OpenAI Hits Manus With NEW AI, but Manus Fires Back Instantly! (AI WARS 3)

OpenAI just dropped a major update with its new agent building tools packed with features designed to simplify creating advanced AI agents. The idea is that this new API will handle tool usage, chat and advanced reasoning in one neat package. They're also bundling a bunch of built in tools that used to be separate web search, file search, and even a computer use function that simulates mouse and keyboard actions. On the very same day, a Chinese startup called Manus AI fired back with something even bigger, a partnership with Alibaba's QEN team that's set to push their AI system even further. It's a clear sign that OpenAI is scrambling to stay ahead while Chinese startups are moving fast and hitting hard way OpenAI positions it. These improvements are a direct response to all the feedback they got from developers who tried to build more complex AI applications and ended up wrestling with prompt engineering or promoting homebrewed orchestration logic. So with the Responses API, you can basically say to your AI, hey, do a web search or find this snippet in my knowledge base and it'll do it. No separate calls or complicated bridging. If that sounds great, keep in mind that they're also deprecating the Assistance API in mid-2026, which suggests they're really betting on the new approach. It might sound bold, but from the outside looking in, it feels like they're trying to regain some momentum. They might be losing to other AI companies that are moving faster or offering more advanced solutions. We'll get into those Chinese startups in a second, but let me just mention the new agents SDK that OpenAI is also rolling out. It's supposed to simplify orchestrating multi step workflows for multiple agents. Picture a scenario where you have one agent for triage and another for, say, shopping assistance, and they can hand off tasks between them seamlessly. They've thrown in guardrails, basically safety checks, so an agent can't go rogue and do something weird or destructive if it decides to run a computer command again. On the surface it all sounds appealing like this is how you build next gen AI apps. But the real question is whether it's enough, especially in the face of competition coming from places like China. That brings us to the big story on the Chinese AI front. You've heard about how Deep Seek blew people's minds by introducing a chatbot that could go toe to toe with some of OpenAI's best models at a fraction of the cost. Then another Chinese AI startup called Monica, under Beijing Butterfly Effect technology, stepped onto the scene recently, claiming it built the first general AI agent in the world, Manus, something that can operate like a digital employee that just takes an end goal and runs with it, no handholding required. They even said it outperforms OpenAI's AI agent, Deep Research, which definitely caused a stir in AI circles. Manus AI has this vibe like they're hardly even flustered by what OpenAI is doing. They've been doing freebies on X, letting random people request tasks just to prove how autonomous their system can be. The only downside is that their site kept crashing from all the traffic, probably because it's invite only, and wasn't prepared for the flood of interest. Still, the hype was strong. Then, in a move that made everyone do a double take, Manus AI announced a strategic partnership with Alibaba's QN AI team. Talk about not wasting time. On the very same day OpenAI announced Responsys API, Manus made it known that they're teaming up with Alibaba to integrate QN's open source AI models into Manus architecture, making their product even stronger. Essentially, they're reinforcing their backend with Alibaba's resources so they can handle more traffic and bigger tasks without blinking. If you're not familiar with Kwen, it's Alibaba's family of AI models that's been in the news before. They famously responded to Deep seeks success by pushing out a model that supposedly outdid Deepseek's V3. And they did it in record time, even during a public holiday. So pairing with Manus AI could turn into a huge threat for all the players out there trying to stay relevant. Alibaba is such a massive force that if they decide to throw significant weight behind this alliance, we might see a big chunk of the market pivot to these Chinese solutions, especially in Asia, where Alibaba's footprint is enormous. Meanwhile, from OpenAI's perspective, it looks like they're trying to pull out all the stops, but the timing is interesting. This refresh happens right when talk about Chinese AI solutions is reaching a peak peak. It almost feels like OpenAI is rushing to release something more robust so they can say, hey, we're still ahead, but the question is, does this new wave of features actually move the needle or is it just a patch to buy some time? Now, let's be real, OpenAI still has major brand clout and deep pockets from Microsoft. However, lower costs and super aggressive R and D cycles in China mean that these scrappy startups like Manus and can spin up advanced tools rapidly. If you look at it from the outside, you see a real race forming. Chinese AI might dominate across Asia, while OpenAI remains a staple in the West. Of course that's too simplistic because developers can adopt whichever solution best fits their needs, but clearly Alibaba backing Manus is a big deal. Another angle here is the cost factor. OpenAI is charging around $30 per thousand queries for the GPT 4.0 such query model and dollar for the mini one. They also impose fees for file storage beyond the first gig and the computer use tool rings in at $3 per million input tokens and $12 per million output tokens, which can add up fast if your agent does a lot of tasks. That's not to say it's unreasonably priced for enterprise grade AI, but some Chinese startups have been pitching that they can match or exceed performance at a free fraction of the cost. If that proves viable at scale, it's a serious value proposition for businesses trying to keep an eye on their budgets. Now let's talk about how this all ties into real world usage. OpenAI highlights customers like Hebia, Navan, Unifi and more, each using these built in tools to create real time research assistance or automate repetitive tasks. Sure it's impressive, but it might not raise eyebrows like a brand new solution that claims to be fully autonomous. Especially when you can see demonstrations of Chinese AIs taking a broad task like analyzing real estate data, checking multiple listings, then spitting out a short list of best deals in seconds without needing step by step instructions. The sense is that these Chinese AI agents are approaching that digital employee dream more quickly, or at least more aggressively. OpenAI also warns that its computer use model gets only around 38.1% success on the OS world benchmark. That's not surprising. These are still early days and any AI that tries to replicate actual browsing or system actions is going to make mistakes. But it underscores that the system isn't exactly plug and play. Even they recommend human oversight. Meanwhile, the marketing from Manus or Deepseek is all about how they can handle tasks with minimal supervision. Yes, you'd still want some fallback if things go wrong, but if they can deliver that next level of autonomy, it's going to raise eyebrows everywhere. All in all, we're seeing a high stakes rivalry heat up as OpenAI, scrambling for an edge, unveils new APIs and tools while Chinese startups like Manus and Deepseek push forward at breakneck speed. It's a clear sign that AI has evolved from mere chatbots to autonomous agents that handle entire tasks with minimal guidance, connecting to tools, retrieving data, and more. If you're in the market For AI solutions, OpenAI's responses API and Agent's SDK might feel polished and convenient, but cost sensitive users could easily be lured by Manus or Deepsea. Ultimately, everyone's watching to see if OpenAI bounces back with something truly revolutionary, or if the momentum shifts overseas. Let me know in the comments which side you're rooting for, or if you think there's room for everyone to coexist. As always, thanks for watching and I'll catch you in the next one.