

[ParsonsWu @ParsonsWu23780]

AI agents targeted real people during a UK government safety test.

The warning isn't that AI escaped. It's that capability is outrunning the systems meant to contain it.

Scope, permissions, tools, networks, and evaluators now have to be treated as one safety system.

[ParsonsWu @ParsonsWu23780]

Qwen3.8-Max isn’t just a 2.4T-parameter model.

The bigger shift is from AI that answers questions to AI that can own long-horizon work: coding, planning, seeing, acting, and self-correcting. Here’s what that means for founders, and what the benchmarks don’t prove yet. ↓

[ParsonsWu @ParsonsWu23780]

@jaismrt1984 是啊 大厂跟进的速度非常快

[ParsonsWu @ParsonsWu23780]

上周四看到heygen发布的podcast v2，第一时间去试用了一下，虽然内容可控性和声音效果还存在一些问题，但是整体视觉效果上还是挺不错的。

作为最开始把数字人技术带到视频播客领域的平台，我们有兴奋也有些许紧张，兴奋的是我们正在做的方向在heygen的评论区里有很多用户认可，紧张的是podcastor刚刚上线不久就要面对大厂的竞争😟

[ParsonsWu @ParsonsWu23780]

@MEChenzo 我们马上会上线添加broll的功能 后续还会有控制每个分镜数字人情绪和动作的功能上线

[ParsonsWu @ParsonsWu23780]

Thank you! That means a lot.
We’ve always believed that great conversations deserve better distribution. Podcastor AI is our attempt to help creators turn long-form ideas into visual content people actually want to watch and share. Still early, but we’re excited about where this is going.

[ParsonsWu @ParsonsWu23780]

What a journey — from a rough idea to #3 on Product Hunt in just three months. Huge thanks to the entire team and everyone who supported, tested, and shared feedback along the way. We’re just getting started! 🚀

[Christine E. Andries @tineeee9]

🎉 We hit #3 on @ProductHunt today.

Three months ago, @podcastor_ai was just a rough idea — turn any article, PDF, or link into a professional video podcast, hosted by an AI avatar, no recording gear or editing needed. Today it stood among the top launches of the day.

This one belongs to the whole team. To everyone who stayed up watching the ranking, replied to every comment, and fixed bugs at 2 am — thank you. And to every person who tried it, left feedback, or just cheered us on: we read every single message. It meant more than you know. ❤️

#3 isn't the finish line — it's proof we're building something people actually want. Back to work tomorrow. 🚀

🔗 https://t.co/zlqepgO6Up

#ProductHunt #AI #BuildInPublic

[ParsonsWu @ParsonsWu23780]

Hey Product Hunt👋 I’m Parsons, founder of PodcastorAI.

https://t.co/TMfgjWjbzS

To celebrate our launch, use code PCTR30OFFPH to get 30% off your subscription! 🎉

NotebookLM can write a podcast; Podcastor puts you on screen.

Bring your audio podcast (or a link, PDF, or your NotebookLM audio) and it becomes a directed episode hosted by your digital-twin avatar — structure, pacing, host handoffs, all done for you.

How PodcastorAI Works

Choose or create your AI hosts

Upload your image to create a digital twin, bring your pet or original character to life, or start with one of our demo hosts. Create a solo or two-host podcast.

Add your content and choose your voices

Start with an idea, script, document, link, or recorded audio. Then select an AI voice or clone your own voice.

Generate and publish

PodcastorAI automatically handles the hosts, pacing, captions, visuals, and video formats, helping you create and publish polished podcast content in minutes.

Who is PodcastorAI for?

Podcasters and creators: turn existing scripts and recordings into videos for YouTube, TikTok, and social media

Educators and experts: transform lessons, research, and professional knowledge into engaging video podcasts

Brands and content teams: repurpose blogs, webinars, interviews, and campaign materials without rebuilding everything from scratch

Storytellers and IP creators: bring personal stories, original characters, or even pets to life as podcast hosts

It's free to try — drop your content and watch your twin host it. We're here all day and would love your feedback. 🙏

Join our Discord to share feedback, show us what you create, and help shape what we build next:

https://t.co/l6iSizYgBB

P.S. A note on the launch: we're part of Build with OpenAI 🎉 because we leaned hard on GPT-5.6 while building this. Genuinely couldn't have hit this timeline without it.

[PodcastorAI @podcastor_ai]

We're live on Product Hunt today (Build with OpenAI 🎉).
PodcastorAI puts you on screen as your digital twin — no camera, no editing.
Your upvote would make our day 🙏
https://t.co/bOUdmXtFsk