Project Genesis: An Interactive 3D Chemistry Mechanism Simulator

Genesis now lets chemistry communicators spawn reactants and watch Brønsted chemistry that keeps its atoms: ammonia taking a proton from hydronium finishes as tetrahedral ammonium and bent water, with the same N⁺ on reaction-built NH₄⁺ as on a spawned ion.

As the pair commits, a curved reaction arrow appears in organic-chemistry convention — electron-pushing from source to destination — gradient-colored from electron-rich (red) to electron-deficient (blue), so who attacked whom is readable without a caption.

Reverse transfer follows aqueous driving force (hydronium and hydroxide recombine; ammonium and water do not). Like-charge push reads as open δ⁺/δ⁻ wavefronts instead of orange jewelry, so arrow language stays reserved for mechanism. Hosted WebGL matches editor sharpness; peroxides and isocyanic acid read as chains; click a structure to inspect without fighting the orbit controls.

Process of experimentation: atom-conserving product wrap, pKa-gated reverse transfer, and charge-cue language (βv4.424)

Uncertainty going in: whether a visible H hop could be the authored product without a second layout pass; whether far-range site–site 1/r Coulomb would court NH₃+H₃O⁺ or explode them; whether closed amber rings could ever read as repulsion, and whether reaction-style arrows at range would steal OChem arrow language; whether a pair cooldown could stand in for thermodynamics; and whether hosted softness was canvas DPR or a stacked URP downscale.

We compared skip-arrange and pose-restore (both failed Play: faithful hop / wrong product, or a teleport baked into the snapshot), then settled families with a scaled authored synthesis after choreography. Site–site 1/r at classroom distances summed nine Hδ⁺↔Hδ⁺ terms and flung the pair apart before reaction — replaced with ion–dipole at range and a short faded site window; electron-flow arrows fire only when the pair is close enough to react (N lone pair → hydronium H; leaving O–H pair → O).

Wavefront A/B smoke: pass: true, two open outward arcs, zero solo arcs; glow was not the hitch. Aqueous pKa / log K gating (always if log K ≥ 2, never if < −2) plus EditMode tests that NH₄⁺+H₂O must not transfer replaced a 3 s chemistry lock. URP renderScale 0.8 stacked on a buffer shrink was the WebGL softness; desktop/tablet now render at 1.0.

Remaining question for the next process of experimentation: can molecule+atom merge, polymer growth, and ester recover-from-blob become the same atom-conserving authored-product wrap — or does spawn/destroy remain required when there is no live cluster to settle?

Chemistry communicators can spawn reactants and watch dehydration, combustion, oxidation, addition, esterification, and proton transfer keep the same atoms from start to finish: glucose to maltose is bond surgery on the live cluster, not a magic swap.

Electrons teach occupancy — Domain Clouds with in-domain drift instead of Bohr planets, σ plus distinct π faces on multiple bonds, and Cloud / Bohr / Lewis modes with amber / teal / gray stippled shells. Polar molecules court and hydrogen-bond; Play-audited geometry covers water through urea-cycle amino acids and benzene.

Save Structure, scene, GLB, and phone AR carry Lewis strokes and Cloud stipple; click-to-place spawn and orbit triangles make a lecture demo steerable. Lewis is the default overlay. The useful outcome is a live engine whose mechanism and electron story match what you would draw on a board.

Process of experimentation: atom-conserving choreography, electron-occupancy representation, and live-force capacity (βv4.418)

Uncertainty going in: whether destroy-and-respawn could be retired without corrupting bonds; whether valence dots could stop reading as planets without a QM renderer; why ChargeField steering vanished after the June splash; how many live-force atoms a Chromium WebGL build could hold at a 30 FPS floor; and whether Play-reported geometry bugs were assets, forces, or kinematic pins.

Glycosylation “Option B” (live bond surgery + orientation → proximity → rearrangement → release) was rolled to every family; AtomConservingReactions smoke ran 12 coordinator-path sections. Circular orbits, Unlit haze balloons, and Gaussian octahedron swarms were rejected (still planets / pink spheres / “mess of squares”); in-domain drift + stippled ellipsoid shells Play-confirmed.

Recovered telemetry showed 16 consecutive Genesis sessions with chargeField.on=false / activePairs=0 — the field was a silent opt-in; it now defaults on. Spawn-only WebGL capacity: interactive through ~126 live force atoms (~25 ms), 30 FPS floor near ~140, collapse past ~150 (~2 FPS). Maltose “missing atom” was a stretched C1–O4 dock; Ornithine’s far carboxyl H was rb.position on a kinematic child. Merge-ladder smoke 59/59; 24/24 reactions Play-completed.

Remaining question: can addition and esterification docks become the authored product the way oxidation Stage→morph→Finalize did, and can post-product combustion scatter be reproduced in batch smoke rather than only in interactive Play?

You can put the same molecule in front of a class and change the electron language without changing the chemistry. Cloud mode is a dotted probability surface on the familiar ellipsoid; Bohr restores classic rings; Lewis is an inspect overlay with solid / dash / wedge bonds and lone-pair dots.

Amber marks occupied regions, teal marks opposite-phase ghosts, gray marks covalent overlap, with a most-likely-to-absent fade across each lobe. Water in Play reads as the textbook AX₂E₂ story in three representations. The useful outcome is an electron visual you can stand behind when someone asks whether you are teaching orbitals, Lewis, or occupancy.

Process of experimentation: Cloud / Bohr / Lewis triad and stippled domain-shell representation (βv4.417)

Uncertainty going in: whether a post–two-mode Cloud/Lewis split could become a stable Cloud / Bohr / Lewis triad without breaking PlayerPrefs ints; whether Domain Clouds could stop reading as pink CPK balloons (and later as a mess of squares) while keeping chemically correct ellipsoid envelopes; whether a cheap object-space probability gradient would read as most-likely ↔ absent; and how to restore amber / teal / overlap-lens coding on a surface stipple without claiming true Orbital (± / *) physics.

Soft URP Unlit haze lobes were rejected (still balloons). Gaussian volume swarms of tiny octahedra were rejected after Play (“mess of squares”). Shared high-res ellipsoid meshes plus a stipple shader stuck dots to the lobe; PlayerPrefs: Cloud = 0 (default), Bohr = 1, Lewis = 3, legacy Hybrid (2) maps to Cloud. Play Mode water: stippled shells, then amber / teal / gray + gradient, confirmed “perfect.” WebGL export of this splash was labeled βv4.417 while Player Settings still said βv3.220.

Remaining question (parked at this splash): can a true Orbital mode encode energy / phase / antibonding without colliding with the pedagogical occupancy palette — and can charge-surface and mechanism-arrow cues land without stealing Lewis/OChem language?

Polar and ionic structures now steer the way a communicator needs them to on stage: they court, rotate to present the right face, and nestle into contact instead of slamming or fanning apart. Prefab-baked and live-built molecules behave the same. Attraction ribbons and dashed H-bonds make polarity visible; macromolecules stay hydrodynamically calm when water approaches an –OH instead of whipping as one rigid rod.

Reactions cannot transmute elements — a conservation guard blocks impossible product sets, and seven previously “working” equations were rebalanced to real chemistry. The stoichiometry HUD appears only while there is something to say. The useful outcome is dynamics you can show as chemistry, not as a particle sandbox that sometimes invents cyanide.

Process of experimentation: unified charge sites, field dominance, and atom-conservation guards (βv3.219)

Uncertainty going in: why molecules fanned out versus the May 22 Genesis4 WebGL; whether prefab-stripped and live-built structures could share one charge representation; whether charge could dominate onion-spring containment; whether off-center site forces would spin polymers; and whether a reported water/hydronium → cyanide path was a conjugate misconfiguration. File mtimes plus IL2CPP metadata isolated the regression without git history. Telemetry: unified-field maxPairAccel ≈ 0.83 (peaks ~3.3) sat at containment/tumble scale — charge did not dominate. coulombStrength 140→450, maxAccelerationPerPair 4→12, maxAccelerationPerSite 8→22; ion pairs then hit the per-pair cap; polar O(δ⁻)↔H(δ⁺) ~1.6 at 3.8 u. ChargeSiteProvider unifies live capture, prebake, and element-identity fallback. PVA giration: extendedComForceShare 0.88 of Coulomb to COM, alignment suppressed 0.92. GenesisChemistryAudit on 112 structures: 1 violation (Boric Acid → Borate dropped 3 H, not 1). GuardReaction blocked 7 genuinely non-conserving families; those equations were rebalanced. Cyanide is not a proton-transfer conjugate bug (audit clean).

Remaining question: which untraced commit path still spawned cyanide, and does polymer merge need the same conservation guard — or segmented-chain physics if local backbone flex becomes a feature?