NEATEdApp

NEATEd — Watch Neural Networks Evolve

NEATEd is an interactive playground for the NEAT algorithm (NeuroEvolution of Augmenting Topologies) — implementing the genetic algorithm published by Stanley & Miikkulainen (2002) that grows neural networks node by node, generation by generation, in front of your eyes.

Train AI agents to balance a pole on a cart, navigate a snake to find food, land a spacecraft on the Moon, beat a Pong opponent, or master Tic Tac Toe. Watch their networks start as a handful of input-to-output connections and gradually develop hidden nodes, recurrent loops, and complex topologies as evolution finds better solutions.

WHO IT'S FOR
• Computer science students learning evolutionary computation
• AI educators and curriculum designers
• Hobbyists curious about how neural architecture search actually works
• Anyone who's read Stanley & Miikkulainen's NEAT paper and wanted to play with it

WHAT MAKES IT SPECIAL

• Live network visualisation — every node, every connection, every weight, animated as evolution proceeds.

• Side-by-side run comparison — pit two genomes against each other in real time. Verdict cards, trophy badges, fitness curves with aha-moment annotations, behavior radars and live action histograms tell you why one agent beat the other.

• Reward shaping editor — tune food rewards, death penalties, and navigation bonuses for each game and watch the agent's behaviour change.

• Six Evolution Strategies — Anti-Bloat, Aggressive Exploration, Conservative Refinement, Diversity Boost, Stochastic-Tuned, and Novelty Search (Lehman & Stanley 2008) for escaping deceptive optima.

• Full configuration editor — every parameter from the published NEAT formulation: mutation rates, speciation coefficients, weight bounds, initial connection modes. With one-tap recommendations when a run gets stuck.

• Built-in glossary — learn local optima vs deceptive optima the moment you encounter them.

GAMES INCLUDED

Free: CartPole, Snake.
Pro: Lunar Lander, Pong, Tic Tac Toe — plus the full Lessons curriculum, full Config Editor, Compare Runs view, unlimited daily training runs, and saving / exporting genomes.

LEARN BY DOING

The Learn tab walks you through NEAT in 25+ interactive lessons: genomes, speciation, crossover, structural mutation, complexity penalties, novelty search. Daily Challenges give concrete goals — "reach fitness 500 on CartPole", "eat 5 food on Snake" — that turn into tangible results.

BUILT IN SWIFT, BUILT FOR YOU

Native macOS and iOS app. No cloud LLMs, no telemetry, no ads. Training runs stay on your device. Saved genomes sync via iCloud Drive. Power-aware: warns you before training drains the battery and pauses if your Mac gets too hot.

DISSERTATION ORIGIN

NEATEd grew out of an MEng Software Engineering dissertation — a Windows tool that wrapped the NEAT-Python library to make neuroevolution approachable for non-coders, exposing every config option through a friendly editor and walking users through the algorithm with built-in lessons. NEATEd is that vision reimagined as a polished native Mac and iOS app, with the NEAT engine independently re-implemented in Swift, plus live animated network visualisations, side-by-side comparison, reward-shaping, and modern evolution strategies the original couldn't easily provide.

SUBSCRIPTION
NEATed Pro: £12.99 / month or £59.99 / year (with a 7-day free trial). Auto-renews unless cancelled at least 24 hours before the period ends.

NEATed is an independent Swift implementation. Not affiliated with, endorsed by, or sponsored by Kenneth O. Stanley, Risto Miikkulainen, the University of Texas, or the maintainers of neat-python. Open-source attributions are available in-app under More → Open-Source Licences.

Privacy Policy: https://github.com/venetsia/NEATed-support/blob/main/PRIVACY_POLICY.md
Terms of Use (EULA): https://github.com/venetsia/NEATed-support/blob/main/TERMS_OF_USE.md