When did plate tectonics begin on Earth, and what preceded it? Published thermo-mechanical mantle evolution models imply that the early history of planets with a composition and size similar to Earth and Venus should be characterized by periodic mantle overturns of 30–100 Myr duration, separated by stable lid phases of 100–300 Myr. This is best described as an unstable stagnant lid because this term captures the Jekyll and Hyde duality of such worlds, which alternate between a stagnant lid sensu stricto phase between mantle overturns, and a mobile lid phase during overturns. Mantle overturn upwelling zones would rework and resurface large tracts of pre-existing Hadean crust and basalt-dominated Archean-style oceanic lithosphere (ASOL). Basal anatexis of ASOL c. 40–50 km thick (or melting within down-drips) could generate tonalite–trondhjemite–granodiorite (TTG) melts and create proto-continental nuclei, while garnet pyroxenite restites delaminate into the mantle. With further reworking, low-K tonalitic rocks would remelt to produce granodiorite and granite, completing the transfer of radioactive elements out of the lower crust. Mantle overturns would generate large-scale lateral currents in the upper mantle that would push against Archean-aged sub-continental lithospheric mantle keels, causing continental drift and orogenesis despite the absence of plate boundary forces such as slab-pull. This is corroborated by the observed displacement of Lakshmi Planum (>1000 km) on Venus, a planet with no arcs nor ridges. Recent models suggest that the Abitibi Greenstone Belt formed as an oceanic tract behind a detached ribbon continent during the partial break-up of the Southern Superior Craton and may be a sample of Archean oceanic lithosphere. The Abitibi Greenstone Belt has c. 50 km of apparent stratigraphy composed of 2–10 Myr mafic–felsic bimodal volcanic cycles that follow assimilation–fractionation trends indicating the contamination of mantle-derived basalts with TTG-like anatectites derived from older basalts. ASOL of this type would be difficult to subduct due to its weakness and buoyancy, but would be fertile and could generate large amounts of second-stage melts. There are no sheeted dykes, precluding a seafloor spreading model, while the absence of basal cumulates or attached mantle means that this type of ASOL should not be called an ophiolite. Archean–Proterozoic unconformities are followed by the deposition of iron formations, clastic and volcanic rocks, which are only rarely affected by sagduction. The increase in siliciclastic input and decreasing sagduction reflect near-global Late Archean emergence of stiffening granitic continents from the oceans due to secular cooling and intra-continental differentiation. Albeit associated with continent-derived siliciclastic debris, many Paleoproterozoic volcanic (and plutonic) rocks resemble Archean rocks geochemically. The similarity of magmatic rocks and hot orogenic styles in the Archean and Paleoproterozoic could imply that the overall geodynamic regime was similar in both. The Siderian–Rhyacian quiet period could therefore represent a stagnant lid phase that followed the 2.5 Ga Archean overturn. When the next mantle overturn ruptured the lid at c. 2.2–2.0 Ga (and again at c. 1.9–1.8 Ga), continents would have been set into motion, forming arcs and ridges. Once initiated, arc and ridge segments needed to multiply and propagate to create a world-girdling system. Mesoproterozoic rocks preserve clear evidence of plate mobility, subduction and orogenesis, but, ophiolites, the geological record of seafloor spreading, are extremely rare prior to 1 Ga. The Earth at 2.0 Ga was probably still largely covered by ASOL, possibly similar to the Abitibi Greenstone Belt, but how and where it was all destroyed and replaced by modern oceanic lithosphere are mysteries. Given the volume of ASOL involved, recognizable by-products of this global-scale reworking process should exist. Voluminous anorthosite–mangerite–charnockite–granite/gabbro suite rocks are mostly of Proterozoic age, requiring either an ephemeral source or a unique process. Trace element inversion models applied to massif anorthosites imply they crystallized from high-La/Yb melts that do not resemble tholeiitic basalts, invalidating the notion that they are flotation cumulates from basaltic underplates. Model anorthosite-forming melts can, however, be explained by high-pressure melting of an ASOL-like basalt source with garnet-bearing residues. I posit that massif anorthosites record the destruction of an ephemeral source (ASOL) at Proterozoic convergent margins. When the last ASOL was crushed between converging continents or consumed by an overprinting arc ( c. 0.8–1 Ga), anorthosite–mangerite–charnockite–granite/gabbro suite rocks ceased to form and the Earth became a modern plate tectonic planet.