Geometry ↔ gravity ↔ spacetime
The whole theory rests on one sentence: mass and energy tell spacetime how to curve, and curved spacetime tells mass and energy how to move. The tensor calculus is how you compute that precisely — but almost every idea that makes GR feel like magic can be reached with the geometry, a little algebra, and the physics you already own. This is that route: eight chapters, six things you can play with, and honest signposts to what the indices would add.
Assumed: special relativity (the invariant interval, proper time, γ) · Newtonian gravity & orbits · vectors and calculus. Deliberately skipped: Christoffel symbols, covariant derivatives, index gymnastics. We name them where they'd go so you know what you're deferring, not dodging.
How to use this. Read in order — each chapter earns the next. The six simulations aren't decoration: run each one until the readout stops surprising you. Every equation is written to be read, not manipulated, so don't reach for a pen — reach for the sliders.
By the end you can
Einstein called it the happiest thought of his life: a person in free fall does not feel their own weight. Step off a ledge and, for the duration, gravity vanishes from your experience — you and everything around you fall together, so relative to you nothing falls at all. This is exactly what astronauts exploit: the ISS isn't "beyond gravity" (gravity there is ~90% of surface strength) — it and everyone in it are simply in continuous free fall.
Turn it around. Seal yourself in a windowless box. If you feel pinned to the floor, there are two stories that fit every measurement you can make with a dropped ball: (a) the box sits on a planet, or (b) the box is a rocket accelerating at g through empty space. The equivalence principle is the claim that no local experiment can tell these apart. Uniform gravity and uniform acceleration are not merely similar — they are the same thing described in two frames.
The move that starts the whole theory
If free-fall abolishes gravity locally, then free-fall is the natural, unforced state of motion — the relativistic version of Newton's straight-line coasting. A falling apple isn't being pulled off a straight path; it's the person standing on the ground, held up by the floor, who is being pushed off theirs. What we call "the force of gravity" is the floor's push, felt because the ground stops us from following the free path. Gravity itself is what happens to those free paths when mass is nearby — and that turns out to be geometry.
There is a catch, and the catch is the seed of general relativity. The equivalence is only exact locally — in a small enough box, for a short enough time. Give yourself room and one experiment separates real gravity from acceleration: drop two balls, far apart. In the rocket, the floor rushes up to meet both identically — they stay perfectly parallel. Over a real planet, both fall toward the centre, so two balls side by side drift together, and one above the other stretches apart. That residual — the part of gravity you can't transform away by falling — is called the tidal field, and it is the true fingerprint of spacetime curvature. Try it:
Sim 1 · Which box are you in?
Release a single ball and it falls the same in both boxes — that's the equivalence principle. Then switch on the tidal test to release a whole grid and watch what real gravity does that acceleration can't fake.
Single ball
identical in both
Grid of balls
— run the tidal test —
Verdict
locally indistinguishable
A single falling ball can't tell you which box you're in — acceleration and gravity are locally identical. Only the tidal pattern across an extended object betrays real gravity: the balls fall toward a point, not a direction. That convergence-and-stretch is curvature, made visible.
The tempting picture
"Earth pulls the apple down with a force F = GMm/r². The apple accelerates because a force acts on it, exactly like any other force."
The GR picture
"No force acts on the falling apple — it's coasting along the straightest available path. The ground is the accelerated object, held off its free path by the electromagnetic push of the dirt. 'Weight' is that push, not a pull."
Because if gravity is really just your frame accelerating, the rate of fall can't depend on what is falling — an accelerating floor rushes up to meet a feather and a cannonball identically. Newton had to assume gravitational mass equals inertial mass (an unexplained coincidence tested to ~1 part in 1015 today, e.g. by the MICROSCOPE satellite). In GR it isn't a coincidence at all: there's no "gravitational charge" to match to inertia, because there's no force. Everything follows the same geometry, so everything falls the same. The universality of free fall stops being a puzzle and becomes the definition.
Where this leaves us
Free fall is the unforced motion; "gravity as a force" is a frame artifact you can erase by jumping. What you cannot erase — the tidal convergence of nearby free paths — is real, physical, and observer-independent. General relativity is the theory of exactly that residual. Which means the whole subject reduces to two questions: what are the free paths, and what bends them?
To describe "the straightest path," we first need a notion of straight, of distance, of time — in a spacetime where those can vary from place to place. That's the metric.
By the end you can
Brush-up · what SR already gave you
Special relativity fused space and time into one four-dimensional arena and handed you its one invariant — the interval between two nearby events, the same for every inertial observer:
ds² = −c²dt² + dx² + dy² + dz²
The minus sign is the whole personality of relativity: time enters the Pythagorean sum with the opposite signature to space. From ds²<0 (timelike) you read proper time dτ = √(−ds²)/c — the aging of a clock that travels between the events. ds²=0 is a light ray. ds²>0 is spacelike — no clock, no cause. Hold onto proper time; in GR it becomes the thing free particles quietly maximize.
Write that interval more compactly. Collect the coefficients of each dx^μ dx^ν into a symmetric table of ten numbers, the metric gμν. In flat spacetime and Cartesian coordinates it's just diag(−1,1,1,1) — the Minkowski metric ημν. The entire content of "the spacetime is flat" lives in those constants.
That single generalization — let the metric depend on position — is the entire mathematical leap from special to general relativity. A gravitational field is nothing more than a spacetime whose metric varies. Everything else you'll meet — geodesics, curvature, the field equation, black holes, waves — is squeezed out of gμν(x) by differentiation. When people say GR is "just geometry," this is the object they mean.
The mental model to carry forward
Picture graph paper you can't fully flatten — like the skin of an orange peeled onto a table. Locally, any small patch looks like ordinary flat graph paper (that's the equivalence principle again: locally spacetime is always Minkowski). But you cannot lay down one global grid where straight coordinate lines stay straight and evenly spaced everywhere. The metric is the bookkeeping that tells you, patch by patch, how the local flat ruler relates to your global coordinates. Curvature is the obstruction to ever flattening it.
You've seen the picture: a bowling ball on a stretched sheet, marbles spiraling in. It's not wrong so much as it smuggles in the answer — it uses real gravity (the marble rolls because Earth pulls it down into the dimple) to explain gravity, and it curves only space, leaving out the time part that, as the next chapter shows, does almost all the work. Keep the sheet as a memory aid for "mass makes geometry non-flat," then set it down. The honest object is the metric, and the honest statement of motion is coming next.
Right here, and then everywhere. Because the metric's components change from point to point, ordinary derivatives of vectors stop being coordinate-independent — the covariant derivative repairs this, and its correction terms are the Christoffel symbols Γλμν, built from first derivatives of gμν. Take one more derivative and antisymmetrize and you get the Riemann curvature tensor Rρσμν, the true measure of non-flatness. That's the machinery we're deferring. The payoff of skipping it is that you can still read every result below; the cost is that you can't yet derive them or handle a metric nobody has solved for you. Chapters 3–4 give you the concepts those symbols make rigorous.
Given a metric, which path does a free particle take? The answer is disarmingly simple — and it hides the most counterintuitive fact in the whole theory.
By the end you can
In flat spacetime a free particle moves in a straight line at constant speed — equivalently, it follows the worldline of longest proper time between two events (the twin who coasts ages more than the twin who accelerates around). GR keeps the principle verbatim and only swaps the arena:
Written out, "extremize ∫dτ" becomes the geodesic equation, d²xμ/dτ² = −Γμαβ(dxα/dτ)(dxβ/dτ) — the Christoffel symbols we're skipping are exactly the "apparent gravitational acceleration" your coordinates report. In free-fall coordinates they vanish (the equivalence principle, once more), and the particle just coasts.
Here's the puzzle that the geometry has to solve. Spacetime near Earth is barely curved — the deviations from flat are of order GM/rc² ≈ 10−9. So how does an almost-flat geometry produce a fall as brisk as 9.8 m/s²? The answer is a matter of scale, and it's beautiful. In one second of falling you move a few metres through space, but you move c×1s ≈ 300,000 km through time. Gravity's tiny warping of the time direction, multiplied by that enormous lever arm, is what bends your worldline into what we experience as a plummet. For anything moving slowly compared to light — which is everything you've ever dropped — falling is caused almost entirely by the curvature of time, and hardly at all by the curvature of space. Watch the two pictures line up:
Sim 2 · The parabola is an almost-straight line through spacetime
Left: a tossed ball traces the familiar parabola in space. Right: the very same motion as a worldline in spacetime — height across, c·t up. Drag exaggeration down to true scale and the "parabola" straightens into a line: the bend that is gravity is that faint.
Height risen
—
Distance through time (c·t)
—
Space : time lever arm
—
At true scale the worldline is straighter than anything you could draw — the ball is going as straight as it can through a spacetime whose time axis is very slightly tilted toward the ground. That whisker of tilt, times 300,000 km of time per second, is the "force" that a bathroom scale reads.
The rubber-sheet reflex
"Mass dents space; things roll into the dent. The bending of space is the whole story of why the apple falls."
What the numbers say
"For slow objects the spatial curvature contributes almost nothing to the fall. It's the warping of time — clocks running slower lower down — that the worldline is straightening into. Space-curvature only becomes co-equal for things moving near c, which is why light bends twice as much as a naïve Newtonian photon would."
Where this leaves us
Motion under gravity is free coasting along a geodesic — the longest-proper-time path — and for ordinary speeds that path is set by how the rate of time varies with position. "Clocks run slower where gravity is stronger" isn't a curiosity bolted onto the theory; it is the mechanism of falling. That single sentence will pay off directly when we get to black holes and GPS.
Free particles follow the geometry. But what sculpts the geometry? Time to meet the other half of the sentence — and the most famous equation you're allowed to not fully expand.
By the end you can
Chapter 1 left a clue we can now cash. Two nearby free-fallers drift relative to each other — converging over a planet, stretching top-to-bottom. Neither feels a force; each is on a perfect geodesic. Yet the gap between them accelerates. That relative acceleration of neighbouring geodesics is called geodesic deviation, and its strength is exactly what the Riemann curvature tensor measures. So "curvature" isn't an abstraction — it's the tidal field, the one piece of gravity you could not transform away by jumping. Where spacetime is flat, parallel free paths stay parallel; where it's curved, they don't. Newton's tidal tensor ∂²Φ/∂xi∂xj is the weak-field shadow of Riemann.
The two-way street, precisely
Matter curves spacetime; the curvature guides matter's geodesics; the redistributed matter re-curves spacetime. That feedback is why GR is nonlinear and hard — gravity gravitates, because curvature carries energy that is itself a source of curvature. Newton's gravity has no such loop. This is the deep reason there's no simple "gravitational field lines" picture and why exact solutions are precious.
Einstein's field equation binds the two sides of our founding sentence. Here it is, and here is how to read it without expanding a single tensor:
The left side, Gμν, is pure geometry — the Einstein tensor, a specific combination of the curvature of spacetime, distilled from the metric by two derivatives. It answers: "how is spacetime bent here?" The right side, Tμν, is pure content — the stress-energy tensor, packaging energy density, momentum, pressure, and stress: "what's here, and how is it flowing?" Crucially, energy of every kind sources gravity, not just mass — pressure and even the energy in a field gravitate. The constant 8πG/c⁴ is just the exchange rate, and the c⁴ in the denominator (a colossal number) is why you need a planet's worth of mass to bend spacetime by a noticeable amount. That's the equation: tell me the matter and energy, I'll tell you the geometry; tell me the geometry, Chapter 3 tells you the motion.
What the indices buy — and what they cost to skip
That one tidy line is really ten coupled nonlinear partial differential equations (the symmetric μν pair runs over 4×4, minus symmetry). The tensor notation is what lets you write ten equations as one and guarantee the statement means the same thing in every coordinate system — general covariance, the "general" in general relativity. What you lose by skipping it: you can't solve the equation yourself. What you keep: once a physicist hands you a solved metric — flat, Schwarzschild, FLRW, a gravitational wave — you can read all of its physics using only Chapters 2–3. That's the deal the rest of this page runs on.
You can add one more geometric term the equations permit: Gμν + Λgμν = (8πG/c⁴)Tμν. Einstein put Λ in to hold the universe static, called it his blunder when Hubble found expansion, and then it came roaring back in 1998 as dark energy — the driver of cosmic acceleration. Physically Λ behaves like an energy density of empty space with negative pressure. It's a one-symbol reminder that the field equation had a free knob all along, and the universe turned it on.
Where this leaves us
We now have the complete logical machine: matter → Tμν → field equation → metric → geodesics → motion, looping back because the motion redistributes the matter. Every remaining chapter is this machine run on one solved metric. We start with the most important solution ever found — the geometry outside any spherical mass — and let it take us to the edge of a black hole.
Solve the field equation for the empty space around a single spherical mass and you get the Schwarzschild metric — printed within a year of the theory, from a WWI trench. It contains black holes whether Einstein liked it or not.
By the end you can
Karl Schwarzschild solved Einstein's equation for the vacuum outside a spherical mass M within months of the theory's publication — while serving on the Russian front, months before he died. It describes the Sun, the Earth, a neutron star, and a non-spinning black hole with one line:
Look at where it breaks. The time coefficient (1 − rs/r) hits zero at r = rs and the radial coefficient blows up there — that surface is the event horizon. It also blows up at r = 0 — the singularity. A century of work established the crucial difference between them: the singularity at r=0 is a real, curvature-diverging pathology; the "singularity" at r=rs is a fake — an artifact of these particular coordinates, like the way all meridians crowd together at the North Pole on a map. Nothing locally dramatic happens as you cross the horizon. What makes it a horizon is global: past it, every future-pointing path leads inward. Not even light gets out.
Sanity-check · the Newtonian coincidence
Set the Newtonian escape speed to c: ½c² = GM/r gives r = 2GM/c² — exactly rs. This is a genuine coincidence (the Newtonian derivation has no business being right here), but it's a handy mnemonic: the horizon is where the escape velocity reaches light-speed. Numbers: the Sun's rs is ~3 km; Earth's is ~9 mm. Both sit harmlessly far below their surfaces — you need to crush the mass inside rs to make a black hole.
The time coefficient does something you can feel. A clock at rest at radius r ticks, relative to a far-away clock, at the rate dτ/dt = √(1 − rs/r). Deeper in the well, time literally runs slower — the very effect Chapter 3 said is gravity. Light climbing out of the well loses energy and reddens by the same factor (gravitational redshift). As r → rs, the factor → 0: to a distant watcher, an infalling clock ticks ever slower and its light reddens toward black, appearing to freeze at the horizon and never quite cross. Play the well:
Sim 3 · Two clocks, one gravity well
A clock sits at radius r outside a mass; a reference clock sits far away. Slide inward toward the horizon and watch the deep clock lose time and its light redden. At r = rs it stops.
Clock rate (deep ÷ far)
0.816
1 year here = far away
1.22 yr
Redshift z of escaping light
0.22
Time dilation and redshift are the same factor √(1 − rs/r) seen two ways — as a tick-rate and as a colour. Both come straight from the time term in the metric. This is not science fiction: your phone's GPS corrects for exactly this effect every second.
Worked anchor · why GPS would fail in minutes without GR
GPS satellites orbit at ~20,200 km, higher in Earth's (very shallow) well than the ground, so their clocks run faster by gravitational time dilation: +45 μs/day. Special-relativistic time dilation from their orbital speed subtracts −7 μs/day. Net: satellite clocks gain ~38 μs/day on ground clocks. Light travels ~11 km in 38 μs — so an uncorrected system would smear your position by ~10 km within a day and keep drifting. GPS receivers bake in the GR correction. General relativity is in your pocket.
Locally, no — for a big enough black hole the crossing is uneventful; spacetime there is smooth and the tidal (curvature) forces can be gentle. The "freeze" is entirely in the distant observer's picture, caused by the light-travel and time-dilation factor going to zero. The infaller sails across in finite proper time. What's inescapable isn't a wall at the crossing but the future itself: inside, r=0 becomes a moment in time you cannot avoid, the way tomorrow is. Tidal forces then grow without bound as you approach it (for a stellar-mass hole, fatally, long before). That asymmetry — smooth crossing, doomed interior — is pure Chapter 2: the horizon's trouble was bad coordinates; the singularity's trouble is real curvature.
Clocks and colours are the "at rest" physics. Let something orbit this metric and GR writes a signature Newton never could: the orbit refuses to close.
By the end you can
Newton's great triumph was the closed ellipse: an inverse-square force sends a planet around a path that repeats forever. Write the orbit as u = 1/r versus angle φ, and Newton gives d²u/dφ² + u = GM/L². General relativity adds one term, from the same Schwarzschild metric:
That tiny term resolved a 200-year-old scandal. Mercury's orbit precesses by 574″ per century; Newtonian tugs from the other planets explained all but 43 arcseconds, and nobody could find the missing pull (they invented a planet, "Vulcan," that wasn't there). In 1915 Einstein computed the leftover from his new term and got 43″/century — no free parameters. He wrote that he was so excited he couldn't work for days. Turn the correction up from planet-tiny to black-hole-huge and the rosette becomes unmistakable:
Sim 4 · Newton closes · Einstein precesses
A test mass orbits. With the GR term off (Newton) the ellipse retraces itself forever. Switch it on and the orbit's axis creeps forward each pass — a rosette. Crank relativity up to see Mercury's whisper become a spirograph.
Precession / orbit
—
Orbit shape
closed ellipse
Real-world analog
—
The same 3GM u²/c² term that nudges Mercury by 43″ a century whips a star near the Milky Way's central black hole through a visible rosette — a precession the GRAVITY collaboration measured for the star S2 in 2020, confirming GR at the galactic centre.
Radii the metric hides
Push closer and Schwarzschild geometry reveals structure Newton never had. At r = 1.5 rs lies the photon sphere, where gravity bends light so hard it can orbit — the ring you see silhouetted in the Event Horizon Telescope images of M87* and Sgr A*. For ordinary matter, orbits become unstable below the innermost stable circular orbit (ISCO) at r = 3 rs; inside it, matter spirals in rather than circling. The ISCO sets the inner edge of accretion disks and the frequencies we detect when black holes feed. None of these radii exist in Newtonian gravity — they're carved by the same metric that dilated the clocks.
Geometry, and it's worth seeing why. In flat space, "parallel-transport" a direction vector once around a closed loop and it comes back pointing the same way. In curved space it comes back rotated — that rotation-per-loop (holonomy) is another face of the Riemann curvature from Chapter 4. The orbit's slow turn is that holonomy accumulated over each revolution. So perihelion precession, the bending of starlight, and geodesic deviation are the same geometric fact — curvature — showing up in three costumes. This is the kind of unification the tensor formalism makes exact; here you get to see it as a family resemblance.
Light didn't get its own chapter yet, and it deserves one — because light near a mass does something a Newtonian photon can't, and it's how the whole theory was first confirmed.
By the end you can
A ray of light grazing a mass M at closest distance b is deflected by
Chapter 3 is the reason for the two. Slow objects fall almost entirely due to time-curvature (the "Newtonian" half). Light moves so fast that the spatial curvature — negligible for a thrown ball — contributes an equal second half. Newton gets only the time part; Einstein gets both, so exactly double. In 1919 Arthur Eddington sailed to measure star positions during a total eclipse, when stars near the Sun's limb become visible. Newton predicted 0.87″ of shift; Einstein predicted 1.75″. The stars had moved the Einstein amount. It made the front page of every newspaper and turned Einstein into the most famous scientist alive overnight.
When the alignment is good and the mass is large, deflection stops being a nudge and becomes a lens: a galaxy behind a galaxy cluster gets smeared into arcs, doubled into multiple images, or — with perfect alignment — wrapped into a complete Einstein ring. Bend the rays yourself:
Sim 5 · A mass is a lens
Parallel rays from a distant source stream past a mass and bend by α = 4GM/c²b — more sharply the closer they pass. Drag the mass up and watch them focus. Move the source behind the lens to split it into two images, then merge them into an Einstein ring.
Einstein radius θE
—
Images seen
two
Total magnification
—
One background source, two (or more) images, brightened — because the intervening mass bent multiple light paths into your eye. Lensing is now a workhorse: we weigh galaxy clusters, map invisible dark matter, and spot the most distant galaxies ever seen through the magnification of a cosmic lens.
Worked anchor · lensing as a scale
The Einstein radius grows as θE ∝ √M, so measuring the ring's size weighs the lens — including the mass you can't see. When clusters like the Bullet Cluster are lensed, the inferred mass sits mostly where the galaxies and gas aren't, one of the cleanest fingerprints of dark matter. And because a lens magnifies, it acts as a natural telescope: JWST routinely uses foreground clusters to catch galaxies from the first few hundred million years after the Big Bang that would otherwise be too faint to resolve.
Static masses bend light and slow clocks. But GR's boldest promise was that gravity itself can radiate — that spacetime can ripple. It took a century to catch one.
By the end you can
The field equation is nonlinear, but far from any source it has small wave solutions: ripples in the metric itself, travelling at c. A passing gravitational wave doesn't push matter around like sound in air — it changes the distances between free-falling objects. As the wave goes by, a ring of test masses is stretched along one axis and squeezed along the perpendicular one, then the reverse, over and over. That's the wave's "+" polarization; rotate the pattern 45° and you get the "×" polarization. There's no "up-down" wobble — gravity's quadrupole nature means it works by shearing. Watch a ring breathe:
Sim 6 · A ring of free masses in a passing wave
A gravitational wave travels toward you (out of the screen). It stretches the ring one way as it squeezes the other, in alternation. Flip between the two polarizations, and note: nothing is pushed — the distances themselves are changing.
What's oscillating
distance between masses
Real LIGO strain h
~1 × 10−21
Shown here, exaggerated by
~1020×
The ring here deforms by tens of percent so you can see it. A real wave from two merging black holes deforms LIGO's 4-km arms by about 1/10,000th the width of a proton — strain h ~ 10−21. That we detect it at all is one of the great experimental feats in physics.
Worked anchor · GW150914, the first catch
On 14 September 2015, LIGO's two detectors (Louisiana and Washington, 3,000 km apart) registered the same rising "chirp" 7 milliseconds apart — a signal sweeping up in frequency and amplitude over 0.2 seconds, then ringing down. It matched the GR waveform for two black holes of ~36 and ~29 solar masses spiralling together and merging into one of ~62, radiating ~3 solar masses of energy as gravitational waves in a fraction of a second — briefly outshining, in power, all the stars in the observable universe combined. It confirmed both gravitational waves and the existence of binary black holes, and it opened an entirely new way to observe the universe: not with light, but by listening to spacetime itself. Nobel Prize, 2017.
Why it took 100 years
Because c⁴/8πG — the stiffness of spacetime from Chapter 4 — is enormous. Spacetime is the most rigid medium there is; only cataclysms (merging black holes and neutron stars) shake it detectably, and even then only to one part in 1021 by the time the ripple reaches us. Einstein himself doubted the waves were real for decades, unsure whether they carried energy or were a coordinate illusion. They carry energy. We now catch one every few days.
That's the tour — the founding sentence, cashed out eight times over. What did we skip, and where do you go if you want the full machinery back?
The whole theory in one breath
Free fall is unforced coasting, so "gravity as a force" is a frame you can jump out of — but the tidal residual you can't jump out of is curvature. Spacetime carries a metric, a ruler-and-clock that varies with place; free things follow its geodesics, the longest-proper-time paths, and for slow motion that path is set by how the rate of time bends — which is why apples fall and GPS needs correcting. Einstein's equation ties the geometry to the matter and energy that source it. Run that one machine on the geometry outside a mass and out drop time dilation, event horizons, precessing orbits, bent starlight, and — from the whole thing shaking — gravitational waves. Every one of those is confirmed. You reached all of it without expanding a single index.
WHAT THE TENSORS WOULD GIVE YOU BACK →
The right to compute, not just read. Christoffel symbols to find geodesics in any metric; the Riemann tensor to quantify curvature and prove two spacetimes genuinely differ; the covariant derivative that makes "rate of change" coordinate-independent; and the ability to actually solve the field equation for a case no textbook hands you. Everything on this page was a solved metric read with undergrad tools. The formalism is how those metrics get solved in the first place — and it's the natural next step, now that you know what each symbol is for.
A. Einstein, "Die Grundlage der allgemeinen Relativitätstheorie" (Annalen der Physik, 1916) — the field equation and the theory's foundation; the Mercury-precession and light-bending predictions.
K. Schwarzschild (1916) — the exact vacuum solution outside a spherical mass, derived on the Eastern Front; the metric behind Chapters 5–6, horizons and all.
Dyson, Eddington & Davidson, the 1919 eclipse expeditions (1920) — measured starlight deflection at the Sun's limb at the Einstein value (1.75″), not the Newtonian half; the first confirmation and the origin of Einstein's fame.
Pound & Rebka (1959) — gravitational redshift measured in a Harvard tower to ~1%, confirming the time-dilation factor of Chapter 5 in a lab.
LIGO Scientific & Virgo Collaborations, "Observation of Gravitational Waves from a Binary Black Hole Merger" (GW150914) (Phys. Rev. Lett., 2016) — the direct detection of gravitational waves; Chapter 8's anchor. Nobel Prize in Physics 2017.
Event Horizon Telescope Collaboration (2019, M87*; 2022, Sgr A*) — resolved images of the photon-ring/shadow around supermassive black holes; the photon sphere of Chapter 6 made visible.
GRAVITY Collaboration (2020) — measured the Schwarzschild precession of the star S2 orbiting the Milky Way's central black hole; Chapter 6's precession at the galactic centre.
MICROSCOPE mission (2022) — tested the universality of free fall (equivalence principle) to ~1 part in 1015; the foundation of Chapter 1, still unbroken.