Start your morning scandalously informed.
A daily curated digest with the tech news that matter + community vibes, delivered daily, in tabloid style. Like you always wanted.
Thursday, March 26, 2026
Jury Says Meta Poisoned Kids' Minds!
Meta faces a furious jury as parents and lawmakers confront how social apps shape children... In Washington, the Supreme Court eases pressure on internet providers over piracy, while in Europe a revived Chat Control plan puts private and encrypted messages under the spotlight... GitHub Copilot starts learning harder from developer keystrokes, even as a new local LLM stack promises powerful models that live on our own devices... A fresh ARC‑AGI‑3 benchmark asks if today’s AI can really reason, while a simple context hijack trick shows how fragile chatbots still are... At the labs, CERN scientists move trapped antimatter without destroying it, opening new tests of physics, and a sodium‑ion EV battery teases cheaper, faster‑charging cars... On GitHub, rival bots like Claude quietly help build OpenAI tools, and we watch the lines blur between companies, code, and machines.
Courts Hit Big Tech As Physics Leaps
Jury slams Meta for harming children’s minds
A New Mexico jury found Meta knowingly hurt kids’ mental health and hid what it knew about abuse and addiction on Instagram and Facebook. The giant payout stings, but the real hit is reputational: lawmakers and parents now smell blood in the water.
Supreme Court shields Cox from piracy liability blow
The US Supreme Court’s ruling in Cox v. Sony pulls back on holding internet providers responsible for users’ pirated music. Labels lose a powerful weapon, while ISPs quietly sigh in relief. Ordinary users just see more confusion over who actually polices online sharing.
EU resurrects plan to scan private chat messages
The Chat Control proposal is back from the dead, aiming to scan private and even encrypted messages for illegal content. Supporters shout “think of the children,” while critics say it nukes digital privacy for everyone and effectively outlaws truly secure apps.
Scientists finally move antimatter without blowing it up
Physicists at CERN and HHU managed to transport trapped antimatter in a portable magnetic “bottle,” something that used to live only in sci‑fi plots. It’s a tiny amount, but it proves antimatter can be stored and moved, opening fresh ways to test the laws of physics.
Sodium EV battery promises dirt-cheap, fast-charging cars
Engineers at BAIC unveiled a sodium‑ion battery for electric cars with around 450 km range and 11‑minute charging. By dumping scarce lithium for common salt, it teases cheaper EVs and less dependence on fragile mining supply chains, if it survives real‑world roads.
AI Labs Race Ahead And Break Things
GitHub Copilot starts training harder on your keystrokes
GitHub confirmed that Copilot will collect inputs, outputs, and snippets from users to build “more intelligent” AI helpers. It sounds convenient until you realise your coding style becomes training data by default, leaving developers uneasy about privacy and company secrets.
Startup pushes powerful AI models onto your own device
Ente announced a local LLM stack, arguing that models are “too important to leave to big tech.” Instead of shipping your thoughts to remote servers, the plan is to run capable models on your hardware, trading some raw power for privacy and real control.
New ARC-AGI-3 benchmark asks if AI can really think
ARC‑AGI‑3 sets loose AI agents in strange little worlds and asks them to figure things out with no hand‑holding. It’s less about pretty chat answers and more about reasoning, exploration, and learning on the fly, which is exactly where today’s flashy models still stumble.
OpenAI repo quietly credits Claude as top coder
Developers noticed Claude listed as a top contributor on an OpenAI GitHub repo, sparking jokes that rival AI models now write each other’s tools. It’s a weird moment: corporate boundaries blur while humans watch two competing labs’ bots collaborate on shared code.
“Disregard that” trick shows how fragile chatbots are
A researcher shows how sharing an AI chat context window with others lets them slip in messages like “disregard that” to hijack the conversation. It’s a simple, nasty reminder that today’s LLMs are huge parrots with no real memory of who to trust, or who’s even talking.
Hackers, Gadgets And Gatekeepers Stir Trouble
Researcher runs Tesla’s brain on a desk at home
A security researcher bought Tesla Model 3 parts from crashed cars and got the Media Control Unit running on his desk. It’s impressive, slightly terrifying, and shows how far dedicated hackers will go to probe modern cars that are basically rolling computers.
Fed-up Apple fan says latest lockdown finally lost them
One longtime user says Apple “just lost me” after tighter macOS controls, Gatekeeper nagging, and a general feeling that the company no longer trusts its own customers. The rant clearly resonates with power users tired of being treated like clueless iPhone newbies.
Ubuntu plans to chop GRUB features for security
Ubuntu wants to strip down GRUB in version 26.10, removing lots of file system parsers to simplify Secure Boot. It might boost safety but leaves tinkerers worried their favorite boot tricks and rescue workflows will vanish in the name of corporate‑friendly security.
Dev rewrites Git in Zig to feed AI cheaper
A developer built Nit, a Git‑like tool in Zig, to store repos in a way that saves AI agents about 71% on tokens. It’s geeky but clever: less text sent to LLMs means lower bills and faster responses, at the cost of yet another tool to learn.
How to stop airport agents rifling through your phone
A guide on dealing with ICE and CBP at airports explains how to protect your devices, from using eSIMs to leaving sensitive accounts logged out. It paints a bleak picture of border searches but offers concrete steps for travellers who don’t want their digital lives exposed.