What's coming next
The chess AI and the racing agent are finished. Tacsmith is working on new bots that carry the same strategic reasoning and physics-based control into new kinds of games and systems. This page lists them and is updated as each one moves forward.
Completed
Chess AI
Self-play neural networks with C++ alpha-beta search. Proprietary, and available through the API.
Racing agent
Deep reinforcement learning for high-speed continuous control and optimised racing lines across many tracks.
New agent types
Tacsmith is working on new bots in each of these areas.
Combat bots
Tacsmith is working on new bots for combat: targeting, positioning, threat evaluation and tactical choices in fast, competitive encounters.
Movement and navigation bots
Tacsmith is working on new bots for movement: pathfinding through complex environments, avoiding obstacles and choosing the best route in real time, building on the racing agent's control work.
NPC behaviour systems
Tacsmith is working on new bots that give characters believable routines, reactions and personalities.
Resource and worker automation
Tacsmith is working on new bots for management and colony games that gather, build, craft and optimise production chains.
Automated testing bots
Tacsmith is working on new bots that play through levels, stress-test systems and report bugs automatically.
Adaptive strategic opponents
Tacsmith is working on new bots that adapt to each player's style while staying challenging and fair.
Platform systems
These are the building blocks that every agent type shares.
Persistent server-side agents
Today's API answers one request at a time. Next come persistent agents that run on Tacsmith infrastructure and keep their state between requests, for multiplayer and live-service games.
Customisable behaviour systems
Beyond today's preset profiles, a profile editor for tuning style, risk, difficulty and personality without retraining, and for sharing profiles across titles.
Self-serve analytics and simulation
The rating ladders, live dashboards and self-play simulation behind the chess AI, opened up to customers for their own agents and games.