C_NCENTRATE
IN THIS ISSUE: Manus rundown, robot muscles, 'freedom cities', titanium hearts, VR prisons, DNA damage fixes, gravity batteries... +
/// Play this ↓
Curated with neuroscientists, ’Focus’ is the official C_NCENTRATE playlist that aids your encoding and recall. Play as is, or hit shuffle.
/// Bookmark these ↓
TNN (The New Normal) - The Future of…WEARABLE/SXSW AUSTIN REPORT, March 20, 22 Bishopsgate, 8-10am - Tickets > Non-member / TBD+ member
The TBD Conference - April 3 - all day - Tickets (pay-what-you-can)
TBD: The AI Salon - April 3 - late - Tickets
/// Skim this ↓
China’s National People’s Congress was overtaken by AI optimism, with DeepSeek igniting investor and regulatory excitement. Meanwhile, Foxconn reported AI server revenue surged to 26% of its Q4 earnings, catching up to its core consumer electronics business.
Meta moved to suppress Sarah Wynn-Williams’ tell-all book, with critics calling it an attempt to punish a defector. The company also rolled out the Passthrough Camera API for Quest 3, though apps using it can’t yet be published on the Horizon Store. In other ‘oh dear’ news, the anti-Musk virus seems to be spreading. The EU ramped up efforts to secure satellite access for Ukraine, negotiating with SES, Viasat, and others, as Starlink’s connectivity in front-line zones became unreliable.
YC’s Winter 2025 Demo Day showcased 160 startups, with multiple companies unveiling AI-powered agent tools to enhance existing AI systems. In fintech, Clearwater Analytics acquired Beacon for $560M and Bistro for $125M, while Worth raised $20M to expand SMB underwriting automation.
Microsoft launched Copilot for Gaming, integrating AI-powered assistance for Xbox users through its mobile app. Meanwhile, Google confirmed Gemini will fully replace Assistant on Android by 2025, except on older devices with limited RAM.
Apple’s AI struggles deepened, as executives admitted Siri delays were “ugly” while publicly promoting undelivered features. Separately, Apple’s battle over UK encryption continued, with a closed-door legal fight against government-mandated backdoors.
Meta, Amazon, Microsoft, OpenAI, and Nvidia were subpoenaed by GOP lawmakers, accused of using AI to censor speech under Biden administration pressure in a sweeping investigation. Meanwhile, the US Cyber Command’s leadership shake-up saw Trump nominate Sean Plankey to lead CISA, with Congress expected to approve quickly.
Klarna officially filed for a US IPO, seeking a valuation over $15B, while Block expanded consumer lending via Cash App Borrow, after obtaining FDIC approval.
Filed under ‘too little, too late’, TikTok rolled out new parental controls, automatically prompting under-16 users with a “wind down” reminder at 10 PM. Meanwhile, DeepSeek’s ties to Beijing were scrutinised, as reports emerged that its employees had to surrender their passports, with travel abroad tightly controlled.
/// Read these ↓
↑ These devices might just bring back the joy of getting handwritten letters. /6 mins
Wind and solar power have overtaken coal (natural gas still number one). /2 mins
A man just survived 100 days with a titanium heart. /6 mins
Tech execs are pushing the idea of ‘freedom cities’ that are run by corporations. /6 mins
↑ Scientists just discovered an RNA that repairs DNA damage. /10 mins
Robot 3D-printed muscles just took a big step forward. /5 mins
China is about to make batteries that run on gravity (bub-bye lithium?). /9 mins
↑ What are the ethics of strapping VR headsets on inmates in solitary confinement? /6 mins
Should AI have a seat on your board? /8 mins
MANUS AI - WORTH THE HYPE?
While it was DeepSeek getting the US’s tech broligarchy’s noses out of joint this week, the big news was China’s latest AI launch, Manus. Touted as a breakthrough in fully autonomous AI agents, Manus, unlike ChatGPT, Gemini, or Claude, doesn’t wait for prompts or instructions—’it’ acts. The agent processes information, initiates tasks, and delivers results independently. It’s an agent in the truest sense of the word, designed not just to assist but to replace human decision-making in specific workflows. Check out the promo video:
Manus’ architecture allows it to break complex workflows into sub-tasks, delegating them to specialised sub-agents that operate asynchronously. Unlike existing AI tools, which require constant user interaction, Manus is cloud-based, meaning it continues working even when the user is offline. The concept is ambitious: a self-directed AI that can conduct research, manage hiring processes, analyse financial data, or build software with minimal human intervention. That is, if you, or your company, trust Manus (see more below).
The early reality, however, is (as usual) more restrained. Initial users have said Manus is not quite the frictionless autonomous agent its marketing suggests. Slow execution, task loops, and occasional errors have cropped up, issues that plague all AI agents at this stage. The reliance on Western-developed models—Anthropic’s Claude 3.5 Sonnet and Alibaba’s Qwen—also raises questions about how much of its autonomy is genuinely homegrown. China is catching up in the AI world fast (and in a lot of ways could be said to be ahead), but Manus does not yet represent a fundamental departure from existing AI development approaches.
Despite the criticism and limitations, Manus is significant, and signals a shift in China’s AI strategy from focusing purely on model development to leading in autonomous agent deployment. The U.S. has been cautious in pushing full AI autonomy, largely due to regulatory concerns. In contrast, China’s approach allows for rapid iteration, potentially giving it an edge in the long run. Oh, wait, someone has already open sourced Manus. Never mind.
Joking aside, Western AI companies have been working on similar projects, from OpenAI’s Operator to Google’s emerging autonomous AI initiatives and China is undoubtedly working on other things. The key question is whether China’s willingness to take risks will allow it to outpace its competitors. For now, Manus is an important development, but it is not the revolution many are claiming. Not yet, at least.
SO WHAT?
The biggest unanswered question about Manus isn’t its capabilities—it’s control and accountability. A fully autonomous AI that makes decisions without direct human oversight introduces risks that go beyond the usual concerns about bias or misinformation. What happens when Manus makes a financial decision that costs millions, executes a flawed hiring process that introduces discrimination, or acts on bad data in a high-stakes scenario? Unlike existing AI models, where responsibility falls on the user providing input, Manus operates independently, making liability a grey area. Beyond this, there’s a lot of chatter about observed security vulnerabilities, lack of transparency, and potential for misuse. One of the big questions coming at a lot of businesses is how much control humans should retain over autonomous AI decision-making.
DO: Scrutinise the technology beyond the hype and marketing materials. Read the terms and conditions of Manus (and and AI tool). Early reports indicate it still struggles with execution, not every launch is a revolution, nor one that doesn’t come with strings attached. DON’T: Don’t underestimate the risks. Manus (and other agentic AI tools) are in an extremely grey regulatory environment that allows for rapid deployment without extensive safety checks.
JOIN TBD+ — the global intelligence collective that’s your gateway to a powerful network, job opportunities, exclusive events, a vast content repository, and more.
Explore what’s next: members.tbdpl.us.