Basic Science

Iterative stochastic-approximation method for finding the root/optimum of a function from noisy observations (Robbins & Monro, "A Stochastic Approximation Method," Annals of Mathematical Statistics, 1951). Pure statistics with no ML framing at origin -- introduced decades before any machine-learning application existed, and became the algorithmic ancestor of the optimizer used to train virtually all modern neural networks.

refinement of Stochastic Gradient Descent

Adaptive moment estimation optimizer (Kingma & Ba, arXiv:1412.6980, ICLR 2015) -- refines SGD with per-parameter adaptive learning rates via first/second moment estimates. Same stochastic-approximation family as its parent, not a new field.

→ 1 link into the applied tree

Classical recursive estimator for the state of a linear dynamical system from noisy observations (Kalman, "A New Approach to Linear Filtering and Prediction Problems," 1960). Pure control theory decades before any deep-learning application -- the state space models Mamba builds on cite it directly as their conceptual origin.

Seppo Linnainmaa's 1970 MSc thesis (University of Helsinki, "The representation of the cumulative rounding error of an algorithm as a Taylor expansion of the local rounding errors," in Finnish) introduced the reverse mode of automatic differentiation -- efficiently computing the derivative of a composite function represented as a graph by recursively applying the chain rule backward through it. General numerical technique, no neural-network framing at origin. Published as Linnainmaa, "Taylor expansion of the accumulated rounding error," BIT Numerical Mathematics 16(2):146-160, 1976 (OpenAlex W2018435387, 309 citations) -- the citable published version of the same thesis result. Split out from Backpropagation (2026-09-03) per BASIC_METHODOLOGY.md Tier 1 SS4 Option B, now that Basic has scaled past the 3-node POC that originally justified keeping this as one compromise-dated node: this is the clean, single-paper, unambiguous mathematical origin, distinct from Werbos's murkier 1974 thesis / 1982 concrete-application claim to have first applied the technique to a neural-network-style model (left as historical context on Backpropagation's own description, not modeled as a third node -- that claim is gradual/multi-source in the same way Many-Body Localization's lead-up was, not a single clean result to split off).

Edward Witten's foundational paper (Commun. Math. Phys. 117, 353, 1988) established topological quantum field theory as a framework connecting QFT path integrals to topological invariants of manifolds (Donaldson theory, Chern-Simons theory). Michael Atiyah's companion 1988 paper gave the mathematical axiomatization of the same concept -- the two are usually cited together as co-foundational. The original candidate-list citation ("Commun. Math. Phys. 121, 351, 1989") does not match any real record; the correct citation is 1988, not 1989.

Independent, simultaneous discovery by David Gross & Frank Wilczek and by David Politzer (both Phys. Rev. Lett. 30, 1973) that non-Abelian gauge theories (like QCD) have a coupling constant that weakens at short distances/high energies -- explaining why quarks behave as nearly free particles inside hadrons at high energy, and establishing QCD as the correct theory of the strong force. Gross, Politzer and Wilczek shared the 2004 Nobel Prize. Bare "asymptotic freedom" is not a safe backfill term (picks up unrelated stemming noise, e.g. Pauli's spin-statistics theorem); a "non-Abelian gauge" qualifier is required.

Independent co-discovery of the J/psi meson (a charm-anticharm bound state) by Burton Richter's group at SLAC and Samuel Ting's group at BNL, announced jointly Nov 11 1974 (the "November Revolution"), confirming the charm quark predicted by the 1970 Glashow-Iliopoulos-Maiani (GIM) mechanism. Richter and Ting shared the 1976 Nobel Prize in Physics for this -- 1976 is the Nobel year, not the discovery year.

Discovery of the Upsilon resonance at 9.5 GeV by Leon Lederman's E288 collaboration at Fermilab (1977), confirming the bottom quark predicted by the 1973 Kobayashi-Maskawa six-quark model, which required a third generation of quarks to accommodate CP violation.

Discovery that Hall conductance in a 2D electron gas is quantized in exact integer multiples of e^2/h, independent of sample geometry or purity (von Klitzing, Dorda & Pepper, 1980) -- the first quantum Hall effect found, later explained topologically (Laughlin 1981, TKNN 1982) and distinct from the Fractional QHE discovered two years later. Query term required the "integer quantum hall effect" exact phrase; the bare "quantum hall effect" term does not separate this from the fractional variant and picks up unrelated noise. Von Klitzing won the 1985 Nobel Prize.

refinement of Integer Quantum Hall Effect

Robert Laughlin's 1981 gauge-invariance argument (a Corbino-geometry gedankenexperiment) explaining why the Hall conductance in the integer quantum Hall effect is exactly quantized -- the first theoretical explanation of the 1980 experimental discovery, predating the more general topological (TKNN) explanation by a year.

refinement of Integer Quantum Hall Effect

Thouless, Kohmoto, Nightingale & den Nijs's 1982 discovery that the integer quantum Hall effect's quantized conductance is a topological invariant (a Chern number) of the electronic band structure -- a more general explanation than Laughlin's 1981 gauge argument, and the direct theoretical ancestor of the entire topological-phases-of-matter field (topological insulators, Weyl semimetals, QAHE) that followed decades later. Per the nearest-parent-only rule, linked here to Integer QHE (the phenomenon it explains) rather than to every downstream topological-matter node it influenced.

Discovery (Tsui, Stormer & Gossard, 1982) that in a very-high-purity 2D electron gas at strong magnetic field and low temperature, the Hall conductance is quantized in exact fractional (not just integer) multiples of e^2/h -- an unexpected result implying collective, strongly-correlated electron behavior rather than the single-particle physics behind the 1980 integer effect. Tsui and Stormer shared the 1998 Nobel Prize with Laughlin.

refinement of Fractional Quantum Hall Effect

Robert Laughlin's 1983 theoretical explanation of the fractional quantum Hall effect: a many-body wavefunction describing an incompressible quantum fluid with fractionally-charged quasiparticle excitations -- the first prediction of fractional electric charge as an emergent phenomenon. Laughlin shared the 1998 Nobel Prize with Tsui and Stormer.

Direct discovery of the W boson (Jan 1983) and Z boson (May 1983) by the UA1 collaboration at CERN's SppS collider (independently confirmed by UA2), verifying the electroweak unification theory of Glashow (1961), Weinberg (1967) and Salam (1968), which had already been indirectly supported by the 1973 Gargamelle neutral-currents discovery. Companion Z-boson paper: OpenAlex W1996888950. Rubbia and van der Meer shared the 1984 Nobel Prize.

Two companion 1985 papers (Devoret, Martinis & Clarke) reported the first clean observation of macroscopic quantum tunneling out of the zero-voltage state of a current-biased Josephson junction, and of quantized energy levels in that same state -- the first time quantum mechanical behavior (tunneling and discrete energy levels) was demonstrated in a macroscopic electrical circuit variable rather than a microscopic particle. Real theoretical precedent exists from 1981-84 (Leggett-style proposals that this should be observable, and studies of how dissipation affects it), confirmed via OpenAlex noise test -- not noise, a genuine multi-year lead-up. Basis for the 2025 Nobel Prize (Clarke, Devoret, Martinis) -- 2025 is the Nobel year, not the discovery year.

→ 1 link into the applied tree

Discovery of superconductivity at 35K in a La-Ba-Cu-O ceramic (Bednorz & Muller, 1986) -- far above any previously known superconductor and the first cuprate superconductor -- ending a decades-long search for higher-Tc materials confirmed as real, on-topic precedent literature via OpenAlex noise test (1968-1982). Bednorz and Muller won the 1987 Nobel Prize, remarkably within a year of publication.

refinement of High-Temperature Superconductivity

Discovery of superconductivity at 93K in an yttrium-barium-copper-oxide compound (Wu, Chu et al., 1987) -- the first superconductor above the 77K liquid-nitrogen boiling point, making practical (cheap-coolant) superconductor applications possible for the first time. Direct successor to Bednorz & Muller's 1986 La-Ba-Cu-O discovery; modeled as a child of High-Temperature Superconductivity via parent_node_id, the same pattern used for Adam/SGD.

Precision measurement of the invisible decay width of the Z boson at LEP established that there are exactly 3 light neutrino species (N_v = 2.98 +/- 0.06), ruling out a hypothetical fourth generation of light neutrinos/leptons. Multiple LEP collaborations (ALEPH, DELPHI, L3, OPAL) published closely related results starting in 1989; this node anchors to the earliest of that series.

The COBE satellite's FIRAS instrument measured the cosmic microwave background spectrum to be a near-perfect blackbody (John Mather, 1990) -- one of the most precise blackbody-spectrum measurements ever made of anything, confirming a core prediction of Big Bang cosmology. Mather shared the 2006 Nobel Prize (with Smoot) for this and the related anisotropy work.

COBE's DMR instrument detected the first confirmed anisotropies (temperature fluctuations) in the cosmic microwave background (George Smoot, 1992) -- the primordial density fluctuations that seeded all later large-scale structure. Real, on-topic precedent literature exists back to the 1970s-80s (earlier, less-definitive ground/balloon-based anisotropy measurement attempts), confirmed via OpenAlex noise test. Smoot shared the 2006 Nobel Prize with Mather.

Bennett, Brassard, Crepeau, Jozsa, Peres & Wootters' 1993 protocol for transmitting an unknown quantum state using a shared entangled pair plus two classical bits -- the origin of "teleportation" as a serious physics term, not a popularization of an earlier idea (bare term tested completely clean pre-1993 via OpenAlex noise test).

→ 1 link into the applied tree

refinement of Quantum Teleportation (Theory)

First experimental demonstration (Bouwmeester, Pan, Mattle, Eibl, Weinfurter & Zeilinger, 1997) of the 1993 quantum teleportation protocol, using photon polarization states.

Joint discovery of the top quark by the CDF and D0 collaborations at Fermilab's Tevatron, announced March 1995, completing the third generation of quarks predicted by the 1973 Kobayashi-Maskawa model. CDF's 1994 "evidence" paper preceded this as a preliminary, non-definitive claim -- not the discovery itself. Companion D0 paper: OpenAlex W1525067064. Note: OpenAlex initially surfaced a near-duplicate, near-uncited CDF record (W3208541762, 3 citations) before the canonical 1,493-citation record below was found -- the same fragmented-record pattern the Basic subsystem's claim-evidence pilots have repeatedly documented.

First experimental creation of a Bose-Einstein condensate, in a dilute rubidium-87 vapor (Anderson, Ensher, Matthews, Wieman & Cornell, 1995), 71 years after Einstein predicted the phenomenon (1924/25, building on Bose 1924). Also relevant to statistical mechanics as a quantum phase transition, not only quantum mechanics proper. Query term required a "dilute" qualifier -- the bare term collides with the decades-long, different-mechanism debate over whether superfluid helium-4 is itself a BEC, and with at least one fringe/pseudoscience paper on "consciousness and Bose-Einstein condensates." Cornell, Wieman and Ketterle shared the 2001 Nobel Prize.

Antoine, L'Huillier & Lewenstein's 1996 theoretical proposal that high-order harmonic generation in intense laser-atom interactions naturally produces trains of attosecond-duration light pulses. Bare "attosecond pulse train" tested clean pre-1996 via OpenAlex noise test (a single, unrelated 1965 metallurgy paper).

refinement of Attosecond Pulse Trains (Theory)

Hentschel et al.'s 2001 "Attosecond metrology" reported the first isolation and measurement of a single (not a train of) attosecond light pulse -- a distinct, harder technical achievement from observing a pulse train, achieved the same year by a different group. Modeled as a sibling of the Attosecond Pulse Trains Observation node (both build on the same 1996 theoretical framework but achieve different specific milestones), not a restatement of it. This is the pairing that is genuinely cited for the 2023 Nobel Prize (Krausz, L'Huillier, Agostini) -- the source list's "2023" year is the Nobel year, not the discovery year, same pattern as several other corrections in this dataset.

refinement of Attosecond Pulse Trains (Theory)

Paul et al.'s 2001 first experimental observation of a train of attosecond pulses generated via high-harmonic generation, confirming the 1996 theoretical proposal.

The discovery that the expansion of the universe is accelerating rather than decelerating under gravity alone, implying a mysterious repulsive component ("dark energy") makes up roughly 68% of the universe's total energy density. Established independently by two competing teams measuring the brightness/redshift relation of distant Type Ia supernovae: the Supernova Cosmology Project (Saul Perlmutter, work begun 1988) and the High-Z Supernova Search Team (Brian Schmidt, work begun 1994, joined by Adam Riess). Riess et al. submitted first (Astronomical Journal, March 13 1998, published Sept 1998); Perlmutter et al. followed (Astrophysical Journal, 1999). Jointly named Science magazine's 1998 "Breakthrough of the Year." Perlmutter, Schmidt, and Riess shared the 2011 Nobel Prize in Physics "for the discovery of the accelerating expansion of the universe through observations of distant supernovae." Independently significant on its own terms as one of the most consequential empirical results in modern cosmology, admitted here on that basis -- not merely as a target for any one enablement edge (docs/APPLIED_BASIC_ENABLEMENT_ADMISSION.md Sec 2 criterion 2).

Alexei Kitaev's 2001 proposal that a 1D p-wave superconductor ("Kitaev chain") should host unpaired Majorana zero modes at its ends -- the earliest and most-cited theoretical route to condensed-matter Majorana modes (4,815 citations), predating Fu & Kane's independent, more experimentally-tractable 2008 proposal (topological-insulator/superconductor heterostructure, OpenAlex W2168226651, 4,925 citations) by 7 years. The 2012 nanowire experiment (below) more directly follows the Fu-Kane/Oreg-Lutchyn theoretical route than Kitaev's abstract chain model, even though Kitaev's is chronologically first -- a genuine multi-path precedence situation, not collapsed to a single "first" paper. Bare "Majorana fermion" is not a safe backfill term -- it collides badly with the unrelated particle-physics concept of Majorana neutrinos; a "topological superconductor" qualifier is required.

refinement of Majorana Zero Modes (Theory)

First experimental signature (Mourik et al., 2012) consistent with Majorana zero modes, in a hybrid superconductor-semiconductor nanowire device -- a zero-bias conductance peak matching theoretical predictions, though not yet unambiguous proof (a genuinely still-debated point in the field, distinct from the topological-insulator or graphene cases where the experimental confirmation is uncontested).

The Wilkinson Microwave Anisotropy Probe's first-year results (Bennett, Spergel et al., 2003) delivered a dramatic jump in cosmological precision over COBE, pinning down the universe's age, composition (dark energy/dark matter/baryon fractions), and geometry to a few percent. Landmark paper here is the "Determination of Cosmological Parameters" companion; "Preliminary Maps and Basic Results" is the other widely-cited paper from the same release.

Novoselov & Geim's 2004 isolation of a single atomic layer of carbon (graphene) using mechanical exfoliation ("Scotch tape method") -- the first time a genuinely 2D crystal was isolated and shown to be stable and measurable. Real, on-topic theoretical precedent exists back to 1992 (carbon-nanotube-adjacent tight-binding theory of graphene sheets), confirmed via OpenAlex noise test -- not noise, a legitimate 12-year theoretical lead-up to the physical isolation.

→ 1 link into the applied tree

refinement of Graphene Isolation

Discovery that charge carriers in graphene behave as massless relativistic Dirac fermions, reported independently and essentially simultaneously in 2005 by Novoselov et al. (Nature 438, 197, primary landmark here) and Zhang et al. (Nature 438, 201, OpenAlex W2118492507) -- a genuinely distinct physical finding from graphene's isolation the year before, not a restatement of it.

A time-reversal-symmetric topological phase of matter with a bulk band gap and gapless, spin-momentum-locked edge/surface states protected by a Z2 topological invariant. Kane & Mele (2005), "Quantum Spin Hall Effect in Graphene" (PRL 95, 226801), introduced the first concrete Z2-classified model and coined "quantum spin Hall effect" in this sense -- the paper the field actually cites as the origin. Real precedents exist: Thouless, Kohmoto, Nightingale & den Nijs (1982) first showed Hall conductance is a topological (Chern-number) invariant; Haldane (1988) showed a quantized Hall effect requires only broken time-reversal symmetry, not an external field; Murakami, Nagaosa & Zhang (2004) proposed an adjacent but distinct "spin-Hall insulator" concept in narrow-gap semiconductors. Graphene's own intrinsic spin-orbit gap proved too small for direct experimental observation of Kane-Mele's original system -- the first experimental confirmation came via a different material platform, HgTe quantum wells, independently predicted by Bernevig, Hughes & Zhang (2006) and confirmed by Konig et al. (2007).

RHIC's four experiments (BRAHMS, PHENIX, PHOBOS, STAR) jointly announced in April 2005 that the quark-gluon plasma created in heavy-ion collisions behaves as a strongly-coupled, nearly-perfect (very low viscosity) fluid rather than the weakly-interacting gas originally expected. The landmark OpenAlex record here is Gyulassy & McLerran's synthesizing review (Nucl. Phys. A 750, 30) rather than one of the four original experimental collaboration papers -- OpenAlex records it with a 2004 publication year (online-first date), while the historically-cited announcement date is April 2005; flagged rather than silently resolved.

Theoretical prediction (Liu, Qi, Dai, Fang & Zhang, 2008) that a magnetically-doped topological insulator (HgMnTe) could exhibit a quantized Hall conductance with zero external magnetic field -- unlike the ordinary quantum Hall effect (1980) or fractional QHE (1982), which both require a strong external field. Distinct phenomenon from generic "topological insulator" (already tracked); bare "quantum anomalous Hall effect" tested clean pre-2008 via OpenAlex noise test.

refinement of Quantum Anomalous Hall Effect (Theory)

First experimental observation (Chang et al., 2013) of the quantum anomalous Hall effect, in a thin film of chromium-doped (Bi,Sb)2Te3 -- confirming the 2008 theoretical prediction 5 years later.

Theoretical prediction (Wan, Turner, Vishwanath & Savrasov, 2011) that pyrochlore iridates could realize a Weyl semimetal phase -- a 3D analogue of graphene's Dirac points hosting emergent massless Weyl fermions with topologically protected Fermi-arc surface states.

refinement of Weyl Semimetal (Theory)

First experimental realization of a Weyl semimetal, in tantalum arsenide (TaAs) -- reported essentially simultaneously by two independent groups in 2015: Xu et al. (Science 349, 613, primary landmark here) and Lv et al. (Phys. Rev. X 5, 031013, OpenAlex W2950417131).

Joint discovery of the Higgs boson, announced July 4 2012, by the ATLAS and CMS experiments at the LHC -- confirming the Higgs mechanism proposed independently in 1964 by Higgs, by Englert & Brout, and by Guralnik, Hagen & Kibble, 48 years earlier. Companion CMS paper: OpenAlex W2167727518. Englert and Higgs shared the 2013 Nobel Prize.

Frank Wilczek's 2012 proposal that a quantum system could spontaneously break time-translation symmetry, forming a periodic structure in time analogous to how an ordinary crystal breaks spatial-translation symmetry. Bare "time crystal" collides badly with unrelated crystallography ("real-time crystal size measurements") and even film-theory literature ("The Scattering of Time [in] Crystals: Deleuze, Mysticism and Cinema"); a "symmetry breaking" qualifier is required for a clean signal.

refinement of Time Crystals (Theory)

First experimental realizations of discrete time crystals, reported simultaneously in 2017 by two independent groups using different platforms: Zhang et al. (trapped ytterbium ions, Nature 543, 217, primary landmark here) and Choi et al. (nitrogen-vacancy centers in diamond, Nature 543, 221, OpenAlex W2540688686). Both realize a periodically-driven (Floquet) discrete time crystal, not Wilczek's original continuous-time-translation-breaking proposal, which was later shown (Nature 2015, "Absence of Quantum Time Crystals," OpenAlex W1652183760) not to be realizable in equilibrium -- the 2017 experiments are of a genuinely different, driven variant.

First direct detection of gravitational waves, from the merger of two roughly 30-solar-mass black holes, confirming a 1916 prediction of general relativity and following decades of indirect evidence (the Hulse-Taylor binary pulsar, 1974) and unsuccessful direct-detection attempts. Also a major astrophysical result in its own right (first direct observation of a binary black hole system). Detected Sept 14 2015; announced and published Feb 11 2016 (PRL 116, 061102). This node anchors to the 2015 detection date, not the 2016 publication date, matching the Adam node's existing arXiv-submission-over-publication precedent. Weiss, Barish and Thorne shared the 2017 Nobel Prize.

Three independent groups (Hensen et al., using electron spins; Shalm et al.; Giustina et al.) reported loophole-free violations of a Bell inequality within months of each other in 2015, simultaneously closing both the detection-efficiency and locality/communication loopholes that had left every prior Bell test open to a local-hidden-variable escape. A 2012 paper claimed a "loophole-free" test via a different, narrower method (local precertification of photon presence) that did not close the same combination of loopholes -- a real precedent worth distinguishing, not a competing discovery date. Builds on the original Bell tests of the 1970s-80s (Freedman-Clauser, Aspect), which is Basic's pre-1976 boundary case and not itself in this dataset.

Schreiber et al.'s 2015 experimental observation of many-body localization -- interacting fermions in a disordered (quasirandom) optical lattice failing to thermalize, preserving memory of their initial state indefinitely, in violation of the eigenstate thermalization hypothesis. Real theoretical precedent exists back to 2006-2014 (Basko-Aleiner-Altshuler and later entanglement-growth/phenomenology papers), confirmed via OpenAlex noise test as genuine on-topic lead-up, not noise -- though multi-paper and gradual enough that this is modeled as a single node (Tier 1 SS4 Option A) rather than forced into one theory/experiment pair like this batch's other candidates.

First gravitational-wave detection of a binary neutron star merger, observed simultaneously across gravitational waves (LIGO/Virgo) and the electromagnetic spectrum (gamma-ray burst, optical kilonova, and more) -- the founding event of multimessenger astronomy and observational confirmation that neutron star mergers produce heavy elements via the r-process. Distinct discovery from GW150914 (a black hole merger, no electromagnetic counterpart possible).

Two graphene sheets stacked and twisted relative to each other near a "magic angle" (~1.1 degrees) host flat electronic bands, in which unconventional superconductivity and Mott-like correlated insulating states emerge -- launching the field now called "twistronics." Bistritzer & MacDonald (2011), "Moire bands in twisted double-layer graphene" (PNAS 108, 12233), theoretically predicted the flat-band phenomenon at the magic angle, but did not predict superconductivity or correlated-insulator behavior specifically. Cao, Fatemi, Fang, Watanabe, Taniguchi, Kaxiras & Jarillo-Herrero (2018), "Unconventional superconductivity in magic-angle graphene superlattices" (Nature 556, 43), made the actual empirical discovery -- alongside a companion paper the same issue reporting correlated insulating behavior at half-filling -- the genuinely surprising result the field is anchored to, not something the 2011 theory itself predicted.

First direct image of a black hole (the supermassive black hole at the center of M87), published April 10 2019 by the Event Horizon Telescope Collaboration -- also a direct confirmation of general relativity's predicted black-hole shadow shape and size. Landmark paper is "Paper I" of a six-paper series; the EHT collaboration's own pre-2019 methodology and simulation papers (2014-2018) are real, on-topic precedent, not noise.

Two independent papers posted within days of each other on arXiv in November 2019 -- Almheiri, Hartman, Maldacena, Shaghoulian & Tajdini, "Replica wormholes and the entropy of Hawking radiation" (OpenAlex W2991362355, JHEP 2020, primary landmark here) and Penington, Shenker, Stanford & Yang, "Replica wormholes and the black hole interior" (OpenAlex W2991561483, JHEP 2022) -- independently derived the Page curve for an evaporating black hole using a new gravitational path-integral technique ("replica wormholes"), resolving a 45-year-old tension between Hawking's original calculation and unitarity. Builds directly on Penington's solo "Entanglement Wedge Reconstruction and the Information Paradox" (2019, OpenAlex W3081775050), part of the same 2019 breakthrough cluster. Anchored to 2019 (the arXiv posting/real-impact date) rather than either paper's JHEP publication year (2020/2022), matching the same arXiv-date-over-publication-date convention already used for Adam and GW150914.

Two independent teams (CHIME/FRB and STARE2) simultaneously detected a millisecond-duration fast radio burst from the Galactic magnetar SGR 1935+2154 in April 2020 -- the first FRB traced to a specific known source and confirming magnetars as at least one real progenitor of the fast-radio-burst phenomenon (first identified as a class in 2007).

Fermilab's Muon g-2 experiment measured the muon's anomalous magnetic moment to 0.46 ppm precision (2021), confirming and sharpening a possible discrepancy with the Standard Model prediction first suggested by BNL's E821 experiment in the early-to-mid 2000s (real precedent confirmed via OpenAlex noise test, not a single new 2021 finding). Status set to disputed rather than foundational because whether this constitutes evidence of new physics is genuinely unresolved: the 2020 consensus theoretical prediction implies a significant discrepancy, but a competing 2021 lattice-QCD calculation (the BMW collaboration) predicts a value close to the measurement, largely erasing the tension -- the measurement itself is solid, the physics interpretation is not.

JWST observations (Labbe et al., 2023, building on earlier 2022 candidates) found several galaxy candidates from just ~600 million years after the Big Bang that appear unexpectedly massive and mature for their age, in tension with standard galaxy-formation timelines -- an actively unfolding result, not yet a settled fact (some candidates have since had their masses or redshifts revised on spectroscopic follow-up). Status set to emerging rather than foundational to reflect this.

Paclitaxel, produced by the Pacific yew tree (Taxus brevifolia) as a natural anti-herbivore/anti-fungal chemical defense compound. Wani, Taylor, Wall, Coggon & McPhail, "Plant antitumor agents. VI. The isolation and structure of taxol, a novel antileukemic and antitumor agent from Taxus brevifolia" (J. Am. Chem. Soc. 93, 2325-2327, 1971-05-01, OpenAlex W2037490361, 4,088 citations). Became one of the most widely used chemotherapy agents (breast, ovarian, lung cancer) and remains first-line decades after approval; billions of dollars in cumulative sales. Classified under Chemistry/Biochemistry, not Biology, matching the statins/cyclosporine nodes above.

Cyclosporine, produced by the soil fungus Tolypocladium inflatum as a natural antifungal/antibiotic compound, found instead to have potent, selective immunosuppressive activity. Borel, Feurer, Gubler & Stahelin (Sandoz), "Biological effects of cyclosporin A: a new antilymphocytic agent" (Agents and Actions 6, 468-475, 1976-07-01, OpenAlex W2080064179, 1,517 citations). Made modern solid-organ transplantation clinically viable by preventing graft rejection -- created the modern transplant-immunosuppression drug market, still generating billions of dollars annually across the drug class it founded. Classified under Chemistry/Biochemistry, not Biology, matching the statins node above.

Fungal secondary metabolites produced by Penicillium citrinum as a natural chemical-competition trait, found to inhibit HMG-CoA reductase, the rate-limiting enzyme in cholesterol biosynthesis. Endo, Kuroda & Tsujita, "ML-236A, ML-236B, and ML-236C, new inhibitors of cholesterogenesis produced by Penicillium citrinum" (J. Antibiot. 29, 1346-1348, 1976, OpenAlex W2055561625, 957 citations) -- the first published isolation, 1976, not 1973 as sometimes cited in secondary sources. ML-236B is compactin, the first statin. Directly founded the statin drug class; Lipitor (atorvastatin), a later compound in the same class, became the best-selling drug in pharmaceutical history (>$125B in lifetime sales). Classified under Chemistry/Biochemistry, not Biology, per BASIC_TAXONOMY_DRAFT.md's explicit Chemistry-Biology dual-tag framing for natural-product chemistry.

Observation that a Penicillium mold contaminating a bacterial culture plate produced a substance that lysed/inhibited nearby Staphylococcus colonies -- a natural chemical-warfare trait of the mold against competing microbes. Alexander Fleming, "On the Antibacterial Action of Cultures of a Penicillium, with Special Reference to their Use in the Isolation of B. influenzae" (Br. J. Exp. Pathol. 10, 226-236, published 1929-06-01, OpenAlex W2177846120, 1,570 citations) -- the actual publication date, not the "1928" observation date commonly cited in secondary sources. Fleming did not purify or mass-produce it; Florey, Chain and their Oxford team achieved that in the early 1940s, enabling wartime mass production. Fleming, Florey and Chain shared the 1945 Nobel Prize in Physiology or Medicine. Founded the entire antibiotic drug class, predating CRISPR by roughly 80 years as a translation of a microbial defense mechanism into an industry.

Discovery that retroviruses (e.g. Rous sarcoma virus) carry an enzyme that copies their RNA genome into DNA, reversing the normal DNA-to-RNA flow of the "central dogma," so the virus can integrate into a host cell's genome. Independently and simultaneously reported in the same June 1970 issue of Nature by Temin & Mizutani, "RNA-dependent DNA polymerase in virions of Rous sarcoma virus" (1970-06-27, OpenAlex W201050536, 2096 citations) and by Baltimore, "RNA-dependent DNA polymerase in virions of RNA tumour viruses" (1970-06, OpenAlex W2013241878, 2040 citations) -- a genuine dual co-discovery, same shape as this platform's existing Higgs/Weyl-Semimetal dual-origin nodes. Temin and Baltimore shared the 1975 Nobel Prize in Physiology or Medicine (with Renato Dulbecco). Directly enabled cDNA cloning and RT-PCR/RT-qPCR (the same diagnostic technique later used for COVID-19 testing) and underlies modern understanding of HIV replication and the antiretroviral drug class.

Discovery that bacteria defend against invading (phage) DNA by cutting it at specific sequences with a restriction endonuclease, while protecting their own DNA at the same sequence via methylation. Smith & Wilcox, "A restriction enzyme from Hemophilus influenzae. I. Purification and general properties" (J. Mol. Biol. 51, 379-391, 1970-07-01, OpenAlex W1506010125) reported the first isolation of a sequence-specific restriction enzyme (HindII), building on Werner Arber's 1960s prediction of restriction-modification systems. Smith shared the 1978 Nobel Prize in Physiology or Medicine with Arber and Daniel Nathans (who first applied restriction enzymes to map a genome, SV40, 1971). Directly enabled recombinant DNA / molecular cloning (Cohen, Boyer et al., 1973) -- the founding technology of the biotech industry -- not itself written as a separate node in this pass. Note: OpenAlex's author metadata for this record renders the second author as "K.W. Welcox"; the historically correct name is Kent W. Wilcox, used here.

Demonstration that DNA fragments from different sources -- cut with the same restriction enzyme (see the linked precursor node) -- can be spliced together in vitro and the resulting hybrid plasmid propagated in bacteria as a stable, biologically functional replicon. Cohen, Chang, Boyer & Helling, "Construction of Biologically Functional Bacterial Plasmids In Vitro" (PNAS 70, 3240-3244, 1973-11-01, OpenAlex W1995924881, 1,786 citations). The founding technique of the entire biotechnology industry -- Genentech, founded 1976 by Boyer and Robert Swanson, was the first company built directly on it, followed by the first recombinant-protein drug (human insulin, Genentech/Eli Lilly, approved 1982).

Technique for producing a single, defined antibody clone at scale by fusing an antibody-producing B cell with an immortal myeloma cell, yielding a "hybridoma" that indefinitely secretes one antibody of known specificity. Köhler & Milstein, "Continuous cultures of fused cells secreting antibody of predefined specificity" (Nature 256, 495-497, 1975-08-01, OpenAlex W2030522563, 17,229 citations). Köhler and Milstein shared the 1984 Nobel Prize in Physiology or Medicine (with Niels Jerne). Builds on, and is a distinct advance over, 19th-century polyclonal serum therapy (von Behring & Kitasato, 1890/1901 Nobel) -- this technique isolates one clone rather than harvesting a natural mixed antibody pool. Directly founded the monoclonal antibody drug class (Humira, Keytruda, Herceptin, Rituxan, Opdivo), now the single largest class of biologic drugs by global sales.

Isolation of a heat-stable DNA-replication enzyme from Thermus aquaticus, a bacterium living in Yellowstone hot springs -- a natural adaptation to its extreme thermal environment, unrelated at the time of discovery to any human application. Chien, Edgar & Trela, "Deoxyribonucleic acid polymerase from the extreme thermophile Thermus aquaticus" (J. Bacteriol. 127, 1550-1557, 1976-09-01, OpenAlex W1629931862, 704 citations). T. aquaticus itself was first described by Brock & Freeze in 1969. This node anchors the natural enzyme's discovery; see the linked node for its later repurposing into a practical, thermostable-polymerase-based PCR method.

Discovery that a foreign peptide fused to a filamentous phage's minor coat protein (gene III) is displayed on the phage surface while remaining infectious -- turning a phage's natural infection/coat-assembly machinery into a physical link between a displayed protein and the DNA encoding it, enabling large-scale selection ("panning"). George P. Smith, "Filamentous Fusion Phage: Novel Expression Vectors That Display Cloned Antigens on the Virion Surface" (Science 228, 1315-1317, 1985-06-14, OpenAlex W2070360032, 4,132 citations). Smith shared the 2018 Nobel Prize in Chemistry with Gregory Winter (who extended the technique to antibody libraries) and Frances Arnold. Directly enabled phage-display-based antibody discovery without hybridomas -- Humira itself was discovered this way -- and founded a standalone platform-licensing industry (Dyax, Cambridge Antibody Technology).

Soil bacterium Bacillus thuringiensis produces a crystal protein (Cry toxin) as a natural competitive/defensive trait, lethal to specific insect larvae that ingest it but harmless to most other organisms -- first isolated as a commercial insecticidal agent in the early 20th century (Ishiwata, 1901; not independently re-verified for this node). Vaeck, Reynaerts, Hofte, Jansens, De Beuckeleer, Dean, Zabeau, Van Montagu & Leemans (Plant Genetic Systems, Ghent), "Transgenic plants protected from insect attack" (Nature 328, 33-37, 1987-07-01, OpenAlex W2062706925, 922 citations) demonstrated the first transgenic crop plants expressing a cloned Bt toxin gene, the landmark that directly founded the commercial Bt-crop industry. This node anchors that translation event, not the original 1901 bacterial isolation.

Method for exponentially amplifying a specific DNA sequence in vitro via repeated cycles of denaturation, primer annealing, and polymerase extension. Kary Mullis conceived the original technique at Cetus Corporation and it was first demonstrated using the Klenow fragment (Saiki et al., Science, 1985), which required adding fresh enzyme every cycle since Klenow is heat-labile. Saiki, Gelfand, Stoffel, Scharf, Higuchi, Horn, Mullis & Erlich, "Primer-Directed Enzymatic Amplification of DNA with a Thermostable DNA Polymerase" (Science 239, 487-491, 1988-01-29, OpenAlex W2032118018, 17,174 citations) replaced Klenow with Taq polymerase (see the linked precursor node), making PCR practical, automatable, and vastly more specific -- the version that actually founded the PCR industry. Mullis won the 1993 Nobel Prize in Chemistry for the original concept. Roche paid Cetus $300M for the PCR/Taq patent rights in 1991. PCR underlies molecular diagnostics, forensics, ancestry testing, and (via RT-PCR) COVID-19 testing.

→ 1 link into the applied tree

Demonstration that Green Fluorescent Protein -- a bioluminescence-signaling protein from the jellyfish Aequorea victoria, with no natural function related to any human use -- can be expressed in a heterologous organism (E. coli, C. elegans) as a genetically encoded, non-invasive fluorescent reporter of gene expression, needing no substrate or cofactor unlike other bioluminescent systems. Chalfie, Tu, Euskirchen, Ward & Prasher, "Green Fluorescent Protein as a Marker for Gene Expression" (Science 263, 802-805, 1994-02-11, OpenAlex W2032579724, 6,785 citations). Built on Shimomura's original 1962 isolation of GFP and Prasher's 1992 cloning of the gene (a co-author on this paper). Chalfie, Shimomura and Tsien (who later engineered the color-variant palette) shared the 2008 Nobel Prize in Chemistry. Became a near-universal fluorescent reporter/imaging tag across modern molecular and cell biology and drug-discovery research -- a foundational R&D tool rather than a single product line.

Discovery of a natural post-transcriptional gene-silencing mechanism in which double-stranded RNA triggers sequence-specific degradation of matching messenger RNA. Fire, Xu, Montgomery, Kostas, Driver & Mello, "Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans" (Nature 391, 806-811, 1998-02-01, OpenAlex W1647075334, 15,432 citations) showed dsRNA silences gene expression far more effectively than single-stranded antisense RNA, resolving earlier unexplained "co-suppression" observations in petunias. Fire and Mello shared the 2006 Nobel Prize in Physiology or Medicine. Directly translated into RNAi therapeutics -- Alnylam's patisiran (Onpattro), the first FDA-approved RNAi drug, 2018 -- and RNAi-based agricultural biopesticides.

Discovery that CRISPR loci function as a heritable, sequence-specific adaptive immune system in bacteria: spacers derived from phage/plasmid DNA are stored between repeats and used, via Cas proteins, to recognize and destroy matching invaders on reinfection. Barrangou, Fremaux, Deveau, Richards, Boyaval, Moineau, Romero & Horvath (Science 315, 1709-1712, 2007, published online 2007-03-22) provided the first direct experimental demonstration -- in the dairy-starter bacterium Streptococcus thermophilus at Danisco (an industrial yogurt-culture context, not academic curiosity) -- that adding a new phage-derived spacer confers resistance to that phage and removing it abolishes resistance. Built on Mojica et al.'s 2005 hypothesis (J. Mol. Evol. 60, 174-182, OpenAlex W1990487336) that CRISPR spacers derive from foreign genetic elements, and on the repeats themselves, first noted in 1987 (Ishino et al.). Distinct from the 2012 programmable Cas9 genome-editing tool (see the linked node) -- this node is the natural mechanism's discovery, not its later repurposing.

Jinek, Chylinski, Fonfara, Hauer, Doudna & Charpentier (Science 337, 816-821; published online 2012-06-28/29, print Aug 17 2012; DOI 10.1126/science.1225829) showed that Cas9 could be programmed with a single synthetic guide RNA (fusing the natural crRNA and tracrRNA) to cut any target DNA sequence, and proposed it as a genome-editing tool -- turning the CRISPR-Cas bacterial immune mechanism (see the linked precursor node) into a general-purpose molecular technology. Doudna and Charpentier shared the 2020 Nobel Prize in Chemistry for this work. Feng Zhang and George Church separately published the first demonstrations of CRISPR-Cas9 genome editing in mammalian/human cells in January 2013 -- a well-documented priority/patent dispute (Broad Institute vs. UC Berkeley) that this node does not adjudicate and is not itself anchored to; flagged here as a real, separate candidate node for a future pass, not folded into this one.

→ 1 link into the applied tree

First demonstrations that CRISPR-Cas9 (see the linked precursor node) works as a genome-editing tool in eukaryotic/human cells, not just in vitro or in bacteria. Two independent groups published within days of each other, both online 2013-01-04: Cong, Ran, Cox, Lin, Barretto, Habib, Hsu, Wu, Jiang, Marraffini et al. (Feng Zhang's lab, Broad Institute/MIT), "Multiplex Genome Engineering Using CRISPR/Cas Systems" (Science 339, 819-823, OpenAlex W2064815984, 15,933 citations); and Mali, Yang, Esvelt, Aach, Guell, DiCarlo, Norville & Church (George Church's lab, Harvard), "RNA-Guided Human Genome Engineering via Cas9" (Science 339, 823-826, OpenAlex W2003171404, 9,503 citations). This is the specific pair of papers at the center of the well-documented CRISPR patent dispute -- the Broad Institute (Zhang) was awarded U.S. patents for eukaryotic-cell CRISPR editing, while UC Berkeley (Doudna/Charpentier, the 2012 in vitro paper) contested them; both institutions hold patents today after years of interference proceedings and litigation. This node does not adjudicate that dispute, only anchors the two papers' shared 2013 date as the real landmark.

Characterization of DNA-guided (rather than RNA-guided) Argonaute proteins in bacteria/archaea, distinct from the RNA-guided eukaryotic Argonautes that carry out RNA interference. Swarts, Jore, Westra, Zhu, Janssen, Snijders, Wang, Patel, Berenguer & Brouns, "DNA-guided DNA interference by a prokaryotic Argonaute" (Nature 507, 258-261, 2014-02-14, OpenAlex W4246016408, 522 citations), in Thermus thermophilus, showed pAgo uses small DNA guides to cleave complementary DNA -- a nucleic-acid-targeting activity structurally unrelated to CRISPR-Cas despite a superficially similar "guide + nuclease" shape. Early-stage diagnostics and gene-editing-adjacent tool development is underway (some groups pursuing pAgo as a CRISPR alternative/complement), but this is real, pre-industry work, not yet a translated technology -- one step behind CBASS above.

Demonstration that Cas13a, once activated by binding its target RNA, indiscriminately cleaves nearby single-stranded RNA reporters ("collateral cleavage") -- a natural defense-amplification trait of this CRISPR-Cas variant -- repurposed into a rapid, sequence-specific nucleic-acid diagnostic. Gootenberg, Abudayyeh, Lee, Essletzbichler, Dy, Joung, Verdine, Donghia, Daringer & Zhang (Feng Zhang's lab), "Nucleic acid detection with CRISPR-Cas13a/C2c2" (Science 356, 438-442, 2017-04-13, OpenAlex W2606915822, 3,955 citations). A "second wave" translation of the same CRISPR-Cas mechanism family (see the linked precursor node) into diagnostics rather than editing -- commercialized by Sherlock Biosciences, including an FDA-authorized COVID-19 test.

Discovery that bacteria use TIR (Toll/interleukin-1 receptor)-domain proteins as phage sensors within multi-gene defense systems -- the same domain family used by human Toll-like receptors (innate immunity signaling) and by SARM1 (the NAD+-depleting enzyme implicated in axonal degeneration), now understood to be evolutionarily related rather than superficially similar. Doron, Melamed, Ofir, Leavitt, Lopatina, Keren, Amitai & Sorek, "Systematic discovery of antiphage defense systems in the microbial pangenome" (Science 359, eaar4120, 2018-01-25, OpenAlex W2784532673, 1,402 citations) identified Thoeris (among nine other new defense systems) via large-scale comparative genomics of gene clusters flanking known defense genes. The specific NAD+-depletion biochemical mechanism (ThsB's TIR domain producing a signaling molecule that activates the ThsA effector) was elaborated in subsequent 2019-2022 work not independently re-verified for this node. Live drug-discovery programs already target SARM1 (neurodegeneration) and TLR agonism/antagonism (immuno-oncology, vaccine adjuvants) -- not yet its own industry, but a real, current "basic science not yet translated" case.

Demonstration that Cas12a, once activated by binding its double-stranded DNA target, indiscriminately cleaves nearby single-stranded DNA reporters -- the DNA-targeting analogue of Cas13a's RNA collateral-cleavage activity (see the linked SHERLOCK node) -- repurposed into a rapid nucleic-acid diagnostic. Chen, Ma, Harrington, Da Costa, Tian, Palefsky & Doudna (Jennifer Doudna's lab), "CRISPR-Cas12a target binding unleashes indiscriminate single-stranded DNase activity" (Science 360, 436-439, 2018-02-15, OpenAlex W2787024379, 4,560 citations). Commercialized as the DETECTR platform by Mammoth Biosciences; another genuine "second wave" diagnostic translation of the CRISPR-Cas mechanism family (see the linked precursor node), independent of, but published within months of, the Zhang-lab SHERLOCK platform above.

Discovery that bacteria encode CBASS operons -- a cyclic-nucleotide second-messenger signaling system, structurally and mechanistically homologous to the human cGAS-STING innate-immune pathway, that senses phage infection and triggers an effector response (often NAD+ depletion or membrane damage) to abort it before the phage can spread. Cohen, Melamed, Millman, Shulman, Oppenheimer-Shaanan, Kacen, Doron, Amitai & Sorek, "Cyclic GMP-AMP signalling protects bacteria against viral infection" (Nature 574, 691-695, 2019-09-18, OpenAlex W2973313700, 598 citations). Not yet its own industry, but its human homolog cGAS-STING is already one of the most actively pursued drug-target classes in immuno-oncology (STING agonists in clinical development at multiple companies) -- a live example of the same "basic bacterial-defense discovery -> drug target" lag CRISPR itself went through, currently mid-pattern rather than already translated.

Faure, Saito, Wilkinson, Quinones-Olvera, Xu, Flam-Shepherd, Kim, Reddy, Zhu, Evgeniou, Koonin, Macrae & Zhang, "TIGR-Tas: A family of modular RNA-guided DNA-targeting systems in prokaryotes and their viruses" (Science, published online 2025-02-27, DOI 10.1126/science.adv9789) -- discovered via iterative structural and sequence homology-based computational mining starting from Cas9's own guide-RNA-interaction domain (see the linked precursor node), not from a CRISPR locus. Found in phage and other mobile genetic elements. Unlike CRISPR-Cas, TIGR systems have no PAM (protospacer-adjacent motif) targeting restriction and are substantially more compact, framed by the authors and subsequent commentary as a potential therapeutic-delivery advantage. Discovered in Feng Zhang's lab at the Broad Institute/MIT. Presented in the literature as a genuinely new, non-CRISPR RNA-guided system, not a Cas9 variant. Very early: 51 citations as of 2026-09-03, no independent replication or applied translation documented yet -- status set to emerging, not foundational, on that basis.

Recurrent unit with gated memory cells (Hochreiter & Schmidhuber, "Long Short-Term Memory," Neural Computation 9(8), 1997) that solved the vanishing-gradient problem in training recurrent networks over long sequences, making deep sequence models practical for the first time.

→ 1 link into the applied tree

Farhi, Goldstone, Gutmann & Sipser's 2000 proposal of adiabatic quantum computation -- a model of quantum computing based on slowly evolving a system's Hamiltonian toward one whose ground state encodes a problem's solution, rather than the circuit model's discrete gates. This is the original conceptual proposal (arXiv:quant-ph/0001106); the more-cited 2001 Science paper by an overlapping author list is a specific demonstration applying the model to random NP-complete problem instances, not the founding paper -- a real correction, not a name change. Basic-vs-Applied framing check: this node covers only the abstract computational model and its complexity-theoretic properties, not D-Wave's commercial adiabatic/annealing hardware, which has its own commercialization path and belongs in the applied tree if tracked.

→ 1 link into the applied tree

Regularization technique that randomly deactivates a fraction of network units on each training step to prevent co-adaptation and overfitting (Srivastava, Hinton, Krizhevsky, Sutskever & Salakhutdinov, "Dropout: A Simple Way to Prevent Neural Networks from Overfitting," JMLR 15(1), 2014). Became a near-universal default regularizer for training large neural networks.

→ 1 link into the applied tree

A learned, differentiable weighting over a set of representations via compatibility scores, letting a decoder attend differentially to encoder states instead of compressing an entire input into one fixed vector (Bahdanau, Cho & Bengio, "Neural Machine Translation by Jointly Learning to Align and Translate," arXiv:1409.0473, ICLR 2015). Introduced inside an RNN encoder-decoder for neural machine translation; Transformer's self-attention (Vaswani et al. 2017) is a genuine architectural generalization -- removing recurrence and convolution entirely -- captured by a downstream edge, not the same fact restated.

→ 1 link into the applied tree

General encoder-decoder framework using multilayered LSTMs to map an input sequence to an output sequence of different length (Sutskever, Vinyals & Le, "Sequence to Sequence Learning with Neural Networks," NeurIPS 2014). Established neural sequence transduction as competitive with statistical machine translation and became the architecture attention was first added to.

RNN encoder-decoder using gated recurrent units to map a variable-length input sequence to a variable-length output sequence for statistical machine translation (Cho et al., "Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation," arXiv:1406.1078, 2014). A simpler gated-recurrence alternative to LSTM, and the architecture Bahdanau et al.'s attention mechanism was built on top of.

Identity shortcut connections that let gradients skip layers directly, addressing the degradation problem in very deep networks and enabling training of substantially deeper models (He, Zhang, Ren & Sun, "Deep Residual Learning for Image Recognition," CVPR 2016). Originated in image recognition; reused structurally as a sub-layer wrapper in later sequence-transduction architectures.

→ 1 link into the applied tree

Dilated-convolution sequence transduction model computing hidden representations in parallel across all input and output positions, reducing the operation count needed to relate distant positions to logarithmic in sequence length (Kalchbrenner et al., "Neural Machine Translation in Linear Time," arXiv:1610.10099, 2016). One of two convolutional architectures the Transformer explicitly compares itself against and outperforms.

→ 1 link into the applied tree

Normalizes activations across the feature dimension within each training example rather than across the batch, removing batch-size dependence from normalization (Ba, Kiros & Hinton, "Layer Normalization," arXiv:1607.06450, 2016). Became the standard normalization sub-layer in sequence-transduction architectures, where batch normalization is awkward to apply.

→ 1 link into the applied tree

Attention model that decomposes the problem of comparing two sentences into aligning and comparing sub-phrases using only feed-forward attention, with no recurrence (Parikh, Tackstrom, Das & Uszkoreit, "A Decomposable Attention Model," EMNLP 2016). One of the few pre-Transformer models to rely on attention without a recurrent network, cited by Vaswani et al. as a direct precedent for dispensing with recurrence entirely.

→ 1 link into the applied tree

Convolutional sequence-to-sequence architecture using stacked CNNs with gated linear units for both encoder and decoder instead of recurrence, offering greater training parallelism than RNN-based translation models (Gehring, Auli, Grangier, Yarats & Dauphin, "Convolutional Sequence to Sequence Learning," arXiv:1705.03122, 2017). The other convolutional architecture the Transformer directly benchmarks against.

→ 1 link into the applied tree

Google's Sycamore processor completed a specific sampling task in ~200 seconds that the team estimated would take the best classical supercomputers thousands of years (Arute et al., 2019) -- the first empirical demonstration that a real quantum device can outperform classical computation on some task, answering a foundational question in computational complexity theory rather than delivering a commercial product. Basic-vs-Applied framing check: this node covers the scientific claim (a computational-complexity separation was empirically demonstrated), not the Sycamore chip itself or any commercial quantum-computing hardware, which have their own commercialization paths. The classical-hardness estimate was later disputed and narrowed by improved classical simulation algorithms -- a real, ongoing debate about the size of the advantage, not about whether the underlying experiment was performed.

→ 1 link into the applied tree

Framework for online function approximation via optimal polynomial projections, giving recurrent models a principled way to compress an arbitrarily long history into a fixed-size state (Gu, Dao, Ermon, Rudra & Ré, "HiPPO: Recurrent Memory with Optimal Polynomial Projections," NeurIPS 2020). The memory theory the entire S4/Mamba line of state space models is built on.

Efficient diagonal-plus-low-rank parameterization of state space sequence models, the first to make the LSSL/HiPPO line of models practical at scale (Gu, Goel & Ré, "Efficiently Modeling Long Sequences with Structured State Spaces," ICLR 2022). Mamba's own selective SSM layer (S6) is described as "S4 models with a selection mechanism."

Deep-learning layer unifying recurrent, convolutional, and continuous-time models under one linear state-space formulation (Gu, Johnson, Goel, Saab, Dao, Rudra & Ré, "Combining Recurrent, Convolutional, and Continuous-time Models with Linear State-Space Layers," NeurIPS 2021). Mamba's own Theorem 1 is stated as a direct improvement of this paper's Lemma 3.1.

→ 1 link into the applied tree

End-to-end language model architecture combining S4 with a linear-attention-style mechanism, generalizing linear attention to more general recurrences (Dao, Fu, Saab, Rudra & Ré, "Hungry Hungry Hippos: Towards Language Modeling with State Space Models," ICLR 2023). Mamba's own block design is presented as directly simplifying the H3 block.

→ 1 link into the applied tree