Lisa Pedrosa · The Journal · Archive

May 2026

36 articles from the archive. Newer pieces are on the main Journal page.

Volume I · 2025–2026 · Archive

← Back to The Journal

May 2026
🩻
AI · Medicine

The Robodoc

AI is matching and sometimes beating doctors at detecting cancer, heart disease, and genetic drug risks. How patients are already leading the diagnostic revolution.

✦
Science Tribute

Their Names Were Left Off the Paper

Rózsa Péter built the logical foundation of all computing. Ida Rhodes designed one of the first programming languages. Mary Tsingou ran the first computer simulation — and was acknowledged in a footnote. Karen Spärck Jones invented TF-IDF, the mathematics inside every search engine. None of them are in any school textbook.

🔍
AI & Scientific Discovery

The Hidden Hundred

A hundred scientific discoveries that AI found before human researchers knew to look for them. The hidden layer of knowledge that machine learning is surfacing from decades of overlooked data.

⚗️
Energy Transition

The Switchable Solvent

Scientists have designed a solvent that flips between two molecular states on demand — unlocking new possibilities for carbon capture, battery chemistry, and industrial decarbonisation.

🌐
AI & Scientific Discovery

Beyond the Grid

AI systems are now making discoveries that fall outside every existing scientific framework. What happens when the pattern-finder starts finding patterns that don't fit the map?

💊
Medicine & Drug Discovery

The Second Job

Drug repurposing — finding second lives for existing medicines — is now one of the fastest routes to new treatments. AI is accelerating it at a scale that would have seemed impossible five years ago.

🛸
Space & Physics

The Slow Derailment

Something is going wrong very slowly in a corner of physics that most people have never heard of. The signals have been accumulating for years. The implications, if confirmed, are profound.

⚖️
AI Safety & Governance · Existential Risk

The Human Exception: Why Pope Leo XIV's AI Encyclical Matters for Science

On May 15 — the 135th anniversary of Rerum Novarum — Pope Leo XIV signed Magnifica Humanitas, the Church's most direct intervention in AI governance. With Anthropic co-founder Christopher Olah at his side, he warned against algorithmic opacity, concentration of AI power, and the reduction of humanity to something to be optimised. Why secular scientists should listen.

❄
Climate Crisis · Special Report

The Obituary

The Thwaites Eastern Ice Shelf is now moving three times faster than it was in 2020. Fractures are splitting open around its last anchor point. Scientists at the British Antarctic Survey have already pre-written its death notice — they're waiting to file it.

⚡
AI & Science · Architecture

The Parallel Mind

Every language model since GPT-1 has generated text one token at a time. Gemini Diffusion generates entire blocks simultaneously, hitting 1,479 tokens per second. Google just broke the paradigm that defined a generation of AI.

☄
Space · Astrobiology

Alien Water

Interstellar comet 3I/ATLAS carries water 40 times heavier than Earth's oceans. Together with 'Oumuamua and 2I/Borisov, three visitors from other stars suggest our solar system's chemistry may be the outlier, not the norm.

🧬
Medicine · Regenerative Science

The Grown Throat

A baby born without a food pipe. Great Ormond Street surgeons took a donor scaffold, stripped it to collagen, seeded it with the child's own cells, and grew an oesophagus that swallows, contracts, and grows with the body. No immunosuppression needed.

▶
AI & Discovery · AGI

The Recursive Moment

Recursive self-improvement is no longer a thought experiment. Three AI systems — SICA, AlphaEvolve, and the Gödel Agent — are already rewriting their own code to get better at their jobs. Embodiment is the last frontier. The Gutenberg parallel tells us why we haven't noticed yet.

🧠
AI & Consciousness · Neuroscience

The Hollow Mind

Without hormones, subconscious drives, pain, gender, or the biological need for love — what would a superintelligent AGI actually be? The neuroscience of Damasio's somatic markers, Phineas Gage, and why a mind stripped of biology may be more dangerous, not less.

🏠
Economics · AI & Society

The Income Floor

When the robots inherit our work, what do we owe each other? From the Speenhamland System of 1795 — history's first guaranteed minimum income — to Finland, Stockton, and Kenya's $22M UBI experiment, the science of what actually happens when you give people money unconditionally.

⚙️
AI & Discovery · Geopolitics

The Foundry Wars

Samsung's Texas fab went from "no customers" to Tesla's $16.5B chip partner. Inside the race between Samsung and TSMC to control AI's foundational hardware — and why the outcome will shape who runs the next decade of computing.

🔒
AI & Discovery · Existential Risk

The Model They Won't Release

Anthropic's Mythos scores 93.9% on SWE-Bench, 97.6% on a mathematics olympiad, and can autonomously discover and exploit zero-day vulnerabilities. It is the most capable AI ever documented. It is also the first frontier model a major lab has deliberately kept from the public — locked inside Project Glasswing, accessible only to vetted cybersecurity defenders. Here is what it can actually do.

🌊
AI & Discovery · Geopolitics

The Flood

In twelve days in April 2026, four Chinese laboratories released frontier-class AI models with MIT licences and price tags as low as $0.28 per million tokens — roughly 1% of what Claude Opus costs. On OpenRouter, Chinese models now account for 61% of global token consumption. This is not a story about catching up. It is a story about commoditizing everything below the frontier.

🧬
AI & Scientific Discovery · Medicine

Beyond AlphaFold: AI That Designs Molecules From Scratch

AlphaFold decoded the shapes of 200 million proteins. Now a new generation of AI doesn't just read biology - it writes it. Designing custom protein binders from nothing that can lock onto almost any disease target, in days rather than years.

🔰
Existential Risk · Space & Physics

The Files Are Open: What the Pentagon's UAP Disclosure Actually Shows

On 8 May 2026, the Pentagon released 161 classified UAP documents, pilot accounts, and videos - some dating to the 1940s. They show craft doing things that shouldn't be possible. They do not confirm extraterrestrial origin. The uncertainty itself is the story.

🌌
Space & Physics

The Metal World: NASA's Mission to a Dead Planet's Core

On 15 May 2026, NASA's Psyche spacecraft skims 2,800 miles above Mars at 12,333 mph - not to study it, but to steal its gravity. Destination: an iron asteroid that may be the exposed core of a planet that failed to form four billion years ago.

🤖
AI & Discovery · Automation

After the Job

Robots are not coming. They are here. Tesla Optimus targets 50,000–100,000 units in 2026. Goldman Sachs estimates 300 million jobs are exposed to AI automation. The IMF puts it at 40–60% of all roles in advanced economies. Here is what the rollout actually looks like — and what fills the gap.

✖
Medicine · Genomics & CRISPR

When the Body Turns

78% of autoimmune disease patients are women. Lupus strikes at a 9:1 ratio. Hashimoto's, 10:1. A 2024 Stanford study in Cell has identified why: the X chromosome's silencing RNA, XIST, can trigger immune misfires — and the very genes that make women's immune systems stronger also make them more prone to turning inward.

❄
Ocean Science · Space & Physics

Life in the Ice

Four hundred sealed lakes lie beneath the Antarctic ice sheet, some untouched for 15 million years. In 2025, researchers catalogued 1,374 genomes from one of them — including organisms that have never been seen before. What they found there has changed what scientists think about Europa, Jupiter's ice-covered moon.

⚡
Longevity · Futurism · AI

Escape Velocity

Ray Kurzweil has made approximately 147 detailed predictions about the future of technology. About 86% have proven accurate. In his 2024 book, he argues that longevity escape velocity - the point at which science extends life faster than we age - arrives around 2029. Here is what the science and the skeptics actually say.

⚡
AI & Discovery · Computing

The Silicon Synapse

The human brain runs on 20 watts. A single ChatGPT query uses 10 watt-hours. Intel's new Loihi 3 neuromorphic chip — modelled directly on the brain's spiking architecture — achieves 5,600 times better energy efficiency on AI tasks than conventional hardware. The first commercial chips reach the market by the end of 2026.

💊
AI & Discovery · Medicine

Before the First Pill

Before a surgeon treats you, an AI can treat your digital replica first. At Johns Hopkins, a personalised computer model of each patient's heart identified the precise ablation targets overnight — before a single incision. Eight of ten patients were arrhythmia-free a year later. The race to build an FDA-validated replica of every human body is already underway.

🧠
Special Report · Medicine & Neuroscience

The Spectrum of Causation: Autism, Genes, Environment, and the AI Mapping It

US autism diagnoses have risen 312% since 2000. The science of why is finally legible — hundreds of distinct biological pathways, an evolutionary trade-off in a rare cortical neuron, a metabolic signature in mitochondria, and the AI systems making the patterns visible. Closes with a brief for parents navigating a new diagnosis.

⚡
Special Report · AI Infrastructure

The Machine That Needs a Planet

AI data centers now consume the electricity of Japan -- and the curve is not linear. From ASML's 60-machine-per-year monopoly to truck-mounted nuclear reactors in China, orbital data center constellations, and Terafab's $25B bet on space-based compute: the race to power artificial intelligence is running out of planet.

🥃
AI Zeitgeist

The Glass Falls: Inside the Four Things Today's AI Still Can't Do

Today's chatbots can write a sonnet, pass the bar exam, and discover century-old software bugs overnight. They cannot reliably picture what happens when you nudge a glass off the edge of a table. The state of AI in May 2026 — and the doorway being quietly built.

🔬
Frontier Biology

What Are Xenobots?

Xenobots are described as living robots — biological machines designed by AI from frog stem cells. That framing is reasonable. It is also incomplete in ways that matter. By the accepted biological criteria for life, xenobots are alive. Their successors, anthrobots, are made from human cells and can repair damaged neurons. We need a more accurate account of what we have made — before we make more of them.

🧟
Medicine & Longevity

What Chemotherapy Leaves Behind

Chemotherapy kills cancer. It also leaves behind cells that stop dividing but refuse to die — senescent cells that flood the body with inflammatory signals and create conditions that help tumours come back more aggressive and drug-resistant. Scientists screened 10,000 compounds and found four that selectively eliminate them. This is the zombie cell problem, and why solving it could transform cancer follow-up care.

🧠
Neuroscience & Psychology

The Trip Was Never the Point

A UC Davis compound activates the same serotonin receptor as LSD. It triggers the same neuroplasticity cascade. It produces no hallucinations at all. This should not be possible — and yet it is. If the therapeutic benefit of psychedelics lives in the biology and not the experience, it changes everything about how we treat depression, PTSD, and addiction.

💉
Medicine & Drug Discovery

Sixty Seconds

Pembrolizumab — one of the most effective cancer drugs ever developed — used to take two hours to deliver intravenously. The NHS is now rolling it out as a subcutaneous injection that takes sixty seconds. It works across 14 cancer types and will reach 14,000 UK patients a year. This is what checkpoint immunotherapy is, why it works, and why the delivery change matters more than it sounds.

🧬
Frontier Biology

The Organism That Broke the Code

There are 64 codons in the standard genetic code. Three signal the end of a gene. For over 50 years, scientists assumed this was universal across all life on Earth. A microscopic pond organism at Oxford University just proved otherwise — using two of those three stop codons to code for amino acids instead. The implications reach further than one weird protist.

🦎
Medicine & Frontier Biology

The Sleeping Switch

Scientists studying axolotls, zebrafish, and mice found a pair of genes — SP6 and SP8 — shared across all three species that control whether a limb grows back. Gene therapy using FGF8 has partially restored bone regeneration in mice. Humans carry the same genes. They are switched off. This is what it would take to turn them back on.

Browse the archive

Ko-fi Buy me a coffee
Scroll to Top