The Landauer floor at the horizon (k_B T_H ln 2) is ~10¹² below the CMB floor, but that is a disposal cost, not a delivered one. A carrier photon aimable at the horizon needs energy E_gamma ≥ hc/r_g; charging realistic delivery losses against that floor is what sets the realized gain.
An event horizon lowers the Landauer erasure floor from the CMB temperature to the (far lower) Hawking temperature T_H — a ratio of ~10¹² at IMBH masses. But k_B T_H ln 2 is the marginal cost of disposal at the horizon; it is not the delivered cost from an orbiting platform. A carrier photon must be aimable at the horizon, which sets a minimum energy E_gamma ≥ hc/r_g where r_g = GM/c². At the paper's fiducial mass (2×10⁴ M☉), r_g ≈ 3×10⁷ m and E_gamma ≈ 7×10⁻³³ J — a factor ~4×10⁹ below the CMB floor of 2.6×10⁻²³ J, not the ~10¹² floor-to-floor figure.
Charging realistic engineering losses (pointing/capture inefficiency, error-correction overhead, scattered-vs-absorbed carriers) against the ideal floor — one to three orders of magnitude — leaves a defensible realized gain of 10⁶–10⁹ over the 2.7 K sky. At the default mass, bits/photon, and loss (2 dex midpoint) this tool reads ≈4×10⁷, inside that band; the slider's full 1–3 dex range spans ≈4×10⁶ to ≈4×10⁸, also inside it. Denser coding (more bits per carrier photon) is permitted up to the Bekenstein channel capacity but yields only modest further gains at this energy scale — moving the slider confirms the gain shifts by less than the loss slider does for the same log-decade change.
The Tolman/gravitational-blueshift bookkeeping is stated in the paper as an O(1) correction at platform radii r ≳ 100 r_g and is not separately computed here — modeling it further would go beyond what §4.2 states, so this tool reports the paper's qualitative conclusion (it does not change the quoted range) rather than inventing a redshift model.
This is the delivered-cost correction to the floor-to-floor Landauer comparison in Bekenstein-Landauer-Lloyd Limit Explorer, which quotes the ~10¹² CMB/horizon ratio as a disposal-floor figure. That figure is real and unchanged; this tool exists because the disposal floor is not the same number as the realized gain a real delivery channel achieves.