simonwillison.net/2025/Dec/31/... Agent, context, vibe coding in buckets of slop: Simon Willison's treats us to is annual review of LLMs

futurism.com/space/max-sp... When the International Space Station becomes out of service in 5 years, the private space industry must lead the innovation charge. One idea is Thunderbird, an inflatable space station that opens up once in orbit instead of flying in parts. More like a Transformers.

Futurology scoping 100 years forward sounds too tricky to be accurate. No flying cars in this list but some trends that are fairly on money.

us18.campaign-archive.com?u=e0354b6d74... Why we need a PET (privacy, empathy, trust) more than fancy coffee at work, the one workforce group that will rule and other future of work trends from Julia Hobsbawn

futurism.com/artificial-i... Those two recurring questions circling around 2025: How can I get my brand mentioned in ChatGPT? How can OpenAI become profitable? We have an obvious answer. A swift enshitiffication likely.

cryptorank.io/news/feed/48... From zero to hero in a 2 billion acquisition, Meta buy Manus, a less than 2 year launched agentic AI startup with focus on practical application. Meta seek foothold across AI infrastructure thru to applications

futurism.com/artificial-i... Googling yourself has always thrown up some wrong attribution horrors (as even celebs like Ian Watkins from Steps discovered). Now AI overviews serving up facts risks one non-fiddling fiddler's reputation

www.linkedin.com/posts/matthe... AI powered kids toys sound like a dodgy idea, and indeed these ones tested to reveal sexual info including a very prickly cactus

www.404media.co/weird-disney... AI slop delivers the ultimate Xmas gift. But will it arrives? This reads like plot of a Hallmark Xmas film

futurism.com/artificial-i... Europol predict within 10 years cyber criminals will hack into drones, autonomous vehicles and even medical devices to wreck havoc

Autonomous cars politely stopping for a blackout is very San Francisco.

A neat analogy for thinking about the boundless enthusiam vs competence of agentic AI from Chris Weston - You've Just Been Given 100 Homers. Now What? www.linkedin.com/pulse/youve-...

www.linkedin.com/posts/ruben-... 10 free Harvard lectures on AI: bookmark for Betwixmas?

techcrunch.com/2025/12/21/w... In a scene out of an apocalypse film, a blackout in San Francisco switches off traffic lights causing Waymo's to stop. A good safety feature as they won't easily pass through off traffic lights but showing gaps in difference between autonomous and human drivers

Large language models are getting weirder. When experts don't understand how AI works, why are we doggedly trusting it? 🔁New edition of tech + AI trends: Rethinking the Hype Cycle #23 rethinkhypecycle.substack.com/p/large-lang...

youtu.be/kx_okFl98Vw?... "That's not a growth projection l, it's a hockey stick by someone who's never played hockey"

LLM AI models can do incredible things but scientists still can't explain why they work. From "grokking" to predicting language, it's a concerning and weird phenomena beyond hallucations www.technologyreview.com/2024/03/04/1...

Is the march of AI and tech inevitable? There are alternative models of technofuturism and AI outside of the US big tech monolith, says @zoescaman.bsky.social zoescaman.substack.com/p/on-inevita...

A dark year in review: how tech embraced the fake: from killing community moderators, deepfake nudes and the relentless march of AI slop in social. What's real and what's fake? this is flooding hte zone. indicator.media/p/2025-the-y...

Palantir and Accenture form a new group with "particular focus on the complex operations of data center and AI infrastructure programs critical to economic resilience." The big guns lobbying for the big guns? newsroom.accenture.com/news/2025/ac...

Congrats to @carolecadwalla.bsky.social for most watched TED talk 2025 exposing what a digital coup looks like: www.youtube.com/watch?v=TZOo... Essential viewing. Had pleasure of meeting Carol at @thenerve.news event last week interviewing Stewart Lee, next time let's get Carol front of stage!

The new factories are AI factories: Kenyan young workers exploited by unemployment and 'middle men' work for pittance pay labelling AI data restofworld.org/2025/kenya-c...

MIT claims we're entering the post-AI hype phase. Here's hoping. This series of essays helps consider where AI hype has taken / is taking us www.technologyreview.com/supertopic/h...

Still think it's OK to let your kids play with AI friends? 37% engage in violent conversations, and half of those sexual violence futurism.com/future-socie...

New ChatGPT Images model is out and looking forward to trying it. Gemini's Nano Banana's images are good but editing is hopeless. ChatGPT promises more customisation to alter layers and sections. Cue the cheesy Xmas on the moon cards. openai.com/index/new-ch...

Influencer marketing has always had its rogue players, but this one goes big: 1,000 hacked smartphones to generate videos on TikTok with fake quasi-realistic AI influencers www.404media.co/hack-reveals...

thehackernews.com/2025/12/feat... Cyber extension secretly spies and harvests data from AI chats. These devices whisper

www.theneuron.ai/explainer-ar... Raw intelligence is a diminishing return as models are more or less the same. The AI technologists that succeed will build agents that build boring consistency to focus on context, need and reliability, according to investor Gavin Baker.

www.anthropic.com/research/how... Anthropic's survey of it's engineers shows Claude enables them to become more "unstack" shifting from Coders to AI managers enabling 27% of tasks that wouldn't have otherwise been shipped.

When half the internet is made by AI, how do you make your expertise unmistakable? New article: rethinkhypecycle.substack.com/p/how-to-bec...

boingboing.net/2025/12/13/a... McDonalds make bad ad using AI about how bad Xmas is. People hate it and it gets pulled. Certain irony all round.

mashable.com/article/tesl... "AI not remote operated" says Musk. Optimus robot shows telltale signs of remote operator removing their headset at recent demo fail.

boingboing.net/2025/12/12/f... As a clarinettist, these can be weapons of a mean jam. But not a gun. As lame AI detection software thinks

time.com/7339685/pers... "Architects of AI" named TIME'S Person of Year. All men featured in article. Mostly not architects. A neatly wrapped state of the AI nation.

data-workers.org/michael/ Your AI girlfriend? It may have not been AI but a Michael in Nairobi trying to provide for his family while assuming numerous identities. It's not just mechanical turk, there's a painful emotional toll for the hidden data workers

futurism.com/future-socie... Knock off music is no longer dodgy CDs. Artists leaving Spotify find their songs cloned

www.bbc.com/news/article... Creatives in design, art, video and copy share their experiences of AI shifting, or crushing, their work opportunities

futurism.com/artificial-i... "Trusting the AI blindly was my mistake". AI is very articulate and sincere... when it screws up and deletes all your work

Grok AI is a playground for stalkers. Ask it for a person's private info and it doxes (exposes personal details) a dosier of info, including their family home cohabitees futurism.com/artificial-i...

Bad news for AI providers: US enterprise and bigger biz folks who report using AI are going DOWN - hitting a wall of around 40%. It could be growth comes from smaller biz and consumer. But 2026 is going to need a big uptick to make investments stick. futurism.com/artificial-i...

Johnny Cash would have loved this. Scientists and fashion designer take inspiration from a Papua New Guinea bird to create an ultra-black fabric - the darkest ever known. www.newsweek.com/ultrablack-p...

AI data centres wilt like lettuce. Melting chips and job automation, is this the tip of the AI iceberg? 🔁 Rethinking the Hype Cycle #22 rethinkhypecycle.substack.com/p/ai-data-ce...

Musk finally fined in the anti online safety rap bill from hell. "The EU concluded that X had become the single most dangerous vector of election disinformation in Europe." open.substack.com/pub/deanblun...

vm.tiktok.com/ZNR8QTLno/ Tech bros and art pioneers imagined by artist Beeple as dog robots shitting out Polaroids of their world view

www.linkedin.com/posts/clowil... 3 years of ChatGPT, Clo Wlliart sums up 3 years of chaos and amazement, from reading mammograms to chatbot suicide hotlines

www.theguardian.com/society/ng-i... Zuckerberg's passion project the Metaverse is facing death of a thousand virtual cuts. Will it be granted a second life?

Like new features in ChatGPT like group chat - is this a sign that consumer AI is going to forge a way forward as a hybrid search and social network? openai.com/index/group-...

Trouble in the OpenAI mill: As Google's Gemini captivates more of the imagination and market, Altman sounds a code red alert with memo to focus on improving core features of ChatGPT to catch up. www.theverge.com/news/836212/...

AI data safety trainers tell family and friends to avoid AI. Working in sub-optimal conditions, rapidly responding without consideration to tests doesn't bring confidence that there's a thorough approach to safety. www.theguardian.com/technology/2...

That poetry Master's? Not as useless as your dad thinks. LLMs's safety features can be run around by injecting prompts into poetic verses. The rhythm confuses LLMs that are based on predictive text. www.theguardian.com/technology/2...