A granularity-based dissolution of the central singularity (a finite, computed core) and a clarification of what the horizon really is — the eternal event horizon is a global idealization; the local trapping horizon survives, one-way, with thermal-only return and no escape. An audit and a framing of established physics (regular black holes: Bardeen, Hayward, Dymnikova; the horizon analysis: Hawking, Ashtekar–Krishnan, Penrose), not novel physics: no derived interior, no escape, no resolution of black-hole entropy or information.
A black hole confronts us with two monsters: the central singularity and the one-way horizon. We argue that both are tamed by a single move — treating granularity, a smallest physical length ℓ, as fundamental — and that taming them is, in the end, one act rather than two.
The singularity is a continuum artifact. Schwarzschild's curvature (Kretschmann scalar K = 48G²M²/c⁴r⁶) diverges only as r → 0; a length floor ℓ forbids ever probing that point, so the blow-up becomes a limit never attained. Replacing the idealized core with a granular one (we compute a representative Hayward profile, verified symbolically) yields a finite center, K(0) = 24/ℓ⁴ — an effective de Sitter core of curvature ~1/ℓ² — while reproducing Schwarzschild exactly for r ≫ ℓ. This is dissolution, not solution: granularity removes the obligation of a singularity, fixing the core scale (~ℓ) and finite curvature (~1/ℓ²) without uniquely deriving the interior. Regular black holes are already known (Bardeen, Hayward, Dymnikova, "Planck stars"); our contribution is the framing, the computation, and the link below. We disclose the costs — the core violates the strong energy condition (precisely how the Penrose–Hawking singularity theorems are evaded), needs an effective non-vacuum source, and carries an inner (Cauchy) horizon prone to mass-inflation instability; granularity fixes the family (finite core, scale ℓ, curvature ~1/ℓ²), not the chosen interior profile.
The horizon splits in two. The eternal, absolutely one-way event horizon — a global, teleological object requiring the entire future and a true asymptotic infinity — is an idealization that evaporation and a Λ > 0 universe already strain. (Evaporation alone leaves only a finite-lived event horizon; what fully removes it is evaporation together with a smooth, singularity-free interior.) What survives is the quasi-local trapping horizon: practically one-way over the full ~10⁶⁷–10¹⁰⁰-yr evaporation lifetime, with information returning only thermally and never by a local route. Locally featureless is not leaky: there is no local escape at any stage. The two threads then join: the granular core is exactly the smooth interior an evaporating hole needs to shed its eternal horizon. The result is a principled singularity-resolution and a clarification of what the horizon is — not a derived interior, not an escape, and not a resolution of black-hole entropy or information.
A black hole is famous for two pathologies, and they are not the same pathology. One lives at the center; the other lives at the edge. Both have been called impossible, both have been called the place where physics breaks, and for decades they have been bundled together under a single word — singularity — as if curvature blowing up and causality slamming shut were one defect of nature. They are not. This work pulls them apart, and the surprise is that pulling them apart is also how they turn out to be one story.
The first monster is the center. In the textbook Schwarzschild solution, tidal forces grow without bound as you fall inward. The clean, coordinate-independent way to see it is the Kretschmann scalar, the squared curvature that no choice of observer can talk away:
$$K = \frac{48\,G^2 M^2}{c^4\, r^6}.$$
As $r \to 0$ this diverges. Curvature becomes infinite, tidal stretching becomes infinite, and general relativity stops returning answers — it returns $\infty$, which is the same as returning nothing. This is the singularity in the strict sense: a genuine breakdown of the theory, an honest confession that the equations have been pushed past where they are entitled to speak.
The second monster is the edge. At a finite radius — the Schwarzschild radius — the geometry grows a one-way membrane. Cross it falling inward and, the story goes, you can never send anything back out: not a signal, not a flashlight beam, not yourself. From the outside, an infalling clock appears to freeze and redden forever at the boundary, as if time itself jammed. This is the horizon, and its scandal is not infinite curvature but infinite commitment — a boundary you can pass through exactly once, in exactly one direction, forever.
Two monsters, two very different crimes. One is about magnitude (a quantity going to infinity); the other is about causality (a direction becoming mandatory). Lumping them together has cost the field clarity, because the cure for one is nothing like the cure for the other.
Here is the thesis, stated plainly so the rest can earn it.
The first monster dissolves. The infinity in $K$ is reached only in the limit $r \to 0$ — and that limit is a continuum assumption, the premise that you may keep probing to arbitrarily small radius. Grant instead a smallest length $\ell$: a floor below which "smaller" stops meaning anything. Then $r \to 0$ is never attained, the divergence is a limit approached but never reached, and the center is replaced by a finite, computable core — curvature large but bounded at roughly $1/\ell^2$, with a de Sitter-like heart in place of the puncture. This is not new physics and we will not pretend it is: regular, singularity-free black holes have been written down before — Bardeen, Hayward, Dymnikova, the "Planck star" picture. Our contribution is the framing — the singularity as an artifact of taking the continuum literally, an artifact granularity simply declines to produce — together with the explicit computation of what replaces it, and the honest accounting of what that replacement costs (an effective, non-vacuum source; a violated energy condition; an inner horizon with its own instability). The obligation of a singularity is removed; a unique interior is not derived.
The second monster splits in two. The "eternal, absolutely one-way boundary" turns out to be two different objects wearing one name. One is the event horizon — a global, teleological construct that needs the entire future history of the spacetime and a true infinity to escape to before you can even say where it is. That object is an idealization. Two pressures undo its claim to be eternal. First, a hole that evaporates is not forever — though evaporation by itself only shortens the event horizon's life rather than abolishing it, since a terminating interior singularity would still wall the inside off from the outside until the very end. Second, a universe with a positive cosmological constant doesn't even supply the kind of infinity the definition requires, so the asymptotically-flat construction is not well posed there to begin with. The other object is the trapping (apparent) horizon — a local, here-and-now surface where outgoing light momentarily stops gaining ground. That one survives. It is genuinely, practically one-way for the entire life of the hole, with no local route out at any stage. So the honest statement is not "the horizon is escapable." It is: the eternal horizon was a global idealization; the real horizon is local, and it still holds you.
And the punchline — the reason these are one story and not two — is the seam between them. What lets an evaporating black hole shed its event horizon entirely, rather than merely shorten its life, is precisely a smooth quantum interior with no singular cut in it: remove the infinite-curvature center, and you remove the very thing that severs the inside from the outside world. Resolving the first monster is exactly the surgery the second monster needs. Granularity reaches the global causal structure of the hole not by softening the gentle, low-curvature edge you fall through, but by going all the way to the core and dissolving it.
A word on what this is and what it is not, set down here at the start so no later sentence can quietly inflate it. This is a principled singularity-resolution and a clarification of what the horizon is — not a derived interior, not an escape, and not a resolution of black-hole entropy or information. The center becomes finite. The edge becomes honest about being local. Nothing here lets anything climb back out, and nothing here touches the question of how a black hole stores or returns information. Those are different monsters for a different day.
Look again at where the curvature actually blows up. The Kretschmann scalar $$K = \frac{48\,G^2 M^2}{c^4\, r^6}$$ is finite at every radius except one: it diverges only in the limit $r \to 0$. And that limit smuggles in an assumption — that you may keep probing to arbitrarily small radius, that space is divisible without end. Deny it. Grant a smallest length $\ell$, a floor below which "closer to the center" stops meaning anything. Then $r \to 0$ is a place no probe and no worldline ever reaches; the divergence is a limit approached but never attained, and the puncture at the center is replaced by a finite core.
To make that concrete, take a representative regular profile (Hayward's, the minimal one with a clean de Sitter center and an exact Schwarzschild tail): $$f(r) = 1 - \frac{2 m r^2}{r^3 + 2 m \ell^2}, \qquad m \equiv \frac{GM}{c^2},\quad \ell \sim \ell_{\min}.$$ Computing the curvature directly (verified symbolically) gives a center that is finite, not infinite: $$K(0) = \frac{24}{\ell^4} \quad(\text{finite}), \qquad f(r)\big|_{r\to 0} \approx 1 - \frac{r^2}{\ell^2}.$$ That near-center form is an exact de Sitter core — a small, smooth, positively-curved heart of curvature $\sim 1/\ell^2$ where the singularity used to be. Far away the granular correction vanishes identically: for $r \gg \ell$, $K \to 48 m^2/r^6$, the exact Schwarzschild exterior. And the black hole is still a black hole: $f(r)=0$ has a genuine outer horizon for $m > m_{\rm crit} = 3\sqrt{3}\,\ell/4 \approx 1.3\,\ell$ (with an inner root as well).
The tell that this is purely a continuum artifact: send $\ell \to 0$ and the divergence returns continuously, $K(0) = 24/\ell^4 \to \infty$. The infinite center was never a feature of the physics; it was a feature of taking the continuum literally. Remove the continuum and the center is finite, bounded, and computable.
A finite center is not a free lunch, and honesty requires putting the bill on the table.
It is a chosen profile, not a derived one. Hayward's metric is one member of a family — Bardeen's, Dymnikova's, the "Planck star" profiles all share a finite de Sitter core and a Schwarzschild tail. Granularity fixes the family (that there is a finite core, of scale $\sim\ell$, with curvature $\sim 1/\ell^2$); it does not single out the member. We computed a representative interior, not the interior. Claiming a uniquely derived center would be exactly the overreach we refuse.
It violates the strong energy condition — by design. At the core the effective stress-energy gives $8\pi(\rho + p_r + 2p_t) = -6/\ell^2 < 0$ (with $p_r = -\rho$): a negative-pressure, de Sitter-like region. This is not an embarrassment to be hidden; it is precisely how the Penrose–Hawking singularity theorems are evaded, since those theorems assume the strong energy condition. No regular core exists without paying this price.
It needs an effective, non-vacuum source. The granular core is not a solution of vacuum general relativity; it requires some effective stress-energy (nonlinear-electrodynamic or vacuum-polarization-like). Granularity motivates a finite center but does not, by itself, hand us the field equations that produce this particular matter content. That gap is real and unclaimed.
It carries an inner horizon, and inner horizons are fragile. The second root of $f(r)=0$ is a Cauchy horizon, and Cauchy horizons generically suffer the mass-inflation instability — so the static regular interior is probably not the end of the dynamical story. (Below threshold, $m < m_{\rm crit}$, one gets a horizonless ultracompact remnant instead — its fate is its own open question.)
So the honest status of the core is: a principled, computed, finite replacement for the singularity, with its costs disclosed — not a vacuum solution, not a unique interior, and not guaranteed stable.
The second monster wears its terror on the outside. Where the singularity threatens with infinite curvature at the center, the horizon threatens with a wall: a surface of no return, a place where — the textbook cartoons insist — time freezes, light stalls, and a falling astronaut is stretched against an invisible membrane. The claim of this section is that almost none of that is real. The horizon has nothing local about it and everything global. There is no place there. There is only a property of the whole.
Take a black hole large enough to matter and ask what an infalling observer actually measures at the horizon. The answer, dictated by the equivalence principle, is: nothing. The equivalence principle says that free fall is locally indistinguishable from floating in empty space — a freely falling observer feels no gravity, only tidal effects, and tidal effects are governed by curvature, not by where any coordinate happens to blow up. At the horizon of a large black hole the curvature is gentle. For a supermassive hole the tidal stretch at crossing is weaker than what you feel standing on the Earth. There is no debris, no membrane, no jolt. The observer sails through a patch of spacetime that is, locally, unremarkable vacuum, and cannot locally detect the crossing at all. No experiment performed in a small enough lab around that event returns a different answer inside than outside.
So where does the "barrier" come from? From a bad choice of map. The Schwarzschild time coordinate $t$ degenerates at the horizon: in that coordinate, infalling light takes infinite $t$ to arrive, redshifts to nothing, and "freezes." But $t$ is the clock of a distant static observer, not a property of the place. Switch to coordinates regular at the horizon — Eddington–Finkelstein, or Kruskal — and the apparent pathology vanishes completely; nothing diverges, nothing freezes, worldlines pass through smoothly. The frozen-time picture is a coordinate artifact, the gravitational analogue of the meridians crowding together at the pole on a Mercator map. The pole is a fine place to stand; the map just can't draw it. Likewise the horizon is a fine place to fall through; the static map just can't follow you there.
And yet the horizon is one-way. This is not a contradiction — it is the whole point. One-wayness is not a local property and was never going to show up in a local measurement. It lives in the global causal structure: in how the light-cones are arranged across the spacetime.
The clean statement is geometric. Outside the hole, the time-translation symmetry is generated by a Killing vector $\partial_t$ that is timelike — it points along the worldlines of observers who can hover at fixed radius forever. As you approach the horizon, $\partial_t$ turns null; exactly at the horizon, the would-be hovering direction has become a light ray. Inside, $\partial_t$ goes spacelike. "Staying at constant radius" is no longer a possible motion for any material body — it would require moving faster than light. The light-cones, which outside lean comfortably toward larger $r$ and toward the future, tip over as you cross, until inside the hole every future-directed cone points inward. Decreasing $r$ stops being a choice and becomes the meaning of "later," as inescapable as tomorrow. Nothing pushes you in. The future simply lies that way.
This is why the crossing is undetectable locally yet irreversible globally. Locally, each light-cone is an ordinary light-cone; you can always move freely within it. Globally, the cones are arranged so that the union of all your allowed futures never again touches the outside. No single local move is forbidden; the composition of all of them is.
The cleanest way to hold both facts at once is a knot in a rope.
Run your fingers along a knotted rope. Every segment, inspected on its own, is just rope — straight, smooth, locally identical to any other piece. Nowhere is there a "knot point," a special spot where the rope becomes different stuff. Examine any small neighborhood and you would never know the rope was knotted at all. The knottedness is not in any segment. It is in how the whole rope threads through itself — a property of the entire configuration, invisible to every local probe, and one that no local manipulation can undo: you cannot untie a knot by working on one inch of rope.
The horizon is the knot. Locally, every patch is featureless low-curvature vacuum — rope. Globally, the light-cones thread so that the inside cannot signal the outside — the knot. The infalling observer feels nothing because there is locally nothing to feel; the one-wayness is real because it is woven into the causal arrangement of the whole. Locally trivial, globally one-way.
A caution carries forward from this. Showing that the horizon is locally featureless is not showing that it is escapable. The knot analogy makes the point precisely: a featureless local segment is fully compatible with a globally inescapable structure. The horizon being "nothing local" removes the wall, the freeze, the membrane — the parts that were never physical. It leaves the one-way causal structure entirely intact. What that surviving one-wayness is, how much of it is eternal idealization versus durable local trapping, and why it returns information only thermally, never by a local route — those are the questions of the next section.
We have dissolved one monster. The singularity was never a place; it was a limit the continuum kept promising and granularity refuses to honor. Now turn to the second monster — the one-way wall — and ask the same kind of question. Not "how do we cross it?" but "what exactly is it?" The answer is that the textbook horizon, the truly eternal one, is a different kind of object than it advertises: not a local feature of spacetime but a global verdict rendered over all of history. And once you see what that verdict actually requires, you start to doubt the universe ever signs it.
The horizon that means "nothing escapes, ever" has a precise definition, and the definition is the giveaway. The event horizon is the boundary of the region from which light can never reach infinity:
$$\mathcal{E} = \partial J^-(\mathscr{I}^+).$$
Read that carefully. $J^-(\mathscr{I}^+)$ is the causal past of future null infinity — the set of all events that can send a signal out to $\mathscr{I}^+$, the asymptotic "edge" of an idealized flat universe infinitely far away and infinitely far in the future. The event horizon is the edge of what is forever sealed off from that edge.
Two features of this definition matter more than the geometry. First, it is teleological: to locate the event horizon here and now, you must already know the entire future of the spacetime. Whether the light leaving your hand this instant ever reaches infinity depends on everything that will ever happen to it — every gravitating mass it will pass, the ultimate fate of the hole, the structure of spacetime a googol years hence. The event horizon is not measured; it is adjudicated, retroactively, with the whole future in evidence. No local experiment, no finite patch of spacetime, can ever tell you where it is. Second, it is global: it is defined by reference to $\mathscr{I}^+$, a feature of the spacetime as a whole — and one that has to genuinely exist for the definition to mean anything.
Both features are about to fail.
The first failure is the famous one. A real black hole is not eternal: it radiates. Hawking's 1974 result gives it a temperature, a luminosity, and therefore a lifetime — a stellar-mass hole evaporates in roughly $10^{67}$ years, a supermassive one in something like $10^{100}$. The hole ends. So the "forever" baked into "$\mathscr{I}^+$ never receives a signal from inside" collides with the plain fact that there is a last moment of the hole's existence. A horizon defined by eternal causal disconnection is in tension with an object that does not last.
The second failure is quieter and, in some ways, deeper. The definition leans on $\mathscr{I}^+$ — a clean, timelike future null infinity of an asymptotically flat spacetime. But our universe carries a positive cosmological constant $\Lambda > 0$. In a de Sitter universe the asymptotic structure changes character entirely: future infinity becomes spacelike, and every observer is wrapped in a cosmological horizon of their own. The neat asymptotically-flat $\mathscr{I}^+$ that $\mathcal{E} = \partial J^-(\mathscr{I}^+)$ depends on is not merely modified — the construction it requires is arguably not well-posed in the spacetime we actually inhabit. The reference point against which "escape" is defined isn't where the global definition assumes it is, and may not be the right kind of object at all.
So the eternal event horizon rests on two idealizations — unending duration and a true flat infinity to escape to — and the real universe honors neither. This is the philosophy-of-physics point, stated plainly: the eternal horizon is a continuum idealization, a global construct that is exact in a toy spacetime and approximate-at-best in ours. It is to the wall what the divergent curvature was to the core — a feature of the idealization, not necessarily of the thing.
Here is where a careless argument would overreach, so we will not. One might be tempted to say: the hole evaporates, therefore the eternal horizon is gone, full stop. That is false, and the reason is instructive.
Suppose the hole evaporates but the interior still terminates in a singularity — a sharp, spacelike edge where spacetime simply stops. That singularity is itself a causal cutter: it severs the interior from $\mathscr{I}^+$ just as effectively as an eternal horizon would, for as long as it is there. Evaporation shortens the sentence but does not overturn the verdict. What you are left with is a finite-lived event horizon — the global sealing-off still happens, it just happens for $10^{67}$ years instead of forever. The teleological boundary still exists; it has merely been handed an expiration date.
So evaporation removes the eternal in "eternal horizon," but by itself it does not remove the horizon. Something has to dispose of the causal cutter inside.
This is the moment the two threads of the paper become one thread.
What does it take to fully remove the event horizon — not shorten it, remove it? You need an interior with no terminating singularity: a smooth quantum interior through which causal curves can pass and, in principle, reach the outside region once the hole is gone. No sharp edge, no causal cutter — just a finite, regular core that lets the spacetime knit back together. This is precisely Hawking's 2014 reading: with a smooth interior there are no event horizons, only apparent horizons — the global, teleological boundary never forms as an eternal object at all.
And we have already built that interior. The granular core of Section [core] is exactly the smooth, singularity-free region this argument was asking for. The same smallest-length floor that capped the curvature at $K(0) = 24/\ell^4$ and turned the singularity into a finite de Sitter heart is what removes the causal cutter that an evaporating hole would otherwise leave behind. Resolving the core is not a separate achievement from dissolving the eternal horizon — it is the same achievement, seen from the other monster's side. One result, two faces: heal the inside, and the eternal outside boundary has nothing left to anchor to.
Note carefully how granularity reaches the horizon, because it is not the obvious route. Granularity does nothing to the gentle, low-curvature region near a large hole's horizon — that neighborhood stays as featureless and crossable as the equivalence principle demands, and we will hold to that in the next section. Granularity reaches the global horizon structure through the core, by removing the deep-interior singularity that the eternal horizon secretly depends on. The wall is dismantled from the inside, not the surface.
Be exact about the ceiling. We have argued that the eternal event horizon — global, teleological, defined against an idealized $\mathscr{I}^+$ — is a continuum idealization that the real universe (evaporating holes, $\Lambda > 0$) does not realize, and that resolving the core is what finally lets it go. That is a claim about a global, idealized structure.
It is emphatically not a claim that black holes are escapable. The locally-real, quasi-local horizon — the trapping surface where outgoing light momentarily stops outrunning the geometry — survives this entire argument untouched, and it is practically one-way over the full $10^{67}$–$10^{100}$-year lifetime. That is Section [survives], and nothing here softens it. Dissolving the eternal event horizon removes a teleological idealization; it opens no local door. Information still returns only thermally, through the Hawking radiation and the Page curve, never through a route a falling observer could take.
What we have, then, is a clarification of what the horizon is — a global verdict the universe may never actually sign, dismantled through the very core resolution that tamed the singularity — not an escape, and not a resolution of black-hole entropy or information.
Dissolving the eternal event horizon does not hand you an exit. What remains is the boundary that was doing the real work all along: the trapping (apparent) horizon, and it is a very different kind of object from the one we just retired.
It is defined locally. Flash light outward from a small sphere and watch its expansion; where that outgoing expansion drops to zero ($\theta_{\rm out} = 0$) you have a marginally trapped surface, and the trapping horizon is their boundary. This is a here-and-now, gauge-checkable construction (Ashtekar–Krishnan) — it needs no knowledge of the infinite future and no true asymptotic infinity, so none of the idealization-attacks that undid the event horizon touch it.
And inside it, you are genuinely stuck. Outgoing light itself converges — heads inward despite being aimed out — so escaping would mean outrunning light, which nothing does. The trapped region is one-way for everything. Not "absolutely, forever" (that was the event horizon's overclaim), but practically one-way for the entire life of the hole: $\sim 10^{67}$ years for a stellar-mass black hole, up to $\sim 10^{100}$ for a supermassive one. The universe is $\sim 10^{10}$ years old. For any conceivable purpose, that is absolute.
The only thing that ever comes back is thermal: the slow Hawking glow, a scrambled, near-featureless radiation emitted over that entire evaporation lifetime. The Page-curve and island results say the information does eventually return — but encoded in the late radiation, non-locally and unrecognizably, not as the infalling object climbing back through. Locally featureless is not leaky: a free-faller crosses the boundary feeling nothing and still cannot turn around. No eternal horizon is not escapable. The robust survivor is exactly this — a real, local, lifetime-long, one-way trapped region whose only outflow is a thermal whisper.
Two monsters guard a black hole, and we have treated each on its own terms.
The singularity dissolves into a finite core. In the continuum theory the Schwarzschild interior carries a real divergence — the Kretschmann scalar $K = 48\,G^2M^2/c^4r^6$ runs to infinity, but it does so only in the limit $r\to 0$. A smallest-length floor $\ell$ forbids ever reaching that limit, so the divergence becomes a place the geometry never visits rather than a place it breaks. What replaces it is not hand-waving but a finite, computed core: taking the representative Hayward profile $f(r)=1-2mr^2/(r^3+2m\ell^2)$ as a stand-in for the resolved family, the curvature at the center is finite, $K(0)=24/\ell^4$, the core is effectively de Sitter ($f\approx 1-r^2/\ell^2$, $\Lambda_{\rm eff}=3/\ell^2$, $R(0)=12/\ell^2$), and for $r\gg\ell$ the metric reproduces Schwarzschild exactly ($K\to 48m^2/r^6$). The scale and curvature of the core are set by $\ell$; the singularity was a continuum artifact that granularity declines to inherit. This is not new physics — regular black holes are an old idea (Bardeen, Hayward, Dymnikova, "Planck stars"). What is ours is the framing and the computation: the singularity as an idealization that a smallest length removes, with the core's family fixed even though its specific member is not.
The eternal horizon is a global idealization. The event horizon $E=\partial J^-(\mathscr{I}^+)$ is a global, teleological object: defining it requires the entire future history and a genuine asymptotic infinity $\mathscr{I}^+$. Both inputs fail in the real universe. Hawking evaporation contradicts "forever" — a stellar-mass hole is gone in $\sim 10^{67}$ yr — and a universe with $\Lambda>0$ has a spacelike $\mathscr{I}^+$ and a cosmological horizon, so the asymptotically-flat construction is not even well-posed. The honest caveat: evaporation alone does not erase $E$, because a terminating spacelike singularity still severs the interior from $\mathscr{I}^+$, leaving a finite-lived event horizon. What removes it completely is evaporation together with a smooth quantum interior — and that is precisely the core we built. The two threads are one: resolving the singularity supplies exactly the smooth, non-cutting interior that an evaporating hole needs to shed its eternal horizon (Hawking's 2014 "no event horizons, only apparent horizons"). Granularity reaches the global causal structure through the core, not through the gentle near-horizon region.
The local horizon survives, and it is genuinely one-way. Locally the horizon is featureless: for a large hole the curvature there is low, a free-faller crosses feeling nothing (the equivalence principle), and cannot locally detect the crossing at all. The "barrier" and "frozen time" are Schwarzschild-coordinate artifacts — Eddington–Finkelstein and Kruskal coordinates are perfectly regular across it. But locally featureless is not leaky. What endures is the quasi-local trapping (apparent) horizon — the marginally-trapped surface $\theta_{\rm out}=0$, gauge-checkable in the Ashtekar–Krishnan sense — and it is practically one-way over the entire evaporation lifetime, $\sim 10^{67}$–$10^{100}$ yr. One-wayness is causal and global: light-cones tip until the future points inward, and the static Killing vector $\partial_t$, timelike outside, goes null on the horizon and spacelike within. Like a knot, every local segment of the rope looks ordinary while the rope is globally tied; no local move undoes it. Information returns only thermally — through Hawking radiation and the Page curve, with unitarity restored via the early radiation (islands / quantum extremal surfaces), never through a local channel. There is no local escape at any stage.
What we have not done. The interior profile is a chosen ansatz: granularity fixes the family (finite core, scale $\ell$, curvature $\sim 1/\ell^2$) but not the member. The resolved core comes at real cost — it violates the Strong Energy Condition ($8\pi(\rho+p_r+2p_t)=-6/\ell^2<0$, with $p_r=-\rho$), which is exactly how the Penrose–Hawking singularity theorems are evaded; it requires an effective, non-vacuum source rather than pure GR; and it introduces an inner (Cauchy) horizon prone to a generic mass-inflation instability. Below threshold ($m<m_{\rm crit}=3\sqrt3\,\ell/4\approx 1.3\,\ell$) there is no outer horizon at all, only a horizonless ultracompact remnant. And black-hole entropy ($S=A/4$) and the mechanism of the Page curve are separate objects we do not touch here.
So the two results are genuinely one — resolving the core is what an evaporating hole needs to lose its eternal horizon — but the claim must be stated at its true ceiling: a principled singularity-resolution and a clarification of what the horizon is — not a derived interior, not an escape, and not a resolution of black-hole entropy or information.