Human validation: two studies, 31 participants.

21 subjects, 30 sessions (3 subjects repeated the test days apart).

Any existing lower-order aberration — myopia, hyperopia, astigmatism — was first corrected with contact lenses. Presbyopia was then induced using strong negative lenses. Lens power increased from −6 D to −9 D in 0.25 D steps until subjects could not recognise any of the 2 mm-tall characters (letters, numbers and symbols) displayed on a phone held 10 cm from their eyes.

This distance forces maximum accommodation demand and constrains pupil diameter to roughly 2.8 mm, as the screen — set at maximum luminosity — fills the subject's field of view.

Participants then read SEDF 4%AV8 3E?6, first uncorrected, then with Glassless pre-correction applied at increasing intensity in 0.50 D steps until they identified their best correction level.

Six sessions were discarded: some subjects could still accommodate beyond the maximum −9 D applied; others could not read anything even at the weakest lens tested (−6 D), meaning that lens already induced complete unreadability with no usable induction gradient.

Remaining participants went from 0/13 symbols recognised to 11/13 on average (min 8/13; max 13/13). Residual errors were systematic rather than random, confined to the finest-detail glyphs: the question mark, and 3 versus 8. Subjects' self-reported optimal correction averaged 2.83 D (min 2.00 D; max 3.50 D).

This test confirmed the validity of our presbyopia model, including the additive nature of pure defocus aberrations (myopia, hyperopia, presbyopia). Lab simulations proved accurate, replicating the eye's behaviour with high fidelity: from our mechanical eye simulations we predicted optimal correction at 2.50 D, with average optimal correction during testing reported at 2.83 D (observed range 2.00–3.50 D).

10 subjects aged 40–70, presbyopic (ranging from +1 D to +3 D), with no higher-order aberrations, screened by our ophthalmologist collaborators.

Unlike Phase 1, this phase was run as an informal pilot without a fixed protocol, to gauge whether presbyopic users benefited from the software in practice ahead of a fully protocolised study. Lower-order aberrations were corrected with contact lenses where present, to isolate presbyopia.

Each subject self-assessed to determine their optimal correction, increasing its intensity in steps until they identified their optimal level. The same correction was then applied unchanged to the whole screen while subjects performed basic phone tasks — navigating the home screen, reading calendar events — first at 10 cm, then at 35 cm, a typical reading distance.

4 of 10 subjects interacted freely with their smartphone without reading glasses, including with UI elements smaller than the 2 mm characters previously tested; the other 6 were unable to perform the tasks at all.

Successful subjects also showed low sensitivity to the change in viewing distance, with the same correction working at both 10 cm and 35 cm. Their self-assessed optimal correction was markedly lower than Phase 1's induced-condition average: roughly 1.00–2.00 D, versus 2.83 D.

The four subjects who benefited from Glassless correction were in the older age range and had more severe presbyopia. Their ability to read UI elements smaller than 2 mm also alleviated our concern that the eye's optical resolution limit might cap correction effectiveness on small print.

The lower correction intensity needed by real subjects (~1.00–2.00 D) versus Phase 1's induced condition (~2.83 D) confirms our induction protocol simulated a moderate-to-severe presbyopia. In these subjects, the same correction worked at different viewing distances: the relatively narrow 10–35 cm range produces minor changes in system magnification, indicating that the correction is dominated by defocus and pupil aperture.

The six subjects who could not perform tasks after Glassless correction presented milder presbyopia. As those subjects experience less baseline blur, the ringing and contrast-loss artifacts introduce more disturbance than benefit. Eliminating ringing and contrast loss will be essential to making the correction effective across the full severity range.