In 1940 surgeons began cutting the corpus callosum, the bridge between the two halves of the brain, to treat epilepsy, yet the patients seemed unchanged. In 1962 W.J. said he saw nothing when a shape was flashed to the left of his gaze, while his left hand pointed straight to it. What kind of answer counts as a report from the silent right hemisphere? Part 3 of the series A History of the Science of Consciousness.
In 1924 Hans Berger began recording the brain's electrical rhythms, and in 1929 he published them as the EEG. The EEG could tell waking from sleep, and a lapse of consciousness, without asking anyone. But mental images left no trace in the waves, and reading REM periods as dreaming rested on what woken sleepers said. Part 2 of the series A History of the Science of Consciousness.
In 1850 Fechner found a way to measure the threshold of sensation, and Donders timed the steps of the mind by subtraction. But trained observers flatly disagreed over whether thought contains images, and in 1913 Watson called for dropping consciousness from psychology. Part 1 of the series A History of the Science of Consciousness.
From 1856 Mendel spent eight years counting peas and found the 3-to-1 rule. Why did his paper go largely unread for 34 years? His letters to Nägeli and the plant he chose to test it, the triple rediscovery of 1900, Fisher's charge that the numbers were too good, and a gene that had a name before it had a substance. Part 1 of the series What Are Genes Made Of?
From the 1950s, America's state hospitals emptied while psychiatrists disagreed over diagnoses. Rosenhan's 1973 pseudopatient study and how its records were re-examined fifty years later, the removal of homosexuality as a diagnosis, and DSM-III's checklist approach of 1980. Part 7, the final part, of the series A History of Psychiatry.
In early twentieth-century Japan, most people with mental illness were locked in a room at home with official permission. From Shuzo Kure and Goro Kashida's 1918 survey of 364 such rooms, through the end of home confinement in 1950 and the boom in private mental hospitals, to the Utsunomiya Hospital scandal and the 1987 Mental Health Act. Part 6 of the series A History of Psychiatry.
Lithium came out of guinea-pig experiments, chlorpromazine from an aid to anaesthesia, imipramine from a failed schizophrenia drug. Three psychiatric drugs that appeared within ten years of 1949, the placebo-controlled trials that tested them, and the dopamine hypothesis built afterwards. Part 5 of the series A History of Psychiatry.
Fever, coma, convulsions, surgery on the brain: in the first half of the twentieth century four treatments that acted hard on the body entered psychiatry, and two won Nobel Prizes. Set beside the comparisons made later, two of the four held up. Malaria fever therapy, insulin coma, electroconvulsive therapy and lobotomy. Part 4 of the series A History of Psychiatry.
Anna O., the founding case of psychoanalysis, was not cured in the way the story was told. The strong claim Freud drew from eighteen patients in 1896 was later withdrawn by Freud himself. And when Hans Eysenck compared treated patients with untreated ones in 1952, no difference could be shown. Part 3 of the series A History of Psychiatry.
General paresis, a killer of asylum patients, was traced to its cause in 1913, when Hideyo Noguchi found the syphilis organism in patients' brains. The approach did not carry over to other illnesses, so Kraepelin sorted them by course instead of cause — and from that came the name schizophrenia. Part 2 of the series A History of Psychiatry.
The chains at Bicêtre were taken off not by Pinel, as the textbooks say, but by the ward governor Pussin, in 1797. From moral treatment — talk, and sometimes fear — and the York Retreat, to American asylums claiming nine in ten cured, and Pliny Earle's 1876 recount. Part 1 of the series A History of Psychiatry.
In 1935 the EPR paper asked whether quantum mechanics is complete, Bohr answered under the same title, and Schrödinger left behind a cat and the word entanglement. What the cat was meant to show, how Bell turned the question into an experiment in 1964, and what the loophole-free tests of 2015 settled and did not. The final part of the series The Rise of Quantum Mechanics.
At the Solvay conferences of 1927 and 1930 Einstein pressed Bohr with thought experiments, and Bohr defended the uncertainty relation. But the proceedings and letters of the time show Einstein asking something else: is the theory complete, and what happens between distant places? The movable screen, the box of light, and the question left open. Part 5 of the series The Rise of Quantum Mechanics.
In 1928 Bloch showed that an electron wave passes through a perfectly regular crystal without bumping into anything. That made insulators the puzzle. From Wilson's 1931 answer (it is how full the bands are) to a time when silicon was filed under metals, and to the solids the theory could not reach. Part 4 of the series The Rise of Quantum Mechanics.
In 1928 Dirac wrote an equation that made quantum mechanics agree with relativity. Spin came out for free — and so did unwanted answers with negative energy. From the full house and its empty seat, through the proton mix-up and the anti-electron, to the positron Anderson found in a cloud chamber in 1932. Part 3 of the series The Rise of Quantum Mechanics.
In 1928 Gamow, and independently Gurney and Condon, used quantum mechanics to show that alpha particles leave the nucleus by slipping through a wall they cannot climb. From a 10²¹-fold spread in lifetimes to fusion inside stars, and the old physics still hiding in the word “tunnelling”. Part 2 of the series The Rise of Quantum Mechanics.
In 1927 Heitler and London used the exchange of electrons to explain, with quantum mechanics, why two hydrogen atoms bond. From Sugiura to James and Coolidge, six years of calculation brought the numbers in line with experiment. Part 1 of the new series The Rise of Quantum Mechanics.
From 1924 to 1927: de Broglie's matter waves, Heisenberg's matrix mechanics, Schrödinger's wave equation, Born's probabilities and the uncertainty relation. How quantum mechanics was born by letting go of the electron's path, and what it kept from the old quantum theory. The final part of a five-part series.
In 1916 Sommerfeld let the electron's path be an ellipse, added relativity, and explained the fine splitting of hydrogen and helium-ion lines. But the helium atom, with two electrons, and the anomalous splitting of lines in a magnetic field would not fit, however the theory was patched. Part 4 of a five-part series on the old quantum theory.
In 1913 Bohr's model of the atom calculated hydrogen's spectral lines from basic quantities alone, and matched the helium-ion measurements to five significant figures. Yet why it worked was never explained. Part 3 of a five-part series on the old quantum theory.
In 1905 Einstein argued that light is a collection of energy lumps. The idea was not accepted even after experiment showed the predictions were right — the man who tested them did not believe it either. Part two of a five-part series on the old quantum theory.
Quantum mechanics is said to have begun on 14 December 1900. But Max Planck did not think he had opened a door — for him, chopping energy into pieces was a last resort, not a conclusion. Part one of a five-part series on the old quantum theory.
After an OpenAI internal model produced hundreds of mathematical results, the statistical physicist Yoshiyuki Kabashima drew a line: AI is far faster on formulated problems, but carving problems out of phenomena and judging what is interesting are different matters. Where AI stands at each stage of asking questions, and how researchers are trying to measure it.
Where has today's AI passed humans, and where has it not? Math olympiads, programming, drone racing, driving, professional work, spatial reasoning and robot hands, sorted using figures from papers and primary announcements. Also: learning from little experience (ARC-AGI-3), self-improvement, and three walls that remain in the physical world. Part 2 of Is There an Upper Limit to AI Intelligence?
If an AI keeps improving itself, how far can its intelligence go? Five questions on the physical limits of computation (about 5.4×10^50 operations per second per kilogram, 10^120 over the history of the universe), on when self-improvement levels off, and on how the compute behind the September 2026 Navier–Stokes result compares. Part 1 of 2.
MIT Technology Review named mechanistic interpretability one of its Breakthrough Technologies of 2026. Work at Anthropic and Google DeepMind is starting to reveal what happens inside large language models once dismissed as black boxes — from circuits that plan a poem's rhyme in advance to chains of reasoning explainable by as few as 200 neurons.
IBM claims a 'verifiable' quantum advantage with 70 logical qubits, Google put reinforcement learning in charge of error correction, and Microsoft/Quantinuum drove logical error rates down to 0.001%. Lining up three very different kinds of 2026 progress to ask what quantum computing actually still needs.
In late 2025 and into 2026, concrete progress arrived in the unglamorous but decisive race to correct quantum computing errors fast enough to matter. A look at where IBM, Fujitsu, and the latest academic research actually stand.
Google and Xcel Energy chose an iron-based battery, not lithium-ion, for a record-setting grid storage project in Minnesota. As lithium-ion dominates short daily cycles, why does a battery built to run for four straight days matter? A look at the efficiency-versus-cost tradeoffs reshaping how grid batteries specialize in 2026.
A New Hampshire man lived 271 days without dialysis after receiving a genetically engineered pig kidney — a record before the organ was ultimately removed. Comparing heavily edited US pig kidneys with China's simpler six-edit approach, this piece traces the real barrier facing pig-to-human organ transplants.
TSMC has begun mass-producing 2-nanometer chips while Samsung struggles with yield, and Japan's Rapidus aims to start production in late 2027. A look at the frontier of chip miniaturization — and the 3D stacking that may come next.
A no-surgery implant threaded through a blood vessel, versus a high-density array that requires opening the skull. In 2026, brain-computer interfaces are entering a real race toward approval — on both safety and performance.
Kyoto Fusioneering's U.S. blanket demonstration facility and Japan's up-to-¥2-trillion prototype reactor plan. A look at how close fusion power came to a genuine "demonstration phase" in 2026.
In 2026, a production line in China began turning out a humanoid robot every 30 minutes. A look at how far Figure AI, Tesla Optimus, and Unitree have come — and the walls still standing in their way.