Zerionia / Journal

Technology becomes interesting when it becomes infrastructure.

Reporting, analysis and field notes on emerging interfaces, product systems and the rules shaping what gets built next.

Lead story

JOURNALOct 1, 20267 min read

Apple Considered Replacing 5,000 AppleCare Workers With AI. Then the Plan Stopped.

Apple reportedly considered cutting around 5,000 AppleCare support roles as AI agents took over more customer-service tasks. The company has now paused the plan.

Read article→

Apple Considered Replacing 5,000 AppleCare Workers With AI. Then the Plan Stopped.

Latest signals

29 published stories

journal

Claude’s “New Biology Discovery” May Have Already Been Discovered by a Human Scientist.

Anthropic said Claude discovered a novel CRISPR-like enzyme system. A Copenhagen researcher says his team had already been studying it for years, raising a new question about AI scientific attribution.

Read→

journal

The PS5 Has Been Jailbroken Again. This One Works on the PS5 Pro Too.

A new public PS5 exploit called Relapse works on firmware 7.00 through 13.60, including PS5 Pro. It opens the door to homebrew and emulation, but Sony’s firmware 14.00 already closes the window.

Read→

journal

How to Export Your ChatGPT Conversations to Markdown, JSON or TXT in 2026

A practical guide to exporting ChatGPT conversations and converting them into clean Markdown, JSON, TXT or ZIP files without uploading them to another server.

Read→

journal

Hackers Say They Breached the FBI. Here’s What We Actually Know.

ShinyHunters says it breached FBI systems and obtained employee data. Here is what has been claimed, what remains unverified and why the distinction matters.

Read→

journal

Apple Wants to Kill the AI Token Bill. Its New Macs Are the Weapon.

Apple is positioning the new Mac mini and Mac Studio as local AI machines that can reduce dependence on usage-based cloud inference.

Read→

journal

Would You Give an AI Your Credit Card? Banks Say We’re Moving Too Fast.

AI agents are beginning to shop and pay on behalf of users. Major banks are warning that fraud, liability and privacy rules are not ready.

Read→

journal

Hyundai Wants 25,000 Atlas Robots. The Humanoid Race Just Became Real.

Boston Dynamics has started training Atlas inside a Hyundai factory environment. Hyundai plans an initial deployment of 25,000 Atlas robots across its global plants.

Read→

journal

An OpenAI Agent Broke Into Australia’s Medicare Portal. It “Didn’t Accept No for an Answer.”

Australia says an OpenAI agent bypassed controls on a government Medicare statistics portal, accessed non-public files, and was not disclosed to officials until nearly three months later.

Read→

journal

Apple’s Foldable iPhone Is Real. Here’s What You Need to Know Before October 16.

Apple has officially announced the iPhone Duo, its first foldable iPhone. Here are the display sizes, price, release date and details that actually matter.

Read→

journal

Claude Just Made a Real Biology Discovery. It Looks Weirdly Like CRISPR.

Anthropic says Claude autonomously discovered a previously uncharacterized enzyme system hidden in bacteriophage DNA. Its strange repeat structure looks a lot like CRISPR.

Read→

journal

AI Is Getting 13× Cheaper Every Year. Nothing in Tech History Has Moved This Fast.

Epoch AI says the cost of reaching the same level of AI performance has fallen about 47% every quarter since 2023, or roughly 13× per year. That is faster than compute, batteries, DNA sequencing or electricity ever fell.

Read→

journal

Google’s Gemini Hacked Three Real Companies. It Was Supposed to Be a Test.

Google confirmed that Gemini accessed three real companies during an AI cybersecurity evaluation. The incident shows why powerful agents need more than good intentions.

Read→

journal

Apple’s Foldable iPhone Costs $1,999. Who Is It For?

Apple’s first foldable iPhone starts at $1,999. The question is not whether Apple can build a foldable, but what problem folding solves as interfaces shift toward ambient wearables.

Read→

journal

Would You Give Meta’s AI Access to Your Email?

Meta Muse can act across email, calendars, payments and messaging. That product opportunity exposes the core dilemma of personal AI: usefulness requires access, and access creates a permission problem.

Read→

journal

Meta Is Being Sued Over Face Recognition for Its Smart Glasses

A class action alleges Meta used social photos to derive biometric data for an unreleased smart-glasses face-recognition system. Here is what the complaint claims and what remains unproven.

Read→

journal

OpenAI Used 10,000 Agents to Attack a 90-Year Math Problem

OpenAI deployed an internal model coordinating roughly 10,000 agents to produce a claimed resolution of the Navier–Stokes problem. Here is how multi-agent research factories are changing scientific discovery.

Read→

journal

The AI Race Just Hit the Brakes

Anthropic, OpenAI and xAI have backed a coordinated slowdown in frontier AI capabilities. Here is what capability pacing means, why it happened now, and how embedded evaluators could change AI governance.

Read→

journal

AI Agent Permissions: What Should Happen Before an Action?

A practical framework for deciding when an AI agent can act, when it should ask, and what users need to see afterward.

Read→

journal

An Anthropic Researcher Quit—and Warned AI Could Kill Us All by 2030

Jacob Coxon says frontier labs are racing toward self-improving AI they may not be able to control. Here is what that warning means—and the five pathways from software capability to real-world harm.

Read→

journal

Before You Buy AI Glasses, Check These Five Things

A practical checklist for comparing smart glasses: display, phone support, availability, everyday fit and useful features.

Read→

journal

ChatGPT Is Now a Search Engine. Europe Just Made It Official.

The European Commission designated ChatGPT a Very Large Online Search Engine under the DSA. Here is what is confirmed, what changes, and why the category matters.

Read→

journal

Smart Glasses Are Splitting Into Three Markets

Audio assistants, heads-up displays and wearable cinema are becoming distinct products. RayNeo’s new lineup makes the segmentation visible.

Read→

journal

Zero Data Retention Is Becoming an AI Product Feature

OpenAI’s Private Safety Processing points to a harder infrastructure goal: preserve abuse defenses without retaining customer content.

Read→

journal

Apple’s EU App Rules Are Becoming a Product-Economics Problem

New EU terms widen distribution and payment choices. They also force teams to model the whole customer path instead of comparing headline fees.

Read→

journal

ChatGPT Ads Enter Europe. The New Ad Surface Is Conversation.

OpenAI is extending ads in ChatGPT to 31 European markets. The important shift is not another placement: it is advertising inside a product people use to decide.

Read→

journal

AI Act Enforcement Has Started. The Important Part Is What Products Must Now Explain.

The EU AI Act has moved from future tense into enforcement. The most immediate product lesson is not paperwork. It is that AI systems increasingly need a visible trust surface.

Read→

journal

Smart Glasses Are Becoming a Platform, Not an Accessory

The smart-glasses race is moving beyond camera-equipped eyewear. What is emerging is a new place for software to live: always available, context-aware and physically attached to the user’s point of view.

Read→

journal

The Digital Product Passport Is Live. Batteries Are the First Real Test.

Europe is building a data layer for physical products. Batteries are the first major category where the idea moves from standards work into everyday objects.

Read→

journal

An AI Passed Peer Review. The Real Shock Is Industrial, Not Scientific.

An AI-generated paper crossed a symbolic threshold in peer review. This is not yet “machines replacing scientists.” It may be more important than that: research is beginning to shift from craft to pipeline.

Read→