GF-T · a ternary-native float that beats the published state of the art at range
φ² + 1/φ² = 3
2.84× and 5.53× more accurate than tekum16 at range · no regime decode · exponent adds natively on ternary
φ² + 1/φ² = 3
2.84× and 5.53× more accurate than tekum16 at range · no regime decode · exponent adds natively on ternary