ION FLOW — WHERE A LITHIUM ION CAN ACTUALLY GO A solid-state battery works only if an ion can move through a solid. Whether it can is a question about geometry: is there a connected path through the crystal wide enough to pass, and how much energy does the tightest point on that path cost? Aufbau now answers it. Open Tools > Ion Flow. Pick one of the two bundled electrolytes — NASICON, the lithium titanium phosphate, or cubic garnet LLZO — and the app maps every position a lithium ion could occupy inside the real crystal, finds the connected channels, and traces the cheapest route through. You get the crystal in 3-D with the conduction channel drawn through it: the framework atoms in one colour, the highway the ion travels in another. Turn on labels and every atom names itself. Scroll to zoom, drag to rotate, and the view tiles the lattice so you are looking at a piece of material rather than a diagram. Underneath is bond-valence site energy, which computes the cost of putting an ion at a point from the geometry of the surrounding atoms. No density functional theory, no fitted machine-learned potential, nothing trained. It runs in seconds on a laptop, which is the point: it is a screening tool meant to sort candidates before anyone runs an expensive calculation or makes anything. It also tells you when it cannot answer. Load your own structure from the Crystallography Open Database or the Materials Project and the app either computes, or explains why not — whether the file is missing information it needs, or the chemistry is outside what it has parameters for. One case it used to get quietly wrong now refuses outright: a sulfide, scored with a potential fitted only to oxides, produced a plausible-looking number that was meaningless. It returns nothing instead. Where a structure contains an element it has no parameters for, it says so, and warns that the barrier it reports will read optimistically low. And it shows you its own error. Beside each computed barrier the app prints the measured range for that material. For the two bundled electrolytes it reads high — LLZO at 0.78 eV against a known 0.2 to 0.3, NASICON at 0.92 against 0.3 to 0.6 — and the screen says plainly that the method overestimates. What it gets right is the comparison: LLZO ranks as the better lithium conductor, which is the answer a screening tool exists to give. We would rather show a number that is honestly too high, next to the number it should be, than round it quietly toward the literature. This is the same engine that computes bonding, reactions and molecular shape elsewhere in the app, now pointed at extended solids. Nothing about the chemistry it already did has changed.

LIGAND FIELD — A TEXTBOOK ORDERING, COMPUTED Put six ligands around a transition metal and its five d orbitals stop being equal: two point straight at the ligands and are pushed up, three point between them and are not. The size of that gap decides a compound's colour, its magnetism, and much of its reactivity. Ranking ligands by it gives the spectrochemical series — one of the oldest orderings in inorganic chemistry, and for a century something you looked up rather than derived. Aufbau now computes it. Open Tools ▸ Ligand Field, pick a metal, and the engine builds six octahedral complexes, places each ligand at the sum of the covalent radii, solves them, and ranks the ligands by the splitting that comes out. Nothing in that path is fitted to the answer: the orbital parameters are a published table, the bond lengths are radii sums, and the series itself is never given to the engine. It is used afterwards, as a check. For the early transition metals six of the nine ligands come out in the correct order, including the whole weak-field end. The screen marks each ligand against where experiment puts it, and shows the score — because copper manages three, not six, and that is displayed just as plainly. Scandium's d level sits above every ligand level, which the model handles; copper's sits below most of them and the interaction inverts. The real series is nearly the same for every metal, so that drop is a genuine limitation and the screen says so. Back-bonding falls out of the same calculation. Carbon monoxide gives its electrons to the metal and takes some back into an empty orbital of its own — and nothing was added to the model to make that happen, because CO's accepting orbital is simply one of CO's own. A bare atom, which has no such orbital, reports exactly zero rather than a small number. Underneath, three corrections to the engine reach well beyond this screen. Valence orbitals for the heavier elements had been built with the shape belonging to the second row, which left iodine's outer shell about four times too compact — every element from sodium down was affected. The transition metals' orbital energies were derived where a measured table exists, with roughly two and a half times the correct slope across the row. And six of the nine first-row metals had no covalent radius at all, so they could not be placed in a structure. The benchmark is unchanged at 49 of 49.

EVERY WINDOW RESIZES NOW — AND THE 3-D VIEWS GREW WITH THEM Aufbau's windows have never been resizable on the Mac. Every one opened at 924 by 694 points and stayed there, whatever the size of the display in front of it. Dragging a corner did nothing. Neither did the Window menu's own Fill command. That is fixed: windows resize, they keep the size you leave them at, and the 3-D labs — orbital surfaces, Bloch spheres, molecule viewers — can be opened as large as the screen allows. The same limit was quietly shrinking the panels that open on top of a window. "Your molecules in 3-D" asked for a thousand points of width and was given about five hundred and fifty, so a thirty-eight atom molecule — porphine, say — was drawn into a box about a fifth of the window it sat in. Those panels now use the window they are in. Opened full-screen, the molecule viewer is roughly twice the size it was, and the aromatic clouds above and below a ring read as rings rather than as suggestions. The Reactor had the same problem in a worse form: on a 1080p display its spontaneity and kinetics readings sat below the bottom edge and could not be scrolled to. They fit now. One piece of text is fixed with them. The Design screen's opening line had been printing its own formatting marks — literal asterisks around a word meant to be bold — since that screen shipped. One chemical result does change. Building a porphyrin from pyrrole and formaldehyde ended with an oxidation that took the ring too far: it pulled hydrogen off all four nitrogens and produced C20H12N4, where porphine is C20H14N4 and keeps two N–H. The macrocycle was always right; the last step overshot it. Oxidation now stops when the ring becomes aromatic, which is what drives it, instead of continuing while a wider circuit can still be traced — and the full route, pyrrole and formaldehyde through porphyrinogen to porphine, is now checked on every build rather than demonstrated on request. A crash is fixed, and it is worth naming because it could end your session. Asking for a route to a target made of more than one piece — an ionic reagent, a mixture you typed, a saved reaction you recalled — could quit the app outright. The planner was checking whether the molecule had come apart in two, when the question it needed to ask was whether the bond it had just broken was the thing that separated them. On a ring bond in a two-part mixture those are different questions, and it reached for an atom that was not there. Routes also name what they want. The Workbench has always had a shelf of common reagents, and a finished route would hand one of them back as a bare formula — you were shown HCl where the shelf beside you calls it hydrochloric acid. The route now uses the same names the shelf does. And a leaf the planner merely stopped at because it was small no longer calls itself a building block: it says it is not a stocked reagent, because telling you to order hydrogen thioperoxide is worse than telling you nothing. Nothing else about the chemistry moved. The benchmark is the same 49 of 49. Beyond that one correction this release changes only whether you can see what the engine computes — which on a large display turned out to be a real limit rather than a matter of taste.

AROMATIC RINGS NOW SHOW THEIR ELECTRON CLOUDS — EVERYWHERE The blue clouds above and below a benzene ring are what make it read as benzene rather than a hexagon of lines. They appeared in one place. Now they appear wherever a molecule is drawn in 3-D, including Sculpt, where the shape is worked out from scratch and a ring can end up at any angle — the cloud follows the ring's real plane rather than assuming it lies flat. Porphyrin is the molecule that showed this was worth doing properly. It had been drawing four huge overlapping rings across itself, because the code that decides which rings to outline asked whether one ring sat inside another. A porphyrin's outer circuit threads through its four inner rings rather than containing them, so nothing was ever ruled out and one circuit got drawn four times. It now shows what it is: one ring around the outside, and the two inner rings that are aromatic on their own. RETROSYNTHESIS TELLS YOU WHAT TO ORDER A route used to end at "C7H6O3 — buy", which asks you to work out what that is before you can order it. It now says salicylic acid, beside the formula rather than instead of it — the formula is what the engine computed, the name is the identity it is assuming, and you can see both and disagree. Aspirin now comes apart into salicylic acid and acetic acid, in those words. ALSO • The lessons run 1 to 7. "Why Molecules Bend" always was Lesson 2 — it sits second in the menu and Lesson 1 hands off to it — but its title never said so, so the course looked like it skipped a number. • A leaf that cannot be bought no longer says "aromatic", which is a description rather than an instruction. It says what the engine actually did: build onto this ring. • The building-blocks list separates what you can order from what the engine could not take apart, instead of listing both together.

LESSONS THAT DERIVE INSTEAD OF ASSERT Aufbau now teaches. Seven short lessons, each working a question out on screen instead of stating the answer — and every number in them computed live by the same engine that answers everywhere else in the app. Why molecules bend is the clearest. Put the things around an atom on a sphere, let them push each other apart, and read off the shape. Methane, ammonia and water come out of one calculation. Hybridization is never mentioned, because it is never needed. Turn the lone-pair dial to 1.00 and water straightens out — the bend was doing work, and now you can see how much. Others walk through why a mixture will not simply react, what a catalyst does and does not change, and how to plan a synthesis backwards. No language model is involved and nothing is scripted. Ask a lesson something it cannot compute and it declines. PROTECTING GROUPS The part of route planning that was missing. Aufbau now notices, by itself, that a step it wants would attack the wrong site — the pieces do react, they just build the wrong molecule — and answers the way a chemist does: mask the site that keeps winning, run the step, take the mask off. It also refuses a protecting group whose removal would destroy what you are making, and reaches for a second one that survives instead. That decision comes from running the removal and reading the result, not from a table of which group suits which case. ALSO • Reaction energies are more accurate. Two errors that sat recorded in the test suite for a year are now half the size — esters no longer fall apart into aldehydes, and methane combustion is 45 kJ/mol closer to the measured value. • Chlorinating benzene gives chlorobenzene. It had been destroying the aromatic ring instead of substituting into it. • Answers no longer depend on the order you list the reactants.

Ask Aufbau how to make something and it works backwards — breaking one bond at a time and keeping only the disconnections it can confirm by running the reaction forwards again. Until now those pieces came back as fragments no one can buy. They are molecules now. Aspirin comes apart into salicylic acid and acetic acid — the route industry actually uses. Ethyl acetate gives its alcohol and its acid. An amide gives its acid and its amine. And rings open. A lactone, a lactam, a cyclic ether — Aufbau opens the ring to the precursor that closes back on itself. Vitamin C, THF and butyrolactone all disconnect now where before they returned nothing. There is no library of reaction rules behind this. Aufbau caps the broken bond the three ways a bond can be capped and asks its own forward engine which one rebuilds your target. When nothing does, it says so: butane and toluene come back marked "no route found", because the chemistry to make them needs reagents this engine does not model. A dead end is now labelled as one instead of being listed as a building block. ALSO • The reaction benchmark reaches 49 of 49 — the last row won by supplying the driver esterification actually needs, taking the water off as it forms, rather than by loosening a threshold. The stricter scoring that once took it down to 46 was never reversed. • Bond solving is about three times faster on difficult mixtures • Sculpt now reports how far bonds drift when you force a shape, so a hit target cannot hide a broken molecule

Aufbau now knows that some reactions need a catalyst — and refuses them without one. The aldol reaction joins two molecules by making a new carbon–carbon bond. Aufbau could not predict it, and the reason was not that it scored the reaction badly. It was that the engine had been taught never to join two carbon skeletons — a rule added to stop it inventing products that cannot be made. The rule was right. It also made this real reaction unreachable, and no amount of adjustment would have found it. What makes the bond possible is a base. It removes a hydrogen from the carbon beside the carbonyl group — a hydrogen that is only loose because of what sits next to it — and the carbon left behind reaches out to a second molecule. Aufbau now models exactly that, and only that. Add a base and it predicts the product; leave the base out and it tells you nothing happens, which is what actually happens. Two things followed that we did not put there. Warm the reaction and it reverses itself, because joining two molecules into one costs disorder. And the reactions that should never fire still never fire, base or no base. Also in this update: adding chlorine across a double bond now puts one chlorine on each carbon instead of both on the same one — the two arrangements contain identical bonds, so only the mechanism could tell them apart. Coordination results no longer depend on the order donor atoms were listed, and rare-earth separation declines a pair it cannot rank rather than guessing. The curated benchmark moves from 46 of 49 to 48 of 49. The one reaction still missed, esterification, is missed for a reason worth stating: it is exactly thermoneutral, and a catalyst cannot make a thermoneutral reaction go. It needs its water removed, which is a different thing entirely.

Molecular geometry now converges. Aufbau finds a molecule's shape by relaxation — nudging atoms downhill, repeatedly, until they stop moving. The step size is supposed to decay so the structure settles. It had a floor. Past roughly 458 iterations the steps stopped shrinking, and the molecule was no longer settling; it was walking, at a fixed stride, for as long as the loop ran. It walked deterministically. The same input gave the same shape every time, on any machine — so nothing looked wrong, and nothing was reported wrong. But the shape you got was wherever the walk happened to be standing when the loop ended, not where the molecule wanted to be. Run the same crown ether longer and its size wandered between 3.03 and 3.30 Å, up and down, converging on nothing. The schedule now cools all the way. That same molecule reads 3.0916 Å whether you relax it for 1,200 steps or 24,000, and across a 32-molecule test corpus spanning water to 21-crown-7, every structure settles. Shapes throughout the app shift very slightly — this is the corrected value, not a new approximation. One answer got worse: converged, hydrazine now comes out anti and disagrees with experiment — it had been right only because the old relaxation stopped just before it changed its mind. The full account is on the Research page at aufbau.app.

Aufbau's claim is that chemistry can be computed rather than looked up. This release spends most of its effort holding that claim to account — then hands you a tool for pushing back on the answer. CHECKED AGAINST MEASURED REALITY Aufbau now ships with an extensive physics and chemistry test suite that compares what the engine derives against published measurements: reaction enthalpies, standard entropies, nuclear binding energies, and stereochemistry worked out by hand from the CIP rules. Ammonia synthesis comes out at −93 kJ/mol against a measured −92; hydrogen combustion at −482 against −484. Just as important, the suite pins the places where the model is wrong, at the size it is wrong — so they cannot drift unnoticed or be quietly overstated. Where a simple model reaches its limit, Aufbau says so rather than rounding it off. MORE ACCURATE PHYSICS • Temperature and pressure now act on ordinary chemistry, not a short list of molecules. Entropy used to come from a small table, so any reaction touching a molecule outside it silently ignored the temperature dial. Aufbau now derives it from first principles — molecular mass, the moments of inertia of its own computed geometry, and its own vibrational frequencies. Ammonium chloride now forms in the cold and comes apart when heated, the way it really does. • Your answer no longer depends on the order you typed the reactants. • The engine refuses what it cannot know. Type an element outside its data and it declines and names it, rather than substituting a stand-in and reporting a confident result. And when a mixture will not react, it tells you what it actually searched instead of claiming your reagents are stable. • Molecules are told apart by structure, not just formula — ethanol and dimethyl ether are no longer the same substance, and a shared reaction keeps its stereochemistry. INTRODUCING SCULPT Until now Aufbau computed a molecule's shape and showed it to you. Now you can push back — reach into the geometry, ask for the shape you want, and see what the chemistry says about it. • Sculpt. Add a virtual charge group to any atom and watch the bonds bend away from it, the way a real lone pair would. Say what you need — this angle at 109.5°, these atoms 2.4 Å apart, that torsion anti — and Sculpt drives the geometry toward it, then reports the miss rather than hiding it. Where a target fights the chemistry, the chemistry wins and it tells you so. Wrap a ligand around a metal ion, settle a conformation with molecular orbitals, or search ring puckers for the lowest-strain arrangement. • Why Molecules Bend — the shape rule in two counters and a dial. No hybridization, no sp³. Put things on a sphere, let them push apart, read off where the bonds point. Hold four domains and step the lone pairs 0 → 1 → 2: methane, ammonia and water appear at 109.5°, 106.2°, 103.5°, against measured 109.5, 107 and 104.5. One number does the bending, and it is a slider. • Bonds that hold their length. A rewritten geometry core treats bonds as constraints, not springs: in an eighteen-membered ring, bond-length error dropped from 0.216 Å to 0.0001 Å. • Also: a larger bond search that finds the better arrangement when you build from loose atoms; damaged reaction libraries are kept and reported instead of silently reset; oversized mixtures say so immediately instead of appearing to hang; and a tidier Tools menu.

This update turns Aufbau into a separation designer — it works out which ligand pulls which metal apart, and plans how to take a rare-earth mixture apart, all from first principles. • Which ligand grabs which metal. A new Coordination Selectivity tool ranks donor atoms — oxygen, nitrogen, sulfur, phosphorus, the halides — by how strongly each binds one metal ion over another. It's the hard–soft acid–base rule, derived from the ions' own frontier orbitals and charge density instead of a lookup table. Pick a target metal and the matrix it's mixed with, and it draws the split as a diverging bar chart: hard oxygen donors reach for hard ions, soft sulfur and phosphorus for soft ones, and the ligand field breaks the tie between close neighbours like copper and nickel. • A separation plan for the rare earths. The new Rare-Earth Flowsheet tool takes a mixture of lanthanides and lays out how to pull them apart — first stripping the few with a cheap one-step redox handle (cerium, europium, ytterbium, because their +4 or +2 ion lands on an empty, half-filled, or full f-shell), then a contraction-based chelation train for the stubborn trivalent remainder, flagging exactly where the separation is tightest. Everything computed deterministically from electron configuration — the same mixture, the same plan, every time.

This update teaches Aufbau pericyclic reactions — and shows you the orbital symmetry that decides them. • See why the Diels–Alder goes. A new Pericyclic tool works out whether a cycloaddition or a ring-closing electrocyclization is allowed — the Woodward–Hoffmann rules — by reading the answer off the molecule's own frontier orbitals, the way a chemist actually derives it. It draws those orbitals as phase-coloured lobes and marks each forming bond green when they overlap in phase, red when they clash. The classic Diels–Alder comes out allowed; the [2+2] forbidden; flip to photochemical (hν) and the rules invert — because the orbitals do. It's a symmetry argument no lookup table is doing for you. • Chiral sulfur, too. Aufbau's stereochemistry now recognises the three-dimensional handedness of a sulfoxide's sulfur — the kind of stereocenter that makes garlic's alliin a single form — assigning it R or S right alongside carbon. Everything computed deterministically from physical law — the same molecule, the same answer, every time.

This update teaches Aufbau stereochemistry — the three-dimensional handedness that can make one molecule a medicine and its mirror image inert. • See handedness in 3-D. Enter a molecule with defined stereochemistry — an amino acid, a sugar — and Aufbau shows it as it truly is: the (R)- and (S)- forms come out as genuine mirror images, each stereocenter labelled R or S and each double bond E or Z. In the Chemistry Workbench, tap "See stereochemistry in 3-D" and rotate it. • Derived, not looked up. Those R/S and E/Z labels come from the Cahn–Ingold–Prelog rules the same way a chemist assigns them by hand — ranking each center's four substituents from first principles — and the 3-D geometry is computed to match, so the handedness you see is real, not decorative. • Carried through a reaction — honestly. When a reaction leaves a stereocenter alone, Aufbau shows its configuration is preserved. When the center actually reacts, it says the outcome is "not determined" instead of guessing — because deciding it would need geometry the engine doesn't model. As always, Aufbau tells you where it's sure and where it isn't. Everything computed deterministically from physical law — the same molecule, the same answer, every time.

A small update with a taste of real chemistry. • Allicin from garlic. Your library now opens with a worked example: two allylsulfenic acids — the fleeting species a clove of garlic makes the instant you crush it — condensing into allicin, the pungent compound behind garlic's smell and much of its chemistry. As always, Aufbau derives the product from physical law, not a lookup. • Two new reagents on the shelf. Diallyl disulfide and allylsulfenic acid join the Chemistry Workbench stockroom, each with a verified structure, so you can build the garlic route yourself — or discover your own. Deterministic, reproducible, honest — now with a little more to explore.

This update teaches Aufbau catalysts — and derives what each metal does from its place in the periodic table, not from a lookup. • A catalyst you can pick. In the Chemistry Workbench, choose a metal catalyst and watch it change the outcome. Over copper, ethanol becomes acetaldehyde; with no catalyst, it becomes ethylene. Over iron, nitrogen and hydrogen finally make ammonia — a reaction that simply won't go without a catalyst, and now honestly doesn't. • A nudge when the catalyst is in the wrong place. Put iron in with the reactants and the reaction stays stubbornly inert — because a catalyst isn't consumed, it doesn't belong among the reagents. Aufbau now catches the mix-up and offers, in a single tap, to move the metal to the catalyst slot — so nitrogen and hydrogen go on to make ammonia, exactly as they should. • Derived from the electron configuration. Copper's filled d¹⁰ shell is what makes it a gentle, selective catalyst; iron's open d-shell is what lets it split the nitrogen triple bond. Aufbau computes those configurations and reads the chemistry off them — the catalyst's behavior follows from first principles, the same way everything else here does. • A more faithful periodic table. Fixing the catalysts meant fixing the configurations: chromium, copper, silver and gold now show their true anomalous ground states (copper is 3d¹⁰ 4s¹, not the textbook-filling 3d⁹ 4s²), from an exchange-energy rule — while tungsten, correctly, is left alone. Every result is computed from physical law: deterministic, reproducible, and honest — including where a simple model reaches its limits.

This update teaches Aufbau to weigh temperature and pressure. Reactions are now judged by free energy — not bond strength alone — so heat and pressure become part of the chemistry. • Temperature drives reactions. Set the temperature in the Chemistry Workbench and Aufbau weighs entropy alongside bond strength (ΔG = ΔH − T·ΔS). A reaction powered by a small molecule escaping now switches on above a crossover temperature — heat ethanol and the engine works out, on its own, that it becomes ethylene and water. • Pressure drives them too. Turn up the pressure and Le Chatelier takes over. Ammonia synthesis is the classic case: favorable when cool, forbidden when hot — but raise the pressure and it switches back on, exactly why the real Haber process runs at hundreds of bar. Reactions that release a gas are pushed the other way. • Honest at every condition. When nothing is favorable, Aufbau still says "no reaction" — and a mixture that shouldn't react stays inert whatever the temperature or pressure. It never invents a reaction to look clever. Every result is computed from physical law: deterministic, reproducible, and honest when the physics can't decide.

This update sharpens the Chemistry Workbench and the computed periodic table. • Reagent amounts update instantly. Tapping − or + on a reagent now updates its count — and the running atom total — immediately, instead of waiting until you test the reaction. • A more faithful periodic table. The computed electron configuration is now correct for the heavy elements — oganesson and its neighbors fill their shells exactly as they should — and the pattern continues cleanly past the 118 known elements. Every result is still computed from physical law: deterministic, reproducible, and honest when the physics can't decide.

This release lets Aufbau answer the question every chemist actually asks: how do I make this? • Find the missing reactant. Mark a reactant as a wildcard — an element with a question mark, like N? for "a nitrogen source" — name the product you want, and Aufbau works backward to the reagent you're missing. It reasons to the answer (deriving, for example, that urea from CO₂ needs two ammonia, with water leaving) and then proves it against the real reaction route. • Honest, even when the answer isn't on hand. Every suggestion is validated by the engine, and the near-misses are shown with what they'd actually make instead. And if the source you need isn't among your reagents, Aufbau names it anyway — turning "no" into "here's what to get." Every result is still computed from physical law: deterministic, reproducible, and honest when the physics can't decide.

This release lets you see the whole route a reaction takes — and share your work. • Reaction pathways. The Chemistry Workbench now shows the step-by-step route from your reagents to the products — each step's mechanism and its energy change — and when a mixture doesn't react, it explains why (how close it came, and what held it back) instead of just showing "no reaction." • A new family of chemistry. Aufbau now computes amide, urea, and carbamate formation from first principles — including making urea from carbon dioxide and ammonia, the real industrial route, derived rather than looked up. • Share & import reactions. Every saved reaction now carries a portable, self-verifying share code. Send one to a colleague and they can import the exact reaction — reproduced bit-for-bit and checked against the engine. Every result is still derived from physical law: deterministic, reproducible, and honest when the physics can't decide.

This release teaches the Chemistry Workbench to ....predict reaction products from first principles... — not look them up. • Reaction prediction. Enter reactants and Aufbau now computes the products from bond energetics, including many reactions that previously showed "no favorable reaction." Each prediction shows the energy change (ΔH) and is clearly labeled as computed. • Much broader chemistry. New coverage across organic reactions (esterification, amidation, substitution, aromatic substitution, additions, hydrogenation, radical chlorination, carbonyl reduction, ether synthesis), acid–base neutralizations and salt formation, ammonium salts, and hypervalent sulfur & phosphorus chemistry — SO₂, SO₃, sulfuric acid, and phosphoric acid. • Honest by design. When no favorable reaction exists, Aufbau says so rather than inventing a plausible-but-wrong product. Every result is derived, never retrieved.

This update makes the chemistry more accurate and the Chemistry Workbench faster to use: • Reagent Shelf — a stockroom of 26 common reagents (acids, bases, oxidizers, gases, organics, and salts) you can tap straight into a reaction, each shown with its real structure. • Truer coordination chemistry — metals now chelate correctly, so chlorophyll's magnesium binds all four ring nitrogens, and heme and the water‑splitting catalyst behave as they should. • Real 3‑D bond lengths — molecules render with true relative bond lengths, so a metal–ligand bond looks as long as it actually is. • A smoother Reaction Lab — the "See in 3‑D" and "Energy" panels open and close cleanly. As always, every result is computed from the model — nothing is stored, and nothing is faked.