Never mistake the
$MaryJ contract address with another.
MARYJkS2kHxmVkQp8baNUL7d3R1pddLr65GgsArJAYY
MARYJ (5-Prefix) JAYY (4-Postfix)
Both ends are vanity-grinded — deliberately mined out of the Solana keyspace, not random.
The math is harder than the naive count suggests:
Solana addresses are 32-byte ed25519 public keys, base58-encoded into 43- or 44-character strings.
Only ~5.5% of addresses encode to 43 chars; 94.5% encode to 44 chars.
The letter "M" is mathematically blocked from the first position of every 44-char address — it can only appear at the start of the rare 5.5% that encode to 43 chars.
Real probability of an "M" first-char is ~0.001, not 1/58. So P(MARYJ prefix) ≈ 0.001 × (1/58)⁴ ≈ 1 in 11.2 billion, not the naive 1 in 656M.
The "JAYY" suffix is uniform: 1 in 58⁴ ≈ 11.3M. Joint expected probability: ~7.8 × 10⁻¹⁸ → 1 in 128 quadrillionattempts at the median, ~17× harder than the naive 58⁹ count almost everyone cites.
Grinding this on secure CPU only — using the official solana-keygen grind with verified CSPRNG entropy, because GPU vanity tools can leak keypairs through GPU memory side-channels and we refused to ship a contract address whose entropy we couldn't trust — takes ~71 million machine-hours on a 32-core EPYC at ~500k keys/sec.
That's $3.5M–$5.5M in self-hosted electricity alone, or $80M–$120M in cloud CPU compute at expected solve.
About 8,250 years on one machine, or ~99,000 servers running in parallel for a month.
Nation-state-grade keyspace burn for a meme. Almost no project bothers. This is the level of detail a signet of perfection deserves.