Hi
Hope you had great weekends. Things are slowly warming up in Hanoi - but the haze over the city is a bit of a drag. If you have any spare sun, please send it our way 🙏
Anyway, enough grizzling - a few headlines from the world of AI for your Mondays:
AI Agents Rising: OpenAI's Elite Tiers vs. Manus' 24/7 Workforce | Agent Economy 🤖
OpenAI are coming for white collar jobs? According to a new report from The Information, they are planning to launch several specialised “agent” products ranging from a $20k monthly tier focused on supporting "PhD-level research”, to a software developer agent ($10k/month), to a “high-income knowledge worker” assistant at $2k/month.
These models will reportedly leverage OpenAI’s powerful o1 and o3 models in combination with chain-of-thought reasoning to simulate doctoral-level expertise. Wild stuff - but then, in what is being called a new Deepseek moment, we got Manus.
Plugged as the "world's first truly General AI agent", Manus reportedly tackles complex tasks through a multi-agent structure. The system - which handles everything from planning, execution, to verification independently and can apparently run 50+ tasks at once - looks impressive at everything from resume screening to complex data analysis. Invites-only at the moment, apparently these are being resold at premium prices due to overwhelming demand but they're speaking to parts of this being open source (read: free) downstream. The speed of development is insane and it might just be that the ceiling is very high indeed…
The Two-Sided Coin of AGI: Urgency Meets Reality | Intelligence Evolution 🧠
Hard not to sound hyperbolic on this one… but the warnings about Artificial General Intelligence (AGI) appear to be reaching critical mass. As Ezra Klein (NYT and ex-Vox) recently said ing his conversation with Ben Buchanan (top AI advisor in the Biden White House): "For the past couple of months I've been having this strange experience where person after person independent of each other from AI Labs from government has been coming to me and saying it's really about to happen... AI systems that are better than almost all humans at almost all tasks... they believe it's coming inside of 2 to 3 years".
This isn't hype from techbro enthusiasts but a sobering reassessment from those with unprecedented access to cutting-edge developments. When combined with Ilya Sutskever (ex-OpenAI Chief Scientist and Co-founder) 's insight that AGI will both improve itself and work on next-generation AGI, we're facing a potential intelligence explosion that could outpace our ability to adapt - and where the gap between public perception ("the funny robot can do monkey tricks!") and technical reality has never been wider or potentially more dangerous (see pic).
Maynard's Critical Disconnect in action... (Wait But Why/Urban, 2015)
Don’t believe me? Take a minute to have a chat (yes, speak) with Maya or Miles - two new conversational AIs from Sesame, a lab focussing on building lifelike AIs and think about how far we’ve come since ChatGPT was released in late 2023 - just two years ago!?! When senior White House advisors sound the alarm about a 2-3 year timeline for transformative AI capabilities, shouldn't that fundamentally change our approach to AI governance? The stakes aren't just economic disruption or job displacement—they potentially include the future direction of humanity itself. Ultimately, as we race forward, are we prepared for what comes next? ⏱️
The Rapidly Evolving World of AI Research: Beyond Algorithms to Ethics | Research Trends 🧠
The AAAI's landmark 2025 presidential panel report has just been released, offering an interesting roadmap through an AI research ecosystem undergoing dramatic transformation - expanding beyond traditional boundaries into ethics, sustainability, and social good while grappling with the growing tension between corporate and academic environments! The field is increasingly hardware-dependent and interdisciplinary, requiring collaboration with experts from psychology, philosophy, and economics to address complex socio-technical challenges. Interesting also to note that researchers are prioritising trustworthiness and cooperative frameworks while looking to address the gap between public perception and technical reality (see above on AGI).
Looking ahead, several pivotal frontiers look to be reshaping our AI future: the pursuit of reasoning with formal guarantees, the integration of LLMs into multi-agent systems (oh hi Manus ), and the growing connection between AI and scientific discovery. These developments raise profound questions about academia's evolving role, the geopolitical implications of AI advancement, as well as whether or not we're fostering enough diversity in research approaches. As embodied AI and sustainability initiatives gain momentum, we're witnessing not just technological evolution but a fundamental reimagining of how intelligence itself emerges and operates in our world.
AI Tutors at Auckland: When Psychology Trumps Technology | AI Backlash 🧠
Interesting pendulum swing from the The University of Auckland in New Zealand. Their recent introduction of AI tutors in Digital Marketing has sparked fierce student backlash, perfectly illustrating psychological barriers to AI adoption identified in a research paper in Nature. Students' raw reactions - "Complete bull****" and fears of "a robot telling me slop" - directly reflect De Freitas, Agarwal, and Haslam (2023)'s identified barriers: opacity (distrust of AI's sources), emotionlessness (doubting AI's teaching ability), rigidity (concerns about "incorrect information"), and autonomy threats ("Why am I paying the same amount?"). 🎯
While Auckland's administrators defend the change as preparing students for "the evolving job market," it looks they've not locked in crucial psychological interventions the research recommends: providing transparency about how the AI works, demonstrating learning capabilities, and ensuring meaningful human oversight beyond weekly tutorials. As institutions rush to implement AI teaching tools, Auckland's experience offers a critical warning: ignore the psychology of technology acceptance at your peril. Technical capability alone doesn't guarantee successful adoption when human psychology pushes back. ✨
Bridging Nostalgia and Innovation: The Animated Rebirth of Los Angeles | Digital Arts 🎬
THIS is what AI in art should be about - enhancing creative expression not replacing it! The music video for Cuco’s "A Love Letter to Los Angeles" is a fantastic fusion of traditional 2D animation techniques, 90s style graphic art, and cutting-edge AI technology.
This awesome project that brought together 20 talented creators - including lead artist Paul Flores, global 2D animators, 3D specialists, and AI animation experts to craft a seamless visual journey that feels both familiar and brand new (excellent discussion of the workflow here). By prioritising artistic vision over technological novelty, the project succeeds in creating something timeless that reaches beyond the technology that helped bring it to life. ✨😍
The landscape has rapidly diversified beyond the handful of dominant players we saw in 2023-2024, with specialised AI agents from OpenAI and Manus potentially redefining knowledge work, while White House advisors issue sobering warnings about AGI timelines compressing to just 2-3 years. The AAAI presidential panel captures this perfectly—highlighting how research priorities now span ethics, sustainability, and multi-agent systems while the tension between corporate and academic environments grows more pronounced. Meanwhile, Auckland's AI tutor experiment reminds us that psychological barriers remain as significant as technical ones when integrating these technologies into education. These are truly extraordinary times where the pace of innovation makes Moore's Law look slow and where the questions about alignment, governance, and educational transformation have never been more urgent.
See you next week for more updates from this fast-moving frontier! ✨👋