Article archive
Published news and blog articles, organized by category. Browse older coverage by month or search for a topic. Undated blog guides appear after dated news.
2,222 articles · Newest first
Inside the LLM Training Run: Data, Scale, and Compute
How frontier language models are actually built - corpora, token budgets, and the compute economics behind a training run.
Claude Fable 5.1: Anthropic's Latest Model Arrival, Explained
Anthropic's Claude Fable 5.1 lands as the latest point release in the Claude model line. N43 explains what shipped, where it fits the family, and what the release-cadence race means.
How the Electrical Grid Is Being Rebuilt for AI
AI data centres are colliding with the limits of the electrical grid. N43 explains the scale of the demand, why queues and lead times dominate the buildout, and what changes for consumers.
Tesla's Cybercab: The Robotaxi Era Begins, or Another Promise?
On September 4, 2026, Tesla began selling paid rides in its purpose-built Cybercab robotaxi in Austin, Texas. N43 measures that moment against Waymo's years of operational head start, and asks whether the robotaxi era has truly begun.
Ultra Phones Ranked: What 2026's Top Flagship Tier Tells Us
Tech Spurt has ranked the 2026 ultra flagships. N43 takes the ranking as a starting point: what the ultra tier actually buys, where the money in these devices goes, and what the flagship race says about where phones head next.
E-Ink's Bistable Trick: How Electronic Paper Holds an Image With Almost No Power
Inside electrophoretic e-paper: charged pigment particles, microcapsules, and the bistability that lets a screen hold an image with the power switched off.
Genie 3 and the World-Model Era: When AI Learns by Simulating
Google DeepMind's Genie 3 turns text prompts into navigable photorealistic worlds at 20-24 frames per second — and the world-model research line behind it may matter as much as any chatbot.
How a Phone Knows Up From Down: The MEMS Accelerometer Explained
Inside the MEMS accelerometer: a spring-mounted proof mass etched into silicon, capacitive readout measured in fractions of a nanometre, and the sensor fusion that keeps a phone oriented.
Why Chinese AI Labs Caught Up: Open Weights, Hardware Limits, and the New Model Race
Export-control-constrained chips pushed Chinese AI labs into ruthless efficiency. The open-weights release strategy that followed is reshaping the economics of frontier AI.
GPT-6, AGI talk and tech euphoria: what the All-In Podcast debate reveals
GPT-6 has landed and the All-In Podcast crew is asking whether AGI is next. A look at the benchmark strain, the hype cycle, San Francisco's mansion shortage and the debate over AI in classrooms.
Large language models explained: how tokens, training and tuning actually work
A plain-language guide to large language models: what they are, how tokens and context windows work, what pretraining and fine-tuning really involve, why hallucinations happen, and where the technology is heading.
Samsung Galaxy S26 Ultra: the most complete phone of 2026, reviewed
Samsung Galaxy S26 Ultra review: Privacy Display, ALoP periscope camera, 60W charging, Snapdragon 8 Elite Gen 5 and seven years of updates make 2026's most complete flagship.
Vision language models: how AI learned to see and understand images
How vision language models work: from CLIP and image encoders to cross-modal alignment, the limits of what VLMs can see, and where video and embodied AI research heads next.
Brace Yourself: Understanding the AI Bubble Argument and What It Means for Tech
Economist Robert Reich walks through the case that AI stocks are a bubble in the making — circular investments, inflated valuations, and a dot-com-shaped hangover. We break down the argument, the numbers behind it, and where the…
The Thinking Game: Inside DeepMind's Decade-Long Quest for Artificial General Intelligence
A feature documentary follows Demis Hassabis and the DeepMind team across five years of the race toward AGI, from AlphaGo's board games to AlphaFold's protein structures and beyond. It is the most complete public record of how the modern…
Titanium Screens and Hinge Lessons: What Foldable Durability Tests Really Mean
The Galaxy Z Fold 8 arrives with a so-called titanium screen and fresh durability claims. We look at what scratch, bend, and hinge tests actually measure — and why the foldable form factor keeps testing the physics of glass, ultrathin…
Why 5G Underdelivered: A Reality Check on Cellular Network Hype
5G was sold as a revolution: gigabit phones, remote surgery, self-driving cities. Years in, real-world speeds have barely improved for most users. A breakdown of the physics, the marketing, and what mmWave actually delivers versus what it…
How Language Models Learned to See Math: Why Vector Spaces Made LLMs Visual
Inside the high-dimensional geometry that turns words into arrows and meaning into direction. A look at why the embedding-space picture of language models, popularized by 3Blue1Brown, makes their behavior finally legible.
The 300-Millisecond Brain: Attention, Sampling, and What an LLM Actually Computes
A forward pass takes milliseconds: tokens in, one probability distribution over the next token out, then sample and repeat. Why that bare loop, the mental model Andrej Karpathy teaches, produces everything else.
The Chip Is the Phone: Why 2026's Silicon Matters More Than Its Specs
As camera and display improvements flatten, the mobile SoC now defines the flagship experience: NPU throughput, memory bandwidth, and thermal design decide what a phone can actually do.
The Upgrade Math Has Changed: Why the Two-Year Phone Replacement Cycle Broke
Flagship phones no longer improve fast enough to justify replacing a two-year-old device, and seven-year software commitments have rewritten the consumer upgrade equation.
AI Agents in 2026: From Chatbots to Software That Acts
AI agents plan, call tools, browse and execute multi-step work. A plain-English guide to the agent loop, computer use, memory, guardrails, failure modes and the economics of delegated work.
The $20 Question: ChatGPT Plus vs Claude Pro vs Gemini in 2026
What ChatGPT Plus, Claude Pro and Gemini actually offer at $20/month in 2026: model families, context windows, usage limits, multimodality and a practical guide to choosing.
iPhone 18 Pro Max: What the Leak Cycle Says Before Apple Speaks
Also explore blog articles and guides or search all DutyStation pages.