2025-10-03 - 2026-10-03
Overview
156 Issues closed from 1 user
Closed
#177 Build ModularBot_garage skeleton
Closed
#178 Battle test ModularBot against varied enemies
Closed
#174 Implement oscillator movement module
Closed
#175 Define Radar trait and implement radar harness
Closed
#176 Wrap radar_lock as radar module
Closed
#171 Implement circular gun module
Closed
#172 Extract and adapt Tsetlin Machine as gun module
Closed
#173 Define Movement trait and implement movement harness
Closed
#168 Implement gun selection logic
Closed
#169 Implement head-on gun module
Closed
#170 Implement linear gun module
Closed
#166 Define Gun trait and WorldState type in Nim
Closed
#167 Implement virtual bullet tracker in harness
Closed
#163 Decide geometric would-have-hit error signal
Closed
#162 Decide input encoding for enemy velocity
Closed
#161 Decide test adversary — wall bot or spinner for training
Closed
#160 Research bullet mechanics — speed, travel time, lead angle formula
Closed
#164 Research existing lead-targeting code in repo
Closed
#147 SNNBot: SNN-aimed gun against stationary target
Closed
#158 Implement SuperSpike learning rule in SNNBot
Closed
#157 Research PC-SNN as backup learning algorithm
Closed
#156 Research SuperSpike learning rule for SNNBot gun aiming
Closed
#155 Prototype reward-modulated STDP learner
Closed
#153 Set up integration test for SNNBot
Closed
#152 Prototype SNN core and aiming loop
Closed
#151 Decide debug overlay layout and content
Closed
#150 Decide learning loop structure
Closed
#154 Prototype hill-climbing learner
Closed
#148 Research Tank Royale debug drawing API
Closed
#149 Research aimTo() mechanics — gun turn limits and API
Closed
#146 Iterate on movement quality based on test results
Closed
#145 Wire DangerZones_Bot integration test with shared framework
Closed
#144 Build minimal GF gun test adversary
Closed
#142 Rebase danger-zones-bot onto master
Closed
#143 Delete old battle test harness
Closed
#135 Make test framework LLM-ready (docs + examples)
Closed
#139 Add OscillatorBot as a second shared adversary
Closed
#140 Update AGENTS.md / CLAUDE.md with test framework quick-start
Closed
#137 Move SittingDuck to shared location
Closed
#138 Add integration test to a second garage (QBot)
Closed
#136 Document exceptions, parseServerOutput, and compile behavior in README
Closed
#126 Wayfinder: Shared integration test framework for garages
Closed
#133 Modify BattleRunner.java to connect to external server and emit structured output
Closed
#134 Implement BattleRunner Java process lifecycle in Nim framework
Closed
#131 Create example integration test in one garage
Closed
#132 Write framework usage guide
Closed
#130 Implement runBattle() orchestrator
Closed
#129 Implement bot compiler and launcher
Closed
#127 Define BattleResult type and server output parser
Closed
#128 Implement server lifecycle manager (lazy singleton)
Closed
#123 Mutation sigma annealing
Closed
#124 Decouple aim() from tickGA - compute input vector once per tick
Closed
#125 GMM-based GA warm-starting from battle history
Closed
#120 Input cleanup: remove redundant distance from sliding window
Closed
#121 Tracker extraction: VBulletTracker becomes VBulletEnvironment
Closed
#122 Enemy-motion-aware bullet simulation
Closed
#119 Input cleanup: remove redundant distance from sliding window
Closed
#108 GA_Gun only: virtual bullets, tick-speed training, hit% fitness
Closed
#114 Battle-wide hit% diagnostic
Closed
#112 Real gun fires champion output
Closed
#113 Weight persistence on new champion
Closed
#111 Wire GA generation loop at tick speed
Closed
#109 Extend ANN to 2 outputs (GF + power)
Closed
#110 Implement virtual bullet system
Closed
#99 Multi-gun EvoBot: three guns, one selector, global-only weights
Closed
#94 Persistence v2: fitness header, best-gating, atomic writes, binary-dir location
Closed
#95 Round-end checkpoint wiring
Closed
#93 Champion transport carries fitness
Closed
#92 Spec: EvoBot round-end checkpointing — best-so-far per-enemy weights
Closed
#74 Evo_Bot implementation design decisions
Closed
#88 EvoBot weight checkpointing: never lose the best per-enemy weights
Closed
#61 Evo_Bot — neuroevolution gun that adapts to enemy movement
Closed
#97 fix: persistent bot thread to eliminate ORC heap corruption
Closed
#98 fix: add receive timeout to WebSocket to prevent silent hangs
Closed
#96 Checkpoint acceptance sweep & hard cases
Closed
#91 Lock the trigger set and write rules
Closed
#89 What does "better" mean for best-so-far weights?
Closed
#90 Which save moments can EvoBot actually act on?
Closed
#80 Spec: Evo_Bot TOPO_Gun implementation
Closed
#87 Energy management and firing discipline
Closed
#85 Weight persistence and hybrid seeding
Closed
#86 Virtual Guns system
Closed
#84 TOPO_Gun with GF targeting
Closed
#83 Sliding window, replay tape, and normalization
Closed
#81 Bot skeleton with head-on firing
Closed
#82 ANN + GA modules with self-checks
Closed
#78 Energy management and firing discipline
Closed
#79 Weight persistence and global fallback strategy
Closed
#77 Virtual Guns bookkeeping
Closed
#76 Bullet power formula and aiming geometry
Closed
#75 Input normalization strategy
Closed
#68 Where and what granularity should we instrument the Nim training thread with chunk-start/end logs + heartbeat to distinguish hang vs crash vs slow for corpse stall diagnosis?
Closed
#60 Campaign-v2 pacing & durability decisions
Closed
#66 Prototype: GA evolution loop on toy prediction problem
Closed
#63 Document the locked architecture as an ADR
Closed
#64 Build OscillatorBot sparring partner
Closed
#62 Define CONTEXT.md vocabulary for Evo_Bot domain
Closed
#65 Research: optimal GA parameters for ~750-weight neuroevolution
Closed
#59 Campaign v2 readiness: implement approved levers
Closed
#58 Campaign-v1 watch: eval-regression threshold fired — intervention deferred to verdict
Closed
#57 Campaign verdict: notebook, results table, scale-or-not call
Closed
#56 Lock campaign-v1 config and launch overnight run
Closed
#54 Build mirror-twin sparring partner + harness readiness check
Closed
#55 Observability inventory: what can we watch mid-run?
Closed
#37 spec: SAC_LSTM_Bot — Recurrent SAC with LSTM for Robocode TankRoyale
Closed
#49 Training harness
Closed
#48 Main bot integration
Closed
#52 Tag v1.0.1 and publish to nimble registry
Closed
#51 Set up GitHub Actions CI workflow
Closed
#50 Implement Tier 1 SharedTestRunner for upstream JSON fixtures
Closed
#47 SAC training module
Closed
#46 Weight persistence module
Closed
#45 Replay buffer module
Closed
#43 Action mapping module
Closed
#44 Reward module
Closed
#42 State vector module
Closed
#40 Skeleton bot with radar lock and colors
Closed
#41 LSTM network module
Closed
#38 SAC_LSTM_Bot: project scaffold
Closed
#39 Radar lock: standalone reusable module
Closed
#24 PRD: Command abstraction layer for PPO action space
Closed
#28 PPO_Bot wiring and integration test
Closed
#18 Command abstraction layer for PPO action space
Closed
#29 [Map] Autonomous LLM-supervised PPO_Bot training pipeline
Closed
#31 Define the tunable parameter set
Closed
#34 Build the training runner tool
Closed
#30 Hyperparam delivery: env vars vs compile-time constants
Closed
#35 Persist Adam optimizer state and round counter across restarts
Closed
#32 Structured output format for LLM consumption
Closed
#33 Stall detection: hardcoded heuristics vs LLM judgment
Closed
#36 LLM supervisor agent design
Closed
#27 Action decoding rewrite for command abstraction
Closed
#26 Network and state vector dimension changes
Closed
#25 Goto and AimTo controller functions
Closed
#23 Prototype goto controller and test in game
Closed
#22 PPO training loop compatibility
Closed
#21 State vector additions
Closed
#19 Output space design for goto/aimTo
Closed
#20 Goto controller algorithm
Closed
#1 RL Bot — Wayfinder Map
Closed
#13 Spec: PPO RL Bot — full implementation
Closed
#17 Weight persistence + background training thread
Closed
#16 Reward + trajectory + GAE + PPO training
Closed
#14 Network forward pass → bot moves via neural network
Closed
#15 Enemy tracker + full 42-float state vector
Closed
#11 Bot scaffold and build setup
Closed
#12 Learning rate and optimizer hyperparameters
Closed
#10 Exploration noise
Closed
#9 Network architecture
Closed
#8 Training loop design
Closed
#6 Reward function design
Closed
#5 Action space mapping
Closed
#4 State vector design
Closed
#7 Enemy tracker design
Closed
#2 Arraymancer viability spike
Closed
#3 RL algorithm choice
184 Issues created by 1 user
Opened
#1 RL Bot — Wayfinder Map
Opened
#2 Arraymancer viability spike
Opened
#3 RL algorithm choice
Opened
#4 State vector design
Opened
#5 Action space mapping
Opened
#6 Reward function design
Opened
#7 Enemy tracker design
Opened
#8 Training loop design
Opened
#9 Network architecture
Opened
#10 Exploration noise
Opened
#11 Bot scaffold and build setup
Opened
#12 Learning rate and optimizer hyperparameters
Opened
#13 Spec: PPO RL Bot — full implementation
Opened
#14 Network forward pass → bot moves via neural network
Opened
#15 Enemy tracker + full 42-float state vector
Opened
#16 Reward + trajectory + GAE + PPO training
Opened
#17 Weight persistence + background training thread
Opened
#18 Command abstraction layer for PPO action space
Opened
#19 Output space design for goto/aimTo
Opened
#20 Goto controller algorithm
Opened
#21 State vector additions
Opened
#22 PPO training loop compatibility
Opened
#23 Prototype goto controller and test in game
Opened
#24 PRD: Command abstraction layer for PPO action space
Opened
#25 Goto and AimTo controller functions
Opened
#26 Network and state vector dimension changes
Opened
#27 Action decoding rewrite for command abstraction
Opened
#28 PPO_Bot wiring and integration test
Opened
#29 [Map] Autonomous LLM-supervised PPO_Bot training pipeline
Opened
#30 Hyperparam delivery: env vars vs compile-time constants
Opened
#31 Define the tunable parameter set
Opened
#32 Structured output format for LLM consumption
Opened
#33 Stall detection: hardcoded heuristics vs LLM judgment
Opened
#34 Build the training runner tool
Opened
#35 Persist Adam optimizer state and round counter across restarts
Opened
#36 LLM supervisor agent design
Opened
#37 spec: SAC_LSTM_Bot — Recurrent SAC with LSTM for Robocode TankRoyale
Opened
#38 SAC_LSTM_Bot: project scaffold
Opened
#39 Radar lock: standalone reusable module
Opened
#40 Skeleton bot with radar lock and colors
Opened
#41 LSTM network module
Opened
#42 State vector module
Opened
#43 Action mapping module
Opened
#44 Reward module
Opened
#45 Replay buffer module
Opened
#46 Weight persistence module
Opened
#47 SAC training module
Opened
#48 Main bot integration
Opened
#49 Training harness
Opened
#50 Implement Tier 1 SharedTestRunner for upstream JSON fixtures
Opened
#51 Set up GitHub Actions CI workflow
Opened
#52 Tag v1.0.1 and publish to nimble registry
Opened
#53 Overnight Training Campaign — SAC_LSTM_Bot
Opened
#54 Build mirror-twin sparring partner + harness readiness check
Opened
#55 Observability inventory: what can we watch mid-run?
Opened
#56 Lock campaign-v1 config and launch overnight run
Opened
#57 Campaign verdict: notebook, results table, scale-or-not call
Opened
#58 Campaign-v1 watch: eval-regression threshold fired — intervention deferred to verdict
Opened
#59 Campaign v2 readiness: implement approved levers
Opened
#60 Campaign-v2 pacing & durability decisions
Opened
#61 Evo_Bot — neuroevolution gun that adapts to enemy movement
Opened
#62 Define CONTEXT.md vocabulary for Evo_Bot domain
Opened
#63 Document the locked architecture as an ADR
Opened
#64 Build OscillatorBot sparring partner
Opened
#65 Research: optimal GA parameters for ~750-weight neuroevolution
Opened
#66 Prototype: GA evolution loop on toy prediction problem
Opened
#67 Wayfinding: Recover Hope via Rigorous Check on Open Issues & Training Harness
Opened
#68 Where and what granularity should we instrument the Nim training thread with chunk-start/end logs + heartbeat to distinguish hang vs crash vs slow for corpse stall diagnosis?
Opened
#69 Should we configure ulimit -c unlimited + coredumpctl capture in the service unit, and treat core dump recurrence as a blocker (stop campaign) or data point (continue with monitoring)?
Opened
#70 What explicit criteria should define 'verified' persistence fix (e.g., buf_size > 0 in first 3 metrics after restart AND training_state.bin rewritten within 2 min)?
Opened
#71 Is chunk size=10 final, or should we re-attempt 20 with the persistence fix (using a measurable criterion like 'no corpse stalls for 2h at chunk=10')?
Opened
#72 Should eval interval remain at 5, or define a metric to re-evaluate (e.g., "if eval variance > X, drop back to 3")?
Opened
#73 Should UTD=3 and batch=32 remain final, or define a trigger for re-evaluation (e.g., based on throughput metrics)?
Opened
#74 Evo_Bot implementation design decisions
Opened
#75 Input normalization strategy
Opened
#76 Bullet power formula and aiming geometry
Opened
#77 Virtual Guns bookkeeping
Opened
#78 Energy management and firing discipline
Opened
#79 Weight persistence and global fallback strategy
Opened
#80 Spec: Evo_Bot TOPO_Gun implementation
Opened
#81 Bot skeleton with head-on firing
Opened
#82 ANN + GA modules with self-checks
Opened
#83 Sliding window, replay tape, and normalization
Opened
#84 TOPO_Gun with GF targeting
Opened
#85 Weight persistence and hybrid seeding
Opened
#86 Virtual Guns system
Opened
#87 Energy management and firing discipline
Opened
#88 EvoBot weight checkpointing: never lose the best per-enemy weights
Opened
#89 What does "better" mean for best-so-far weights?
Opened
#90 Which save moments can EvoBot actually act on?
Opened
#91 Lock the trigger set and write rules
Opened
#92 Spec: EvoBot round-end checkpointing — best-so-far per-enemy weights
Opened
#93 Champion transport carries fitness
Opened
#94 Persistence v2: fitness header, best-gating, atomic writes, binary-dir location
Opened
#95 Round-end checkpoint wiring
Opened
#96 Checkpoint acceptance sweep & hard cases
Opened
#97 fix: persistent bot thread to eliminate ORC heap corruption
Opened
#98 fix: add receive timeout to WebSocket to prevent silent hangs
Opened
#99 Multi-gun EvoBot: three guns, one selector, global-only weights
Opened
#100 Strip per-enemy persistence, go global-only
Opened
#101 Upgrade ANN to 91→16→8→1 two hidden layers
Opened
#102 Implement CMA-ES module
Opened
#103 Implement GF_Gun — basic guess-factor histogram
Opened
#104 Implement virtual guns selector
Opened
#105 Wire multi-gun system into EvoBot
Opened
#106 Write ADR-0002 and update CONTEXT.md
Opened
#107 Spec: Multi-gun EvoBot — three guns, one selector, global-only weights
Opened
#108 GA_Gun only: virtual bullets, tick-speed training, hit% fitness
Opened
#109 Extend ANN to 2 outputs (GF + power)
Opened
#110 Implement virtual bullet system
Opened
#111 Wire GA generation loop at tick speed
Opened
#112 Real gun fires champion output
Opened
#113 Weight persistence on new champion
Opened
#114 Battle-wide hit% diagnostic
Opened
#115 Spec: GA_Gun only — virtual bullets, tick-speed training, hit% fitness
Opened
#116 Spec: Phantom Meteor Gravity Movement — bullet dodging via learned danger field
Opened
#118 Enhance EvoBot GA gun system
Opened
#119 Input cleanup: remove redundant distance from sliding window
Opened
#120 Input cleanup: remove redundant distance from sliding window
Opened
#121 Tracker extraction: VBulletTracker becomes VBulletEnvironment
Opened
#122 Enemy-motion-aware bullet simulation
Opened
#123 Mutation sigma annealing
Opened
#124 Decouple aim() from tickGA - compute input vector once per tick
Opened
#125 GMM-based GA warm-starting from battle history
Opened
#126 Wayfinder: Shared integration test framework for garages
Opened
#127 Define BattleResult type and server output parser
Opened
#128 Implement server lifecycle manager (lazy singleton)
Opened
#129 Implement bot compiler and launcher
Opened
#130 Implement runBattle() orchestrator
Opened
#131 Create example integration test in one garage
Opened
#132 Write framework usage guide
Opened
#133 Modify BattleRunner.java to connect to external server and emit structured output
Opened
#134 Implement BattleRunner Java process lifecycle in Nim framework
Opened
#135 Make test framework LLM-ready (docs + examples)
Opened
#136 Document exceptions, parseServerOutput, and compile behavior in README
Opened
#137 Move SittingDuck to shared location
Opened
#138 Add integration test to a second garage (QBot)
Opened
#139 Add OscillatorBot as a second shared adversary
Opened
#140 Update AGENTS.md / CLAUDE.md with test framework quick-start
Opened
#141 Wayfinder: Battle-ready danger zones movement
Opened
#142 Rebase danger-zones-bot onto master
Opened
#143 Delete old battle test harness
Opened
#144 Build minimal GF gun test adversary
Opened
#145 Wire DangerZones_Bot integration test with shared framework
Opened
#146 Iterate on movement quality based on test results
Opened
#147 SNNBot: SNN-aimed gun against stationary target
Opened
#148 Research Tank Royale debug drawing API
Opened
#149 Research aimTo() mechanics — gun turn limits and API
Opened
#150 Decide learning loop structure
Opened
#151 Decide debug overlay layout and content
Opened
#152 Prototype SNN core and aiming loop
Opened
#153 Set up integration test for SNNBot
Opened
#154 Prototype hill-climbing learner
Opened
#155 Prototype reward-modulated STDP learner
Opened
#156 Research SuperSpike learning rule for SNNBot gun aiming
Opened
#157 Research PC-SNN as backup learning algorithm
Opened
#158 Implement SuperSpike learning rule in SNNBot
Opened
#159 SNNBot: lead-predicted aiming against moving targets
Opened
#160 Research bullet mechanics — speed, travel time, lead angle formula
Opened
#161 Decide test adversary — wall bot or spinner for training
Opened
#162 Decide input encoding for enemy velocity
Opened
#163 Decide geometric would-have-hit error signal
Opened
#164 Research existing lead-targeting code in repo
Opened
#165 ModularBot — Modular multi-gun bot with pluggable harnesses
Opened
#166 Define Gun trait and WorldState type in Nim
Opened
#167 Implement virtual bullet tracker in harness
Opened
#168 Implement gun selection logic
Opened
#169 Implement head-on gun module
Opened
#170 Implement linear gun module
Opened
#171 Implement circular gun module
Opened
#172 Extract and adapt Tsetlin Machine as gun module
Opened
#173 Define Movement trait and implement movement harness
Opened
#174 Implement oscillator movement module
Opened
#175 Define Radar trait and implement radar harness
Opened
#176 Wrap radar_lock as radar module
Opened
#177 Build ModularBot_garage skeleton
Opened
#178 Battle test ModularBot against varied enemies
Opened
#179 T1: Tracer bullet — Gun harness + head-on gun + bot skeleton
Opened
#180 T2: Add linear gun module
Opened
#181 T3: Add circular gun module
Opened
#182 T4: Add movement harness + oscillator module
Opened
#183 T5: Add radar harness + wrap radar_lock
Opened
#184 T6: Extract Tsetlin Machine as gun module
Opened
#185 T7: Battle test ModularBot against varied enemies