Hexanaut Github [repack] May 2026

Leo had been staring at the terminal for 17 hours. His Hexanaut bot—a sprawling, hexagonal territory-capture algorithm—kept failing on the third expansion wave.

Hexanaut wasn't just a game. On the private GitHub repo hexanaut-ai/hex-core , it was a simulation of geometric conquest. Each hex cell represented a server node. Each border push mimicked a DDoS wave. The goal? Hold the largest contiguous cluster while starving enemy daemons of processing cycles.

He opened the repo again. 47 forks. 12 open issues. One pull request titled: "Feat: Dynamic territory reallocation via min-cost flow" hexanaut github

And then he watched.

His bot—now named HexVector-1 —didn't charge forward. It retreated . It gave up three border hexes to consolidate power. The enemy overextended, starving for resources. Then, in one devastating turn, HexVector-1 reclaimed twelve hexes in a single loop—a legal move the game engine hadn’t seen in three seasons. Leo had been staring at the terminal for 17 hours

He clicked through. The contributor, @hexVector , had rewritten the scoring function. Instead of maximizing cells held, they minimized distance to supply hubs —a classic supply-chain hack turned into a combat edge.

Leo smiled. He forked the repo again, added a single line to the README: And somewhere in a server farm across the ocean, HexVector-1 expanded one more hex—quietly, greedily, perfectly. On the private GitHub repo hexanaut-ai/hex-core , it

Leo’s bot was brilliant—except for one flaw. It didn't understand sacrifice .