A PNAS study found all 45 main-study participants improved after learning six whole-face qualities, with mean AI-face detection accuracy nearly doubling.
Read article
AI cheating in schools nearly tripled in a year in the UK. Detection tools are failing and creating false accusations. Here is what the evidence shows actually works.
Read article
A landmark clinical trial confirmed AI therapy chatbots can produce significant reductions in depression and anxiety symptoms. Here
left over from before. Worth clearing via Edit excerpt so it doesn't conflict with the actual meta description.
AI is unlocking 175 gigawatts of grid capacity the IEA says policy alone could never deliver — here's
AI tools now deliver over $172 billion in value to US consumers annually, and personal finance is where
Most people are still using AI wrong. Here's how to save time with AI across email, meetings, scheduling, writing, and planning, with no technical background required.
Read article
OpenAI, Anthropic, and SpaceX are heading to public markets within weeks of each other, chasing over $200 billion
ChatGPT, Gemini, and Claude have converged on similar pricing and capability in 2026 — this guide breaks down
OpenAI filed for a $852 billion IPO, then watched Anthropic top it and SpaceX's rocky debut spook the
The headline "AI will eliminate half of all jobs" predictions haven't materialised in the unemployment data — but
Hallucination rates range from 22 to 94 percent across the leading AI models, and the failure pattern is oddly predictable: accuracy collapses specifically when a false claim is framed as something you believe. Here's how
Read article
A 2026 academic study found AI genuinely can pick future stock winners — but only winners. Here's what
As the EU tightens AI surveillance rules and the US loosens them, a digital resistance movement is building
Google DeepMind's new Mixture of Recursions architecture cuts memory use in half, doubles inference speed, and outperforms Transformers,
Simpler AI models just proved they can match giants: DeepSeek matched GPT-4o at a fraction of the cost,
AI product design in 2026 has stopped being about flashy features and started being about earning trust after two years of users learning not to take AI outputs at face value. Here are the four
Read article
Figure, Tesla, Boston Dynamics, and 1X are all shipping humanoid robots in 2026 — but the gap between
The EU AI Act's full enforcement lands August 2, 2026, just as Meta's open-weight Llama models spread through
AI bias mitigation faces a hard new deadline: the EU AI Act requires documentation from August 2026. Here's
AI sycophancy is structural, not a bug: a 2026 Science study found AI affirms user actions 49% more
MIT research shows 7-year-olds attribute real feelings to AI agents. The APA warns AI companions may displace healthy relationship development. Here is what AI is actually doing to a generation growing up with it.
Read article
Generative AI now makes individual writers measurably better, a Science Advances study found, but when everyone uses the
Compression theory holds that curiosity, beauty, and intelligence are the same process viewed from different angles. A 2026
Google DeepMind says open-ended AI is essential for superintelligence, but no current system has it. Here's what the
A randomised controlled trial found AI coding tools made experienced developers 19% slower, not faster as they expected.
A randomised trial proved AI tools made developers 19% slower, yet 69% kept using them anyway. Cognitive ease, not productivity, explains why: the work felt easier even when it took longer. Here's the psychology behind
Read article
Disney's Stuntronics robots now perform aerial stunts too dangerous for human stunt performers, powered by Nvidia's new Newton
Meta is cutting its metaverse budget by up to 30%, and consumer platforms are ghost towns. But spatial
Ambient AI is growing 24.5% a year, faster than cloud computing or smartphones ever did, and it's already
The AI carbon footprint could nearly triple the electricity use of three major nations combined by 2030. But
The single biggest lever for cutting AI's energy footprint doesn't require new hardware: it's matching model size to task. Here's what the 2025-26 data says actually works.
Read article
EU AI Act enforcement arrived in 2026 with fines up to €35m. AI governance platforms went from optional
Greater AI knowledge correlates with less trust in police facial recognition, not more, and the evidence explains why:
Self-hosted AI models now deliver 70-85% of frontier quality at zero marginal cost per query. Here's why running
Robots in education are already deployed in 206 real classrooms across 28 countries. Here is what actually works,
Extended reality (XR) is moving from gaming into classrooms and workplaces. Here's how to prepare for XR properly, based on what worked and failed in earlier deployment waves.
Read article
China now has 515 million users on generative AI, larger than the entire population of the EU, reached
Bryan Cranston shifted OpenAI's deepfake governance through union advocacy alone, no legislation or lawsuit required. New research shows
The AI emissions paradox is real in both directions: data centres consume growing amounts of electricity, but the
The Louvre heist took just seven minutes, four thieves, a furniture lift, and one angle grinder. Here's the
The AI workplace divide doesn't run where most people think. New MIT research shows generative AI often helps lower performers more than top performers, complicating the standard high-skill-wins narrative. Here's what the data actually shows
Read article
Explainable AI just went from theory to measurement: researchers can now dial up a single internal vector in
Generative AI in creative industries has already crossed its tipping point. Here are the licensing deals, lawsuits and
The AI regulatory landscape split in two in 2026: the EU pressed ahead with binding AI Act enforcement
88% of organisations now use AI, but only a fraction have a real AI governance framework — and
Six months after launching to viral acclaim, OpenAI abruptly shut down Sora 2 following deepfake scandals, copyright fights and a pivot toward robotics ahead of its IPO.
Read article
AI workflow automation isn't just replacing tasks anymore, it's reconfiguring how work gets done entirely. McKinsey finds employees
RL environments have gone from research curiosity to the most fought-over infrastructure in AI, with Prime Intellect's $130
China's AI datacentre in Qinghai is already running at 3,579 petaflops on domestic chips alone, on track for
Albania's AI minister Diella just took a formal role awarding public tenders, the first cabinet-level AI appointment anywhere
