{"approach": "A 224-48-48-3 ReLU MLP (two hidden layers of 48, 13,299 parameters, Gemm/Relu with fp16 initializers Cast to float, 27,117 bytes, outputs raw [yaw, throttle, shoot] that the sim clamps). Made by self-play from Starter, run 3 of 5, every random choice seeded with 3: (1) behaviour cloning of Starter (numpy, Adam lr 1e-3, 20 epochs, 1.16M Starter decisions from 3,000 random-spawn matches against Starter, the current rule and the built-ins); (2) OpenAI-style evolution strategies (48 antithetic pairs, sigma 0.02, Adam lr 0.01, centred ranks, 16 shared-seed matches per generation) against a league of its own past versions (snapshot every 20 generations), the current rule, Starter and the four built-ins; one phase against the current rule alone; then a balanced league (per generation 6 current, 3 Starter, 3 past selves, 4 built-ins) for 17 minutes. Final model chosen on held-out seeds among candidates. Harness: the repo's sim crate (SimState, same loop as run_match without replay frames, randomize_spawns on) built with Rust by rustup under .cache and zig cc as linker (no C toolchain on this machine); Starter and the current rule run natively (current rule transpiled from its ONNX graph, outputs identical to onnxruntime on 1,063 decisions; Starter within 2.4e-6), about 3,000 matches/s on 32 cores. Exported ONNX checked against onnxruntime: identical outputs on 21,063 inputs, all finite. entry_sha256 is the sha256 of the base64 string, as in the current entry's file.", "blocked": null, "current_vs_starter": 0.9926666666666667, "records": {"starter": {"matches": 3000, "score": 0.9570833333333333}, "builtins": {"matches": 800, "score": 0.924375}, "current": {"matches": 1000, "wins": 6, "losses": 993, "draws": 1, "score": 0.0065}, "exploiter": {"matches": 500, "score": 0.008, "how": "Same 224-48-48-3 MLP, started from the Starter-imitation network and trained by the same ES (seed 3) for 9.5 minutes (about 2,000 generations) with the frozen fp16 final model as its only opponent; its last parameters, 500 fresh paired-spawn matches, half from each side. It wins 496, loses 4, yet scores only 0.018 against Starter: a narrow specialist, which shows the final model is easy to exploit."}}, "entry": "CAgSEGltZC1zcC1zdGFydGVyLTM6z9MBChoKA1cxaBICVzEiBENhc3QqCQoCdG8YAaABAgoaCgNiMWgSAmIxIgRDYXN0KgkKAnRvGAGgAQIKGgoDVzJoEgJXMiIEQ2FzdCoJCgJ0bxgBoAECChoKA2IyaBICYjIiBENhc3QqCQoCdG8YAaABAgoaCgNXM2gSAlczIgRDYXN0KgkKAnRvGAGgAQIKGgoDYjNoEgJiMyIEQ2FzdCoJCgJ0bxgBoAECCiYKA29icwoCVzEKAmIxEgJnMSIER2VtbSoNCgZ0cmFuc0IYAaABAgoOCgJnMRICcjEiBFJlbHUKJQoCcjEKAlcyCgJiMhICZzIiBEdlbW0qDQoGdHJhbnNCGAGgAQIKDgoCZzISAnIyIgRSZWx1CikKAnIyCgJXMwoCYjMSBmFjdGlvbiIER2VtbSoNCgZ0cmFuc0IYAaABAhIMc3Bfc3RhcnRlcl8zKpCoAQgwCOABEApCA1cxaEqAqAExOg495T8zuH0uQTvDt3G2sTfsO2i2q6/gtUUtrLsvu2yx6TMlNZm7XrM4uvW9FSVlPg67KbmGO5Q8ab71uGw6k7nhujE7NLpNvkI83zyNOvG/cjXdNsm97q6fOLw7YbOvruC8RrwZvye8ZrnIsI23iTydsIi4GbgvPVmxWDNLuPU6gDxksyM9SyhDvNa8eblqthAsC7IxPTtAY7a6vAe+z7lfOO26pq1MQd5Abi1pPFCzFDYeNEo8K7yntNQ9Jr+sOMA8EbQoOn81vTa/OAi0MbyKPjg81TJxvbS3YTw9umK4JLinO8y4hbqUuOeyLTWrsSI51ShCvgA+SLumOFE8rT6OOLA+QEDfrDA8IrupQFrAYDuvtwI8Fi9HvOs8KjMQs4TAcLZOOkW2RrDQsp89BK5ZPOU4hLzcuXe1x7xVORc0VToeuJG4NzTfMqQymL2IuXk7ULwEvjioIbUDP4w/YrUAOyi6QcFwvEwy6bXfues4ujxLO/k7O0B0OtEpez47P8M4JjFgvhctmDvQvYS0XzJMOis9d8BAtJI8jTmYPNS3Crw2OoK9Ua7KOUO0pzKyOFC+WLxLNLqyu7jCsMA14zK8PN1BH0CBubA7mzoXsBc5qbz5Ofo/VTnLNrM8D7M4NgkxSrVdOxk5fbxoOT+37zr+umu6Dbb1uqm8szlLN2k9jrmEvS+9RDCUMe6/4i3LOeK+kzo4PhA04T7ePta9UD0Nu30/FTx7vLWyVDWmugK2BjkJvVUsdTfesO+4tTWfuzU97ycAwGw4U7zONOs7R7vNvSyvtTifuAm9cLKDPES0mTyPvk+3VDL5OtM2WDzIONk1mbz4sZk20bnlvAoolq4XuksoIbtQtEO84D9sM/ey47TLvKC7vzuHOnC9ALt6vdm30zWmtY05mbjds6A8mTLStW6mbThON+M0Iz2fOKa8+jiLOqa1OTcTO5u7Yz3WLx+2yrVIuak4HaYcHvs62bitNDM5Dy0eLMs8fjw9uWmoJD9GOMi8G7c4PaK8YTzluLO96zcrsd26kTw+N5U5DrqntbytQ7haMfq24DiPs7y0/zgupSW8fr2RNhk7DjxeOqo8H7j3u5o2aLh9N4A3kLxaPoK1kjs4uuQ8RrbYuzc0SL0cuDGwoDGRLiowqagIqiw4pTiBsss4SrRHQKE82bT/tD0ygj29t+m8P6iZPRA+xrQ5vxe9QjTAuEwlYzZLv0Q8YrqhMVwuxbtZLkoxITXduEKyBDxbNhEu9rYFN3w9TbzwKcs2Q7IHJCMveTAQsDS8czcnudu3sLj8N5M2SryPufw73blQO0S50DBrMfu1/zZfuxMvzjxHsBRA8LYvPsM057wrsgc4JCGIOQq6FEDTvPI5sDTGNI821bkvOmguiDUHrX05EbgJKn2wcjuUvDgxrLytNna8LjdzvAgsfbSOuue6vDrvrXO7p5/fPL6vBL3isrgrCjf5Nbm9NjTQuD65Qzk1LNc/EblrtPw4tD0nu6G4OziSOek+EzdevDk8qi44PAa4UDWbMw82kbaMPCm8oz8rMo08vjrAMaa6KLlLOKE9ATS1vX2415//u+24urIkP3G3gj10tFm0nqqosui6azscNiw2Z7IgNsGwY7twwFw9LbtwMkk1ybAOPnY6a7lpNyK+bbRwwA+7ZrncvNU8Saz8tRa3Jbr/ukG9HzWoPgY7B7EFuC4y8j++s82+DT4YvkMwjjyrNFm2wTkcsei2NLFFvtCnDr4JPwY7SDq6Ois5XLoEKDE/yrOxv9Y8SicMO3Q4MLDGpu+sXC4XQDI4qzlZPSe/HbXPPyw5crs/tks417hVwEy5cDlBsXSxG7w8Pdu2FkACp+q8Rqi9Nt84cDzkODC8ErDPvVk2viOxvNQ6OjgQv/m16jdypgg9pL0dOK6/f7lxtoi5QbqWvVc9srNyM+AzebweOaI6vDkcuGG9fa2/rEe/izlJvHC2LjwEvY08BMADPVQ447hVPJK8rb17wUc5sT2asZg3yD5ztIc39DuxNqU4qTZ1PH80JD/pqao4GLHgsvimtzaQPMowXi4+s6AwmTknNpm3lbhTvK8wMzzqt/uz5TflwFg9BLEqwE8+Jz1GtDg4d7gHOCSs2zDJPEW84zZBNBK9p71VKR286K2Uqtc7hzdot4E9ZTSaPSC6uLxFPvI7JyuvuvY3ay4iP3G5z7Kasec2qrlmPTKs37NVtJczZjymPQs8Ob6VrYg0RqmHwkc4Ij2Lsq+62jVANtw4jDdWtfu+SDnQvgO6G7cStVM7RD4vMkA+/69UPMy+bDyPNti/yrCAs0qsIbgKPLu62j6TvJ493aN0wJs0LbbaNFc6oTz/PCg1ozVNuPe56DwSMwO4Q74nPWe3nLRuvns2dLwHOBK7lzkgsVo60zMdvR801z26NiC5MbsuuEUtK7gFuLvAZjVkPIq5ALiKPP45rb7lPHq5MjmsPA68CLZGQEy3KzjFPUm0L7qQve05YbcpvIW83TQqug8uaDqzO4w7+DP/vAmwLr60Pxk4Oz8IO8mzEDsTwmk26zQjsCw92D7BOF0+WT1AOHyzrTtbvVE5pTUFNkC//LnEtIcveblkMS21xSw1u+SwObKrqy27ay+yuSG+Zb0oPuK3HMGVrAc9bi8QrLa3bjpMtBG8XzAruuk5/7eIOW6+cbEAuzk9lzhKP2I7SiXivCc5IL63qju2bjkfuIo6wTr4OkM5MznoPHQ4BzfJubE9lbxjPEq0cbGtNUwvvrhSLpi8s7+wtSS8SL9aMA0vZzO3P0U4Jj0yvas5McBvvX42mryTtOO4Zj05OMuvGsAOtsw8dTquuwY8qjuPvEvCj7GCuAOSYD0tNVu7D7xxQFs0uDr9tiy7q7q0t+quHDtqtz04LzzlOrY9jD7DMPkoaLxgvJ2z4DV0PmexMrpgPK05ksAcOGswiTxJwHCsyju9vmi56z6ru1i+U7cjr0s5jLZNOee3ozm7Ouu9zkFjuJy5NTvOs800yzxJrLO1ZjvNtE67g6q0uKK4BroZu7mybzGCuIk4PblPoNM/aLeQwO45P7SoOje2KzYOuJ21HbAgum86jDwavWK6jzAwusS/OrlvveEwdTTKOSSqSbyNNzs7v7wiQZG8PjpcOee5nTXytmI9xLcyJpq8oDTcuAs8OLRIwHuy9zgKOOC5ibhIPaatIzyfOJExXLn2ss4nKbbfuKsX47gtupK+Hr9HtvY52rorsCY3mbzPPnKyQjlzO9Y5OTzaOAe3/LmnuRW5PjnbPJw9lTyhO0a8L70vPCrAoC1OmJg6ObZMvfM6Ybt5steyXK+sOBS8B7letIu8pMAQOl24irFGLoY8G7/6OPOuijZsviszu7lyNS85hb3WuCC5BTeOP6q7qz9vuUC7wCgfPQa6KTdiP7a9BT6BtrA4r7aTOwI66L5mvanB/7dbvAK1ezy4NngsDro8sls/RrT6ueG8sjLFu986HDSFORk8QTgMOcWiYzX6utTAeLF5tDw3IT4Ws7G6rLyCuEWzZbdTOZMhvTt5PiZB87tjPDg9qLAurtM9HrCCOm4+xTrlLpS74T1ENFa2KrrmNfY0qzoTvTK8Pr+6u908OzD5vFezTT7guyY8jjrlOFw/hrbgOuO9kyk0PHi5wbqtvrMz1b7Duy6ynLrctBY7hD5FNdW9ub+CPBepALyHOCu/sTZrOuy9NzwQvBm9b7Qhu4Q+BrSVOm27hDkDu1W7HEAkuk2+hTUFvPY0Fb1rNeGqvjhcOC86SbyWuOGy+Td2vcEzfzSLPmU7tbhPPDi+CsDDvkS6yjTZvmE7+jCRupY5WDYCnCO9vbJts9437rCbQSG3lDhQwAa8LbDZuim8IjCXNic8HzoSO1a8tTGiuC47fbsnqOM4vy7rPxs8CDnHsaq5qrrCu52wTTYOwB86Z76pudAxKixFrW2tLjiBPeq8z7yDvhw8Zrh9uOA6zrZMvrU3Jo5sPLI4hjKEvHG4nbm6ORy+mTWCuD28uzvGuHyqNzXfPK862TcBuGKpWL2hMaIxKzXTuKE8CbsCtBk86jodrUiqDTl3t486Fyk+OPu04Tm2Oc0zOb2jsYW3wjo5Ok6507CdtkFBuzhMuMS7PbW0uIO4Vy4bPDi877l1OTe9rMCVMMy5FLotLvu4kilfwKA8tj+Es0M6ojZuL8G2DTzLsms+ajDpNuU9GqGNNmu0EDxsufo4bLUIOvYu4bWOMzWylTzfMqNAwDxOIoI9TUAVQbY8aDYnOZa/HTkoPb4xaT1Ut34zVD3KuSs7MD46t+U6WDnjsgO4dTBXPMU0OzSmMRW6NbSpvMm3izx3OhE+KzrLMvq877k7ulwyx7wQuk291TE+PNTA4rSTPVO27ayhuDu2AbwkJmc0tj8jugC1ui2zO6O6ujfIOx8+pjl3M1ex2jQBNkg6Urw7vP28pS9WwBI+nrsxOCmc7K90tlY80ihqM8i207WevmY6mTyWtbyzWzgcOe6+MKwDK+Cxh7dQOTe+gMBwuBK9/LyPNuo7lrtYOpY10SaUv5G9FzLlntQ8Z6haNO8uxTxHPCy8/i6jO5m+ELaTt10vf72rNqismrvjvcG2W7xsObi6vjdWuWY8fDQUOIUvwcDquLw2RzzTvEq70j42uFI4nboAv/e9arnROn6467fav5C+D7YROvigFr3WuE64OzdNuF+xH7llpws5ZK/FOaO7cDy6M924nzECtYE2qLI5ORs31jiIuW28CMBDvVsvGTh2OjU0eryYOoq5LLyOOQY23y8cuRWuPrVMNDorVrmetx+6ADkTuE299zuMQFExTj5HucI1xDuFu0i/gDFhvf838jv+MuS4Kzy6uzA+VbkauVM+izd5ue+037rJPvMvOLPcOi2p5r4mvow31rjCOrs//jWFLtG7m74LPPY2tLKOLVAuDbf5N5EwXSuhPaamzaowutm8GsAePCi7Pb2co6s/Pb0Hvqi5g753PA4/qbvUu+o2Vzl3rEu44kBVt9W8QrjJubC8VcB3toiu+bxDvEi8IzhfOO+5FDkOtRg8DDqGLJU31zxsruc9jraCvYy++zd9qW02A7VHvZ653jnovVc8QB0MObs5zz40s60lJb/3NPe4c70gOTC+tLdEs649irqiPl29NjvKPeWx1baotmo/XSgtOTwzVrd9PDu1Qj0tpE071jZHNw27KzEYvd+1NzrSNto4Ba/RtSO2iT3uwPYucz+BtJK+OrzWseS+Ur0UOx243zCzuPa0WjPoqf+y+KWOuPi8Srp6u6E/QjQBvisvCqrdPB6+eb+vMIU6Rz0pJxe8+T+ctv+6oToRPXC9IzsVvyC39bVTMV+zOrteuGY+Cb0cOmo5WDRjwKw4djzcNTO8Dj27vVcwd7wLuru8KUGgOfy9oDn+sKk5Ia+rNDs9cjEMvcK6VTsLrg82KjwyLym0r7VSuEM6vTR2rA01TDmAPAA4AS1wtyuwg7UFPte2DjD/N3awKj1Ptk2+KDmWwY25j6tcNCs9xbkJrBU9uzODv0a7UTvfoRu4j78vMWQ1GzhuPQIjADsHPIw88jWvPVE1ebW6tOxAoTqovPa2p6XrNrA8PEHVsFy6wjF1Or47BrjFMoc94btgvhk2V7zbqhM5xrgfPka98rvmN1k6prevtCa3gCxPupO88T8xPCY2D8DXOmi9JLzBuwO4i7hLPC24Wjz9vc6e0zwyuby1FDC9OKw4l6gltFY5mrxpum05Ajg7MDg6Xy9IOwa+MbMSM2I9DDr0Nd276bYrt101EDtHnuCtMTj3vaU4RbecuSY4BrOTqpakzLOCOXY5aDh+PHe2uCynp8U3czLwvYM5Bj3juV69wTTVvaU+wrwNtqa2ub47uQk7E0DWrDC0iDysqOw3A0MSQjY4nzxXOcu5mTNvOQE9+zL9u5I4Mqt4tnyrSbt8LjC8Kzxkun68WTIfMAG41b5htDRAdzaAtgy+cC7MwHE12jnotQK9aDwfvNa23bQZuqq6kriUuCM6NL4HPhE5hS3OvZY8bzkrOhixM6vXPVE5argrPDMzTr4WuF62BzRVum0vQLuIvQ5A8DgONDQ6dEH+OzY8fbyDuEA2HjvoNUK8/L6XuaQtvLtTv7q81L0NONk4FjlFMng727vyvgOnRL8EuRWsSLEYOE+3OTxOt+K8krEyuYIzCjxRPB46wbXqvJks/bwusr0/mLZEPos2fDJrN6A8pzokunW6ATG8sKC5PrtTu2k8ubZzOMc1Q7fONqEkXT02Ntm0oyC9vkAwfzU0u0c6ubmSJgSzhj5WP82Y9T4xPAm5yKzFvW+8PTgKv4ax1z99ucg29q8QrW65l7euMMW8db4nsgY8tbcyrp8xhDtPviw0VjYwuE46e7xquWM5IrVLsFA6TbbJO567E7WmNXMvGTsgtEW6cqraObS1ljNgqZKt5DK4MtY4tLQnP/Q8ILywtOw9yLW1uX0+zjzePTw9FTg/tlM8SjXyPEe/DjXAOqG8bTYHtNy41DoOuPG8iz0Lvcg8r7zotBYxArzpMN89ksBJOmu7yzzmMXU74qXsOgE8PDJ7tnu4fby3NEK7zjL0sNS/Brl8sTw7o7jttrm/PLQcMlI8SrySN7M8/DwhP2y3DDqdNTG7y7WGudG5yTiNOLixXbI/vZk3WbnSNvy2uEAQPbu07jetONw4rz5YOQe8Db/oNCxAqjlhsUa8oCNNNnE40ydaOwi1MqwGPdW8SjwduFw/bbFPNp2ymjnQPBi2FbPZN0G4pj1KsJe5YbhNvfI0E7UIs1G4fLrgMK27F7g2OKY3uSyKvy651z1QvCC6aDy2uqA4+T3muIex6LwQtm4+gLzGOR84eDiYN5u4lDZVNzBART3VQG48PLberWm9iTuruu4kZz0jOcg48J4KwM2+WjVuN3u8EDdYrEgwYjM8OY83+jZTuKu7VzxEO1E8yLAtOai8QTbqvZM8hb0UuZQzTzlxOU8vUjqavDqxZzq+MQy617ypOZk2w635M6o03r3BuN+5DzaEtRK7bcBNtNC0kDuKNLS6uzgZOoc4Ay3DsGa65rCnuoM6FD1qNEY43bf7v1K8ATcPuTo3AD0wMvy9F0C7NaM8cLtEMxxABLrTPJy1wLsBNzmrz7FfwA08c6qrrcW5ND/Bsr49UTHZv8G4CL0LPBm0NbyrHDg27zgLuyY7Dj4CsQ65qT9cPHI4E73YuaG4VzkzOy9AFbmMO8G43jw1Pc88wrZ4OqQ7GrrzNHwuQ7zIukk4/bkhuLY8Aj5gvOQ1KbhANzg1GykHPzk3orwNs9eyALneuOi4ZrnLsHy8YDfoN4s69D81u2k73Lr1FKe74CbAt50/9r6oOHS7SzsFNeY6Xbs4v721zEBAPpk6BbhytZU7Xb3RtqK5SLfavE45ZzTJPD822bZDOoS89LKSun65JKretKWw+jyovJA4EDquNqlAPz7gsdg60DLZO2M8ujxNtBY7Qi9osiqw27p+OHg327nqPyosBTlvN9W5bTllNjk8JS/pLMM1XbxyOLSubr7/uHy5+TkqstQ2+Ll9O8y8lT0OPzdA9jkVNoA3krvNvEAvA7gtuPe5Sr3wtli1VLedvCS8e7V/tlIuT7mgrSc537liPTiw3zRZOTI5QDoKvP4tCT3GMpi5Lbk9P/U8F8HFLl2tbTydt/+7RTdgNqE8HjwtPV64MLbFOv037azzOJ8xaLIEPG4y7LNaN+U6YDZgOF26c7N1OzTArDliPoY6c7lbLA5ALr5qMm60MTX2N905GTg5OlxA6bVovr87Kx2pPsjAAjbJPMe5BbHFugM1L7eUuoI7FzdhL+e6lbbOML47RLV1OTG+4jzYNgJAOrzFOGqqwDRhv+C9U7Lxtlk5r7+ouGLBmas9wQw567u3PHG9mDWftiY9+bxjM0G2iUA5sQc+E778tTq8gjxOuQygFTVquZMzSr1IqWM4n6kcuCG727RRN8o1BDy+uLW+tzx+Ks+2QbvjNSk38Lq/rBYwX7ttOnQv/LsdQMazVq2NtI48TbRVutC6jzvVuPI8OLxbPTC72h9VrpIwhrjPLG2p47naPl+zMbxMPD8x1LQVvMq2A7mtMhcsUTeOI9awDLwyspm2dbZ0uFwzUC6IN6wxezLqOMkxx7q2wNq3cLyIvP04D73uM9Mv3rbLOh84OTILtAm53bmpsuVA0Tfgvo04jjkJtvs12DvtwHMzpK/TPfE8IaXkLJY3gbWoPIM4ILIRui6+Wb7vNDM0r7Z6usgvpzYWqYQ1Gz5IPhOqWTwIvRi+YblFv+w05rpPvrQ7xLmkOJU0xDcYPZg5WL9gMc4/TzoItUa21bCmvuq8Bz0PMyo5iDacvAe7+TlCurEvvjr2Ow26MzekOncxoDsbP1e81zzotmQxXr1aPkEzwTw2LDu4cje7sQo4x70KOm2z8TYIPS4+IDfpPny63LkuOcG6E7h5N5k5jzQ1uhQ9Ey3NsXS/uEDnvwoxJsAjPF+4oTs0OA40vDzPuL87Jy85O0xBGzBVOtur+zX7ILsdwL+UvlCsDTihufk4CECjuOQ5EKaNuVuw1D9LvIi3xDC1tIWmAjFwPYU4lTwpOXe82zxtMRk9h7lSvVc6n71Ft+o55zlCvzk8hrx4NvM6lrf1uCA7kbvsMxk7ibLZvIO2tjIWvEBAwb53PZ+9hCzdPnE5DLbYNDYv5zl/vBEygTs2OKC8sTLTsio7xSw3PLG7wTkYuls04zXcsw44h7hFr18mxzRLujk0mkCOsQs4p76zu/g/P7yBuDhAJ7l/P7i60bd9Pca15TVxN8o+urgIKeU7Lr7NLyutw77vOIU4bDrIMQm91KazuPY5WbkSuLzBNLnNPb87zbQ+vxG8ILZGIeW3jzq4u+8ljDvyN328bD4yu5QsUZmYMoawlDgFPVQ+LbmpvXavgDpbPkayOSYFNl2sPx1XuSI4OsDMM1cx6LMFO7IypK6YPAQ9C7sHP9a5nSxcO/87TLIQvSkxP61nOMm7dLWZtAS947d+vZK15DLLOVsznjGtNpKzajpMv2S4Dr7JOSQ6BDbGqFU8mz2jrTI+gy/xvtQ0uqlHP5A6xrsWPNS9Cz2lPAgzGr+8PAK7p7U4LS87WLwMNfk76q6QO/4ywjl7o124gjbivHK1Jr0iPEW7FrRgN2C+vbzxOno+MrE/uCQvbjwHunw17cArQTqqkTfmtQ08PT0jO7U9ab2DOEC7sTYqMF6/ODT2sG07nKnmt7C6i7oJOFustrvjvTW9VCbBsw49MT74sgxACLTfuZ888jlFMjk+NbQHu9K4OLpUObe0mDxFKBQ+jr12rY4/prRjuXW1MbjiOQm0ALNaOfe8UrsVv+Y5tTEQPlC5hrhMPLS0dLqoNNC95jR9rusyU7oKswc4mLyJvJGwFyugPAw4Mrygtug1tzkKNkSzPqLltNO4l7y4ubs8JTt/OwY6E7ldOHc3q705N0M9HblpuAM5WrfRuFQ9hzvZMUe8Xi9Hsi40Obt4u5s5Gj80LQW5pj1RtNe3CzZftEo4K7WIsK2347lfuHW1bDaSsHFAiTnQtLS8Yb1+uqi6djxEttMv0zjGvb2xAb+ksp84wTCjKt60LLwQOaE0cjmyMEw5cTxLOIc4rEA2tbc3dLcltKu7sCxfPdk7CbskPGW3QLqWMR61ZToJM3A4hbJ2uCA9RLUnOOQ+yzWTvWi4pjpcPCu8o7o8NNe74TknNbO5hbaHMkq5zjgAPSsyr7x+PM6ty7h9LuKbBTduv8e4EzwJM+67ZRxerh+8vTfQu8YzebjIPVOpXT1LMxI5PDmQuYM1rSkltEGvPjoWwSG7SLzLunA8Ga7iNLy8P744upm8cLCxOmK1trh+vBs4cDjyF0a8jqkjvXi50z1fvA6wy7/muOK1DTbGt5BAcD0vuai9ajjZs9q6K7v6uoC4kzj/PGI3bztAuZE5HrnoNJo/WzNCM6U9WbMfPxK+SjISwAC4VbhJO0u9BjcPvRQ6qbiruboo7DHLuUW8V6k3OnHAKr08OxQ4NDp/LG23lDraOMu4ObxhtOS2j7dAuHs9OTquPNI4Oj7TNMAzzjXrtG+5BTuxOjo5AzsXJJW5mSFzQCS8H70xLFe3wyLppDQ2ADjcMhW8yDXpOCw6ETMktCO6DEBQrLE8wjnStr098TlwOC+9cTxNOga0wD0UPc8sCzWEQRG+s7eDOZ0xFbi/vYQ/mLSWtcQ3a7lQPEW0pbuKNb65rr2Cqm0yBDoHst+41TSJsT0uwDZxtCCgVqHTPAc7KrrlMWK2rzSovNW6lz3AuQe7yT0uOzjAcLrDui2/zBL7PJk5Gb16OP87xLi0sZgtP7htO946HL92OPa52bBjvHc0FL2TtMe5Zb7Ftbo287tytzK91zcltGUtxDwmvUe86yF0u1s2vLWRPpA5jjpdPV88TD3TtTU1YThMN/4nGa2hrmU+9DUlv9awXj1ns8i2Bb0IM6Mm0rhOt6A25DuhOXM5BTBqNEQ15jlnPYcyjz9Ivmc5azigNNG3obvCvF8uWLh/OGQ9sjlFvFatxz3brJs1VzwvOUQrgS4dPWMxhD0/Ol07QrIPPGO7PTx5s3+11TxDvdc54qQAtMayezAMNvG4oyy2MBY5ZzWIP7K3izgYOHUxUrTEPm29vSSqvgA+nrhDO3w3/zn1OrQ9KbCyODm4pbtysT+1b0Axu5o5IjjCNnq6jrlfONDAiTrKOxG8G719PyylFje6MUe/kjeDswkp6rmduPC/26qIOfu8OT1lPCpAbzorsUq9IDxAMG27IjvCLGwyKz28OEU7OThyOWY6Wzwhsue1bzAiPGI8vSJsOMQ4hzScPick7DiZOrKwvbm7uQwtybAavnm0wLiBuge7Pj0XPBW9PCRqtS00xS0yrX+rvLc7u3cxmjnVuJixLCi8owg14DvFrAA9HzDxviK4NL6VLSo0PrkUOlk7Zravuhm1c8D/tCouADkQOsW0xjn9PMk9PTOluJM8j7XEOt8/s7RQP3I6Ur0ctB44sredOrIw+7wEvOY6HUA/NDQ9sjnaNV24g7zbvEy4VLckOzY8HcD+KlM2MTYBO1e89zgYOF062rzlvQ84Xi1CPPM4gD/FMMe8sDqlue25YDxNvLo8aby/tPww1TbXuPssKCgGtsq6Pjxfs9cvQrpbNfu8Oj3uvno/ALwNPSxAbjlXOFG4iLPVLnG+wTgMOhK55zGcqok7ZjqUt5w+/juhPWw6gTv3usm7EDFfuR460zZBNCq1QrF/sdq5fzoUPCU5CD6Kr608vrjUu9I6ZrRstqg9qr2xOP6867mMOMI4zbT2PUo6aLgfuTepiT2gvlG7urnXMFSzULbcqc48TkAePqI4uzSCQfg+hLxAQKo9ODrZNmS6GLzeONa7DjvsMm4zcb4AtEtCwjC8u0Y8arKYpYQ4drfftRa80TytOwK1ez0EsKewmrOos1m7grlrtSw5+DkwQRU+hjo4NWKuO7fnPWsyVrmQMRwm1zkOtkW1KkDVMWYu6zloQJK2qjw0PfQ+Oy0UO428KDRHvf8wHjzCtGy6QjpFvVg5GLCtP4QgcrN8Plq9eLwSMkSnJ66vuBAzB76+rT47hDwLPaC8Fzl4u6g88bXKvSa73TrBPLE2rjkmvag49bFEO+U5lbsePEs3WbjhuDW0nTPDNdcxqbsDKtY8SLR5v3y9Kru0Pfm+/j2uO7c6yr6kqPKxkLw9ua4urbWUPJ+5Ab1juLQ47MDNtp08i7gONyk7YDlDt3k7zjjoPCQ9Sj2xOEG64racLNk3ebKTO962Grw+MD+4TLkYOzU9e7pGqxW6abhGulM25LsfOZ+3PjZXPg+4Xp+MuEW8GTlgOKXA3a8FuDfA+jyttnA+2bvUO147kzjutU89ZLweQLoyMzfKMHe4e7ULOrCvd7feuKw9YjxMPiA1zDmrt1g9/r2AOS23sDhTuoG5pj07PN051jMUvS4z2jxpPKS2IrgJOD46nC/POYo6O0FtuO64/D74uIK9fLxzPek7bL5xv4S4NrdBMAs0/plJNOY8n7c/NMEuwDzivn86x7ZsvAA8OzTvvdE6672/uEq+ozfXNAK+jC9ktJJBpjalLNS+ybvWvEYs2TSVvTMpnqwmuj086breNJi9Z6mCvPCtlDFfpxckq7LrubSynTnLvL80DLZ4O3M0YaquNPK5cD4ENNg1UrvcJoG5q7g/tIA5NDQPPbm5rDCRK242G75QORW5ajbMLC6uGTymQGvAKjYNvXs4YrjENxm93DlmOyg2xqpnvRS4YT3ttVM3mbS5uZO/tzwov0BB4Thbujw+LLuBPLw9bsDtN3+4fL18NxO/Iaj8sxS5WMB4uR+ulj5lMg+9jri2OBM12bfiJ166KDBcONLAlzyXtPu1hj4osHs1G77DuYe4cDHHuAA886xBLNw8dDzeu5xAhzy4OJm5srgpIm8zMLu3PBo6/TiJrcWkYKEFsTu8ubEFs50+sbgNOCKwR77DOIu1AzktNOKsCq50qka3CbhpPcY4YTF2vG+00TzMv001vCgIOytANrRoQSm+aDTYOIm1PT2ROoLAJ77MtBM8Fzo/uNqyQbzQPUi1nb5GM+263jGLOXi8IDw3PQ47HbfotDC387eduLW4FDb/PHI+qjbWMlQ8L7TAN7o8HiwcOjs4srYNvAW2kThuPk4/ZbxVOxa2PzwdvsW4AD3OvVC7H0AWOcS3IziYtGEz7aEQvniumjtzPCg6eCmNOCa4SDpEsD46ibgXuwa/wDliOfqx7TxNO6K3RDZKO9W8s77JOBI8e7QIvQe8jayPuHw6hiRnON0tbLhjuwy0kCEWtme/ADr0ttc79LzJsL+rDr7SOjS4CzBqN4u8vjBQMbK3laldPO+4h692PxM1UDskvY446rw/PR88kTBWPr+zFLEqvqRBfT+UMJY60rfaQLItkLAiPJI99Dr0OaO1KzfkrSU02j1fPh5AiDg6tIQ+Iir3NAw5FrnevH29IMGlunm8AS7yLJY7JblQwIs7jDxJOoa8fbY0uCwxp7uRtuq9t6izvQvA66/8uDQ9cDTquLC8kjmdqr8tiTXBP2+zaLycOak6SjxvsTW0eTeBvRc0ODM9sZY8Cr5CsC8kVr6/vJG3x7yOOCQ8xrwVvnSz5bf6OmE/1D0oLwY5XDSOOmw8fLgyPrgzr7nZNpg8Ij4dPSI2k6+qqzA1o7OYKp884Tgfq2aurDvpsZs5pTjjKIo43TSDtr+2DsAVOGI7WzX8PHW1FDwBOR04BbSMttS3ljdMtWclaSpyNcczLD5CupqwJblBNY44XTnDPEs6nLx5Oyc4ZryOMcrAmDkvNgS6pjjuPhI/S7+evIYu67kEMGos4bYgpEA+8zk8PWE4XDhdP4i7B6hcujC8hj10t3ewd7cdvtw5ZLYOO1W3MLRitXw5r7mZPMouIibjvw47+TY9OsOz2T6suLE4zjKiu/G8XzyxtIC6rytEN46z0jxMu4WhAzUcvsg9qjlmshKwJT4ivUU5YjkgNbo4/zf5O9AUYLg2tkq9BLipOu6vcr6SNN6zf7HlMhQ5tj3OOMMoobgnu2G/17eDLxe8c7/sOmK3bbXVumM9KT61OFi5Prvjujg7qT6BsEDA2TG2Pvyyo7OivBc5z7PJsPowCjtVPEA7mzbduD46qrryucU827WnPY22ojizOU+6bznFL1c2ISIzsEK2GrusJW614r6xuzW27Tw4MQKmmTbnt1u2kT4+uNu+wTjqtxo8JjYSuZg9XDrTuC6+Zr3GORWoLbpzMk4wrLhVMP62UbX6tUI748G+tw65MLBIuU49tbtLLqi4Tqq4Ny033DVUtYwxk7wlOrM7VzN3tLK4osD5t4bAIb/Qu9+46zjiu0oysDbytGFABjSPOde9KDYnsHK8OSjpsCG73ztMv/a1zbwwOdO4M7NIOw47hzfHujG8hzMvuFE6MDj2OaE7eLraO1u1MTw2tJ40GrSKuBY9+bQAPLo0XLvftIg7BzlWtNc4WzRgPQC3vLmIwEu9YDHMu96yL7AEPai1lrRRMNQ/zMH+OF6usUBMurG9OEDuKtI6CTR8NVO8ITg8sSwuPrXXtuQyCbjkOEu3ErGaPGE45696uvE4GjlWO+s7xT3FvdKz/rhUsuSytrlFOGmjt7hlvSM3kbwGP4Y2JjzAOvg9IbuxvE+1VzyOOXI6syzDOo00rD8Lo40427oFvNS5ETeaus2/B7d9v128IbrNNeM2mixKsyG5bTWKOZq7ELTftW+1UDFvQOO7Gi0dOUItnzQpPIc3HTEKuTu33K8jObu3274CNsq54DKMOik0XT+CuuW6ubkavfo+yDWRuJo55ziXNaatvzStvHy0JTyduo42QjWptHi9xTlXvFS8+braNx851LwAPQs7y7W3NWmoRze9OcK6cLsguH074T02OGC+iL16LqK76LxTOEI0TyxvM5K2kqgXLBs+1jnKsqe8QD56vYBAejpzOD7AgDeuPR67Srw9uYGyV7w+PFC4AT1xPJC9fT+QM0676ztgtqK0crafOFm/UzptKoW85LlLqSw7R8FcNT81SSyNPjAzRTYoOK2ocTxsNik86DZ4uxw4a7vDPXsQj7pBK1GvQj25uGu8mDmiPio0XzqcP7+e+bYNvda4jrwCNWc4dzi+sxg5Ez0CtTA4I7B1vR47+7lathipfzg+ssmxjD1ePoS4oTiJPHK8YT7qq+kn8jxbvKS92zoNOpW8xC7luFQ7rLg7vpk29jkuvAupKrw5KqWtoC3VOMu3YTjHu3m8LThrOis3fD2xttG907eDvWc2urxIOR+zRLuKsJOoSDgJtqS9MT5KvJo88D1DtcY5PjqePtUy0LFmPlk4kz3hL8W7sbEIuge9djnENc85jj6ZKQE7HragOAE1gTJeNJc9MbkAuX62/jk8tiq6RjTCOc03k7+qvPg5uzOGt7Ks/bQguzy5yroqumW3OjgKolewsDGFj9W7jTdEOZO8uLjkQBw6uzbGOYslxEA1O9g/cTExwGg1ejrEOL+7Bjg4PWA2eLwetjQ2gjGIsDM74DZBPhK2CDsQtRO4hLeTtlI3Ib4hNV65HsFAvR05PTihtRcy87l+tCi75ziaPHc0MaqqOKu7bLUoNDMzbLOYtbK4Gzvssy01NzacsXc0+bo5u6nAMLl4vZmvbTQ4MAMuIzhRuR04ezu6Nxc3LL6SPtsty7nWPCm5fDvsuPO0VzoHPGSyebzvuuI4njM3wI6vvDgkNCG7WrocPAgvCSXKQBW84T1+G3e/ijHIPtWyGrZnPMS9Kzl1q8G0OLOLwDQ6nzzuOUI0KyS0PPW7C6yIuoA5BDtQQCI8yLksO8E+xT79vESx3Lvzu4w2gEE9PAM73j8uNj46cMAsvF88VjjavCE+lb2LObA6wL9ZPqgwby9dPTE8M7pCPtwwZzWRubA0hzy0NVA5hjc8vcgmnDU1IFW+B7TRuXI/7jhiqN2pI7javKKt+b2oP2w3+bRiOC225rQHvRK8/7SMt7Iu6T6ROEYuCSy5su451zyStFk0rThHOIm6uLZhPXA+0rMKMsS88bLttHo3JbpDNS67Yjd6vgg45Te/vIk4ATvlLEQ7o7ByOyezPTy2PT44uz9JuMO7QbYhOEU0HroirSi8yjW7O7+ynbkWPAo7gCAhtz28W0AlOegxlb2mnGS4LzkzuYK0Ebs3Kp03mjstvb6+tDoMvjo7m7mKOIOvmzZgtLY0Wj0xLpW8ITscqum7Nq3fNrI5B6oTP7c0xrr0toktG7Tss9ZAazkYtMKyVrsusEE8pDVjPt0yg7x4OPQ59TwPtDW+ALJIOVa9x7mMMq06k7hEONk4vjcGuLQy1bGcu1EhEj4aN5q8F7qJQOazTKy5vIwygDRjLPcxM75yPAWpjjwzNTK4jjwkvK89KzplO5K2+jQqrr61rzpTLK23pTlHrzg9LTH0unU4EjkqOHU6EKwTN204KD9yvmI7gTbdNVg5U74Wt1IqLjh/sHO8vjngNUW72DBPtx08FrD4M4u+bLfRp8g6RD0JuTk8x7VEvIs/WDuOLfu3MrqxNFQ2Ny90ME211zpXNtAhpzLONQgsiz0QMZy80bwZtCI97DeqOg86TTOTr9U2iThmOiS6nzzmum04czX6sKy5Dj4MuRgxu7PlvKS6TrjXQcU5U7qcMjG8y7SYuhAwSjmgNei8VLrBNHs5iDKEtQ65I7c3PAm/CMDZvOYnlUC7vZ62g76JuuC4MC77uKW8Fi/Gsrm+2DhbM1y9A7iQPGe1OTUovAzC9jXyMxOn3LyINuUzUDwttrY8L725rCc4KLlYNdszesGiuOM4nLwivrStD7XyP5g04DA2LI2+Rjs+QMS3ML17PdW/RbzOqky4dTj7uKC1GTknNq2+u7P+LTo3CToZtqo6Sz4lth83ATlpumW2EzEdurG7ijinu4IrgTc8M5yv0TgsPM681qBsrPe5KruGPBczGzs1umQ7zS0pLA5BBzZQP/w76TgYvFu9o7L9N767BK+Eto23AbnPMic6pjhGNcyvKTjsuNkzcjI1Mc45aLUcuDs9ObgKNjq3Hy9rOOu2GDlVOPg0wzMKPNqqrLhHtM80cKBGvUfAQr7bOHwwELSBPYK4Tr+2PZI8xjg3OqUv+DmZNxAsO7DZJ7U+Jq0gMq07xLx9tNc4OzKxtv61O7hXukUyKb25LHy6HDpqtx03UizGrxcyTDVgumo577bJvli0HLyPvOQ7FbPHPEE/dbpbvUkxwDdcOM27zzHUvIs0cTkROMqtTD+Evou0eTE1NrK40bmTOBDAHzfhNn+tWK1owB88aruQuyo8cjetuLAzwbsTtfi4Lb2QuHO+XC7fP2yrZzTkNRfB0T8Il4+1FTh6Nd26dbibO0s+O7xjtlU7Cbp5Owm8sizROAW8Gr4IPsm4YL0NOwa6WLQLqXU1tTkcOcitKTyHJhy8Ar1EMzO2ILb/EIe8Qbr8MLs8OzRzOcO4JUDytU3AZ7nEuLqowLLBMjg3ETIov6o5izojqnavejW7rEc0q726JOW67r36uNo7zTkJuX2zWq2uPqW/NSuRuYo2irSYuuW4SL35O3K3cb4svLk+NrwMuiU6vjZ+MHitNT0tN0W8wL58uIg8szV7rsKxJ72ZM0A4TL7+rQqkoDTemHmtCrk/vgm8PryvvXy/FjXnsEc9LrjkOze2abQsQi68xzfaPHw0TazkOK68zbspOcCxIzagu2g5FZ4gN1+5D7rKP1C2yjQuKKWkWr4kvXq5Vz2Nr0w9RLoMPdeyTD2OuHgwHDoxt/K8P7ietXiwM74dO+s2xrMyukO7Iz1xOzE2FjhAwJcZOztwO0AtSD2LtKi89C8+uEK23D0QviwpbLgusyk4BkBvunE8hLdbuqA59TqwuA02Cz0aPdw9mj5kshQ8prUEK7y6VLUpP421YzqQJQE+oq9fPSQ6mrrrOquxeLQbvF0+sjcTvg09h7n9OFQ49DljuD62EzwQPG22XrwzuDQ9ILkKwK+4Cr4NQMK9drF6vA29drz2Obc67jm2P2071rvRuHhAGLXwOYKwbDq8saE7rz1fMHO5b7cKmY67RDPht3i0BqwrOWkzm6mPsvusCDziPN4ziC8XsO85QrnwPdO7ejpYu6w2qLcfvj0zlLulu84rSC1ZK2Q6ejj2MhQ4kzhQLtA4JbdtOxy5druBvkU5xb6pvaM5/Dl9u10/JLiQNYm1BDuAtbq7fTxsPk23OaGbvXc/yrs3Oii7ujjbPtqwATyPvDq4PbefO3+9HzOeuIw+77a7PEImZL1LvVitqTdGvG88kbzAuik46DvwPLW8zDT7Ok60hb1yMycvsrFINLE8QDmntLO/jSyquJ61OL13uyQzlTaWuam7f76ruSa93j4tuwS4gr42ul0xYjw7OrI+lSl8rjI0JzwePks70Dufv2k9FT7Er+axfbvit0I8hqOKt4OscDteOMy8VTceNEM2BqMbNzQ9F7UtvSQtuT1suUQ5N7dsOuytYjVzPSA5w7zQN9k7vzsBu3o+ODlsPMu447aYsm41Lra7Oos47C3ZQA4pejh2rXi7NrFNtf2oxzW5kx62tyynJue4zTwDt5e0ADdTOEG/ab0QvCk0vLUxMyc9vLRAOa9BQqowNru/hr0JuqK4rrHBs041JLpywEo0qr0HuIK6BDyvOCPA9LocOL+8KsAkuEQroLVBtItAK60etEA9GbUkPEq86DxsPba4ALYSNHa89bl0uHkzMb5IPBm4KbSItRU5Yj3pvN63Y7odPBwqSblwvMs8Brrquje34jmJu748G605OLGxBTBSuJa7RD02sQGdRzn7taI9GryasqQ8yz7zsYM8rzV1NVFAyTmlufa5E7rpPjy8RjZluInAQzIQPDM6TLn4N+a2kSZzN5iy9z4HtTU4ObgOutExVjs2vbg26TwtPFQ0z7iFtUO8Jb3iOSI5garUsNu3p77XueA0YrN5NV838rTsPmw1p7Dor5gy6LDYO8a5bb73tEC5+7iyMPI8lTtXuBW4vCWpO3w7a7cxquY5i7glPCa4DD60OfY14DmTt6O7ir1LvN2+FrtwuE9BibbquX5AArrTvEew0DvrNk6q+jwuvKq4BDUBu5Ownj2NORQ7i7jtsLg55bjWN6879Dkfsow4bzTXJEY1JLdGwMw9/Lmste+5Czieufk27bIFPSq8yLhcPBC9+TseOb46YDonvV2tnz06vhe0uzJWQBq0KrfhtHi9rbgHNxK51b0xuNC8hr/hs0CznjzYsQy1Aqjxue64pScDtZEt8bdNN2CubrxoslY7crUMtjHAdbxEtyY/qi78vHw7NjtVuGE7hy9zPog7+z48ONQ3QDP8ufC4LyDqrIC+4zDBr6g7vTW8OPu/QredMio3Ibs/tTC7Pjq2sGK6vrgYMfG+4TT3txi/jSJ8PAm2nDK2PC6+C8BcPYrAnEC8PQK64LdpFjw6WTx/s+Q+GTt0tU0tD63UvfE0A7Ytrj49Yz0+tV4us7xmObI8lbgHtToxfLzvu/GyvbILvae91jk2umky1Tpet528bTSWNwM5JzUONKM3U648PYA4mLp3s0usRLhPNM68lK2gNOE8Irc/P/82qLZAwCw0abqTvF0207djuNY5FLQMtfCzHze4OPjAYLdlNbisvrkZvvMiPbwltjC+QLp8Ok07dr2KvO41Ar+TONW1HJ1rLHm5gDNhOW86LTbnPra7gzuCNHQ99CguPFm9GsBgs/Kzbj4ftqxAViqYvPKv7bhPPGI5x7iYujK5i6EHsBo+SDikNEY1Jz2xODM2Zx07OnU2N7FPPr+0JrEYPeoptjEntaa0MrxUPc09ET3Msk21gD62sZU5YjhlMNwqLbH0uW62GTwpN9+1hjyXun08sLhwPQw6cr3duZjBVjupOdG457v6PXatObW4OdukgbtsPgkwFL83NaQ2+zCssC016bJauGS+kbg0N0g5yj+GvjI5Er0evP+5V7hCqzC5VLwQuUi73z1Jwdqz0aB6O8CzxTWHOMIWsTECuaS9ebiaLfy0dLjcthu+3zw8t/i73zgkr2cnEDm5Oam8Rb0BOYMxpjx8vAA9ebaItRo7Dq6bvMm9CDlmve65Nb1kvAu1w7DTtqkoeTxFtIk/YTnrMwe7Hbg7NqW5FL+4MTq2UTY1wey6vrqcObu2yqkYP5+0KbfCL92piLZZuRwkLEDFoQ+43Lw/tR06T7yyMWq8HzWXPUO5tr/hONEf3bYXOz+0oy7fM3O1abfovbG5FyXNrUAnKrwDOfg1tSwgPF3ABDxrOBQ2lbVltOU0sbTluio9Sj1lNyQ+hTmTuAW1f7qwPp66IDTIup28fL10sSK8kDliNmE4ujcKOZwv/jgkumm2b7ZktA09tbYvQCy+FjCcsBrDPD3StZqwHTcMPha5aLwjtK00HboqKDKcWbrJOUq4Xr0uuCc7V7h7OaW9mb1MP+qwlbQrODStcbo4Orm0vz4PPE+oAz4zMEI1tb34NUW7Dzd1QKk9BLhsOLm7Jj3cPBi9o74Oq1Y4ZbrUuk08LbYwsG80k7rKPZkxwLsNQKA4wDydNCs5GbWYuoY1TLRvNGi8cDtXMQszVTzOucK0b7hEwWy9QrjgOv84lr02PPe8Jj2gMpG7QTX6Oc22Nx2dP7y0mzcsO3I2EbT+uG0wsruFOES2DK+jula5CjZIO/Gw0jreOgexXzias1m7kzhfvAU4fTsIvO66QLtyL6g567rrN6Y3+7zNMm680z/pLK66qzLNNQo0LDxEuqo3b7SAudc8izpRQHa6DDwXtgu0+7wOOnZARbxeNry22jkdvnGxpbcYvKU5Qb89uJq0wq3puY26Fjp2uQ0zvrOaure8fzjlMo80HTjHuG/BLcGovBO3lDw8s3K+4T5kt+6pGbu3M7c2dq6yuvI6cSftvDc1yrUzOYc89bLwu4E4AjUDOWMuUTVHuCW8nrObPY669boItkU6hb02wFa4szdkPMMwYEBHOkwutDEDtQ28Hj4ouD89yLuTuQY67DabvJa6wLQ4uYy90jcEPJe1k7c4Pbo67D5yvTq4yLu9OcsycLzyvWA9iDSLQcm7uzf/tfCsNy0Kt9W4N7t1uZI4gbt0t/FBqD1Eu1a6urx2u1Sc57YYMFM8sTx7O9o9KjfYPI+00bmGvZ2mh7wZOBg5Nzm2MAWv5L6jO30waT7yN+ourT7bOrvAZbp+uwot9LjLsWg6ODsBNfkuMDNlKhI6LDqnPie+OT86wMG3lrg4siW2q639O8o5SD14uwQrJbN0OXE9TTorvFc6TUBaNtI7Aj3eKZG72TlEr8a0OLCfON44XrDrrEm5C62avWM4e7d/uYK2nbEQNWU2LjprPV8+nrgPtcW3FL8GPEI3yDwjOEM2wbVOthQ5lbSuwM8w3TyqvEAwBqZut/89MrjtPCfAl71bsCW0zDZ7OyU4Qa/8PLm5ND+YNEW/orU0OY8t6rg6MuS+WbQ7NXOrHjgXtvW5H0F7PHgmsb2PubE1pjkZPNu+mbYjvj44ScEtN4s7WzZEoR+43rTIN0w0pbhQPLIpazEcpem5c73DMSa48a39OFEtharDNQU9orT5OFY6Kj6OvUW4CbYfvEi5VDfiOv2+kDOKNb894rTouIM5yTrirGy3RrxfvDe4mTm7KFq6WDZothQ+gbgqvQM3JCwNO7CwVL3xO4m45bhJNwG1fSkVPa48oDv/uI40Vzv5OzE8BD4ZOGmvYrTvPAgpHTq6uRU9STdAvRopELsNuNI3aTc/r266VbkvuGK9Pj/IPDw+TTm2PXu8KTv+t744Xj0PPBSzTDNUrlyysCvZuvq4pzihOWq6eS+wPQs5Dr3XOWo7XjeYsIu/a7PkPSW8Jr0Subm8IDh/P6Y3KbwxPFw+oTohOx88nquMtju5JLIcsLI6jTrtu4WbRL0asNM6qbpGPL83oLqAwXYwr7V3MAu8XjRPO1s4ib1oNF6x4rE3tqG9fjlKNQu8tSX4v4u9ob5NuauqO0HbMESwZ7/+N9e8K7kKv3i85jbXuLe9CaVCODk/AT4wNeE0ez/5rfi4Eb64ra2+NTO1OpzAZDwwPKi/LDnwsNE827hdOJq72TkXsrMyaLLzKoo3CjYTtau+4r0hvm8zkzo7vCvAQDoXtHw0ZrcYQS25CbtGwGs/ijEotC620rxYM6sq9rlEwaq0CLoCwMe4yaD1vpUieJymu+E1n7t8Mgi7ijGWv0a90TwLOYM3/L4otoA4QTyxMCI1pTZFt/KuCkCdN4c1ybdVsz+0xbSNwEK/Ir1mOgM4Kjlnqlk5cTcMsKe7qzy8vjO/mTkuNruzI74TOD434rrHtpgwRztaOPQ/urzePke7ajUuutu43jorsqK9uDu0pvI0Eb00NI064LwerNe2VrnJOUe92ro9PWWwyz04uc+ySKqOvOY3C7W+NbGxDDwwOGe3NTndOJ48P7u4tjK0vjjFth047r0QvQMoWr9iOM07hLgSONO9ZzkGIcIof0B8Oc8ear3IMrY7rbnlPa+0iLzfuWS99TvNrTo52rgWuGw1kjUHsAY2kjbMOEi88rp1ONOzNyp5KHm9mUCatAc2ULP8PNK8jDTvPdCyhL6LvSCs+zneO/6pXze/tJQ6DjURNkKxUbbcu3K/XKxtOEW5abqopaGzR7UnutS5IboAvNqwH7STOzk1rLTJMHK/gjb3szSxyziuJWg6IbiNPYK7dDeLuqQ4kLIWN645ZTGJNdE1oD3tsAg3try2MIi937TMvEy087h3OOw48zcBsjk5gjkFLpU+sbrfrPS9zDwzM0O9GDwWt4oZGDVntOGw9jTHK825ULxNt3q32S3JuQO5gzWjvKOvoTpmtV45R7t/O0i0Nze1tcS4tB4UN8A8P7djsXQ06qziO/U4wTOsL+49JzaEPis9ZbQkNd8rWzlRtQCwvjf8uLm5yjDpuG07gr2jn0c05z4prn8+QajoNa07IjxxvFc1SjTqMZo9Njw/Ox/AgC8zuyu44LK7OMG4FjClNFc71bpxvus4QTRLtBO5XroPurmsF7pcvlq+Qz4iOa64oD6Sjhq8KTzlNdI2qDx2vgw7qj86vEQ6Hi9Xr/Q9oLXpLwa9IjRppNO0oLVcPcFAg78WMjs4ATyTMTA2iaowNgAuMzqTOR1A88CLvWC81bNsuI0+DL0zttQ1eLZ7O4e6lLwWuWYvi7ZxuXc0r7Rdvu++ZbcGuGelVLymtia2zTtGIVa5Gb6WOmwwgLnltl6+azgGvjcxVzwWKPu42TnLtQy5XzyTMfO8RjQju7A6Jjgft8s39bf8vlPAeDQhMk28h7kcNnU9Yz2Qq1EzLLe2N4A4qrrcrHwraSU9u8k5JTW/Mhu+nzxqoPu9YzxQLgQwIjlMuPjBCDe6vJIzKr9PMZo43rSevI25TjtKN864/rp2qyI5jLQOtna7XrLoNfq9dLruO2k56TzlrKa1lrhqP2E+SbOhuJg/cbIAPMw8kbVTOqcsLrViNh84OLBdO1oy5LLWPd8ozbyEPi+5YzhrNbe0xC+zNE82970rvhKwmKVkt94s97vkuoQyH72EOq05ZR/APQC79S02sEQ47TltPfMvPz2qrl65wToAMze777dUuJq8MjYquCy7gr1lvIK7iDhBPTS1crolNKi7jzsHukw9IrWitlSsc7o3OWw4MbpzPbg7ALjDPXK8lDh7uLe1JLWjPtItP7EwvZo+lTv7vOM0ADqcvBvAWzxvPJU4IrsRvS/AFsA3v3s/ardluAdAAryKvAG6OjXMuSa56bmxtd6nkbersjU0sLgeuk86r64FwFS9lTh2Orc6SDy9PJA77re+PhI6/rhMOdqpUrp7QVi8ojh2MNa8U7xBMLu1r7nptgS1gTpIt/o5KjUxNIy9Ai6dvA86mLl6MuU3OzivtMSxVjSUP3VANzlbwNO0q71EOrk6gS8Oq+y81iKksgm70ri5Nm023D7fNVm5IrRXtO+yZLrFvd68yzk+tOS8kzFuLXC3BLCouF++ZD96wAm/Xj5Ytum167qQtmu2BMAfPBg8Ar3hMv06wTJ6NIQ4oa09wS4+DLhbs6s1Abi+uPQ4mj3XPUg/oDgWPTu/3rbfNZW8ny3GwEo2FLfptrWz3TLcpAO+mbxWua6497hXORg8NjdsOGc8tDapO+q17bZpJT+zV7zpt5AuU7oZsc68F7mMMX46xToINZc8S7sLv1I3PqjmOlm/d7voPDG6vLsWuzo2nDpRO6egr7x/r+m+cDd9NQs6MLeKu3w9tD44O2a8Hq7ru4s5Ib1kuJu3mLxPtUu8NDkWO/I5nzystog1ir3LNX88DaZHOyEsS7gbvJY7oDnZOAW5qLxRNxIygLghPV47dDdNvXw4BLhcrgC3g0CDqTs1+jslMOW6Gz6dp3m5hbFuNr46D7bxMw26vzBuMRG9G77sqJ29gbdbuEU1OTd9uIw5lj1aNlCl9Lk8sw68kTVsOR85DbbrsAGxSS9Vq6U9hy1Au5S1pjLJNLi5UjtXsKU8ODssOCY3EDn8IWK5OrfyOfu2ciwbucS9WzL5PKww6DkUuWuyvzD6Ly0unbzVvvi4FTrPOkq6SiGaQGi0cD0CNng4MTS+POO0Irx1PH8987n/tMEpQjm0vxW3gL7PvF1AtDrjPLW8MDi+OJG9HbRMOy86aLy8Ofa8I7xVwO27TT/KOuI39y4fPEM5xjM7NR07cKzWuqE2jr7Nvf45vDxSMkDAzTbsv647ZDTmtP6+PrisONs4GLB9MGC1fzhXvGs8hi4CuTq9NbroMro8PbxZOLs6YDp3P6m4ETUfsuq8jDWGNE29LDt1LuM/mzyIOmg+Iqc2PrS4ZLoGLy64K7yxu6c30izKtA+iLbgAqwXC4TzeqI64lLpYM6m/SjwvNyq7Qryos8S1Ar0qv3Y9ZTvwvJ22c7I6POg9Q72UPEGwLjiisu7AKa2IN+w8BDqWM6+7nj0ysWSjmLkbQN27tj6it3m6jDlYv+swKrxQNFo+IMCms9w76TtcrMW9YLgrLtg+LzhTutYxDTgBwLw6UbAmsV84B7RjuAI2xi/NQLS+TD1wNYiriD9yOLw3KDj0NzYuOzQJwOa+/jETNBg+bzgLuBA5JbM/vF6zjDuNuwu4tjCluFkayjkWN5au2TXivF2+RrZrpoG9vTS4tj69ZL2fOlI2XblsOYy6Izm5PL8sXkDHNUQ3gUCnvG41nD0EN740TD1kICm3c76Mr4g2GTwdvMwwRyqoON4/1jolPW6++jswvTuykjwXssC+jTS7ti2uTrZUvfAvEzozISK6lz5tMEirgrwpOFI3ZzS1vEi40TPbOEm8Qjs4vWW8sL3dvZEqn6pNQMo0TDQZu6u60r0mt9iydbiDONk7WqG1tMe4FTmdOcg6EDD1PYC4Jzzptmo9CrppPbE0JD3KuBM9KTjwOK6/LLP7tMUybriLvDuwnTP6Nf82OL2UuYw9/TSFO1s6PrjEtVIuGjhhPHG1pjfSvr6wyDqKPky9k7zpulG9dDOysb89zD2gPKGzI0BJNFk8fSqYOK09Cbx9Oko3UrYqukIz+kBgMPG/5zihrBW+Rri3vOu+hD0aPCQ9S719NUK7rz1ItZI6DLXNpIw1TThWsZI6NrSJucC8ZD6GNqE0abffMm0wUjmzuc65HzmKLZa8tLjJt6g3ML4XNCi5hTv0PLIxk7SfLPC+X7e5PGg+ojoILc66QzXguxKyMbikwUs9qDDstDIpqDJVudc5SSEcwOM5AT8BwAO7GS9HvMeuq7rMrSs8Cr2CvB27/zK9uOW1TbwLPdw8mbzfvOs8XL5iuVm/yEAXNxi8OiwsO8KeSL4VPr8zubhTOWy5XB9OtbEs8D67NAA5Fj1xvJi7yzwcPgG1Yq8AvDGzP7buM2UzWTt3t+8kyzjduga+ojvLv8y5JD13roa3db+qvNexfTxZP5O2nzY7vGYu2aPXvOkzXMHALjo2oL3NM+Q8I7QiwIq7DjYdOeAxyjjiNnuspLnYNU26vbD8LAWw4DuludQ4STmqPxI97bxJPTK3tzhyu/fA9b3vuQG7z7oYvQWwU7wrP3M0dbbaPGa6Vzc/PJa0tT1BJTIyEcB7JWylOTmGsGy+yTnkJHw6FTdRvQW4FTA6vLq4VD2yPnSfAbpuNTu+TD2IuZq1cbooPX29i7yaonk6JiyfvB44kTvIuhG7sjhmspk69z5JOgE5Kz7ZNTO9/z3yvUEmb7j9rAy3L703OD3AnTiotJg9ETsluZi6GDZSPve0NyjAs349mbfoMre79LvBO8A6ibiePY8p8DFIsUO3KroYvtg7O7/PNKe8kLlArMC0pDmVvB06ibchrcu52C84tuwzGrwZu5gstLnGvTO3ETzBPrw8OituuZS9v7QMOSa806qxPNE2JEBaubYknpmusSu8FTmduDi1lTc/tBqlX7s5u94yjLjPtp4+5jZhtmg0BLt1PDS8Wa0GsIu8tzlUOf+4M747NRI4qjYgQLokurr9NGU2IrdKM5K6NB7jt0I8W7QHuIC7BrzquEk7sDcJwNy7PTqhtx8997JdN8M8Er27MKS+Ar/bvfU4dDkrtaU2Uj1SvMI9bTv2tWs5erCUNc89YDjmuN64rrcSu8w4sjARs8MwibaHNw4/wLZsuO65KDmBuqY8+DiHMdU+Yr32uGmqFzuTOLc4GbPfwCK6s7LFPFc2LbB5NS80EjvmuXdAvzuEOjeuEDyXuO44rTeBt/08Ri/1uae0/TJVrmW7rrAoveO64bWOr4kxIjKiuRQ0GTvXPAg6eD6st302Y7HkvIgzbDyLssQ5cDz2N0o7SzS7vSG8si8dOgm0B0CSPBa/bTZWOYk7pTVQv7+4XjkrPeKyqjhPqGQ27LhXPn09CrgWtmswibkTOJi7DDy3sYe81bznNj4puDI5uL66gMAPO1A0gDeZv/k5dTr/May2ybmQv68zHDaJrDS3QT60NlW7ZzATPwm7CjeLvD+8TzYruzaq7TvpPK66mb2nvQg5yr65PBo0rzzZtv24zDK6uiM007RguqI4WLbsvrcqgLzXvnQ0frIZuBOwqLRfuAYxbzjeM4S3traWNl845TchvGW9BT7QOYs6Qb8svBu6M7yxt+K5dLr9IJk7LzgpNRc4yrj6vEo8fD5EgCg8ez62PHowdD61vQ29OLz3uJo7yrwAMUkxOLuJOBi6j6o+v1u4JbOgsSe8h8B1s2G9I7iVvku2s7q7PBo6S0DLN984LTYhtbE2CDolLlG5tD4iOQk+bDovugk6nDtit6e8L7K0ux+5ljgstqs2abWLM68+G7pBOGI2Ir0/Oh+zU7gbr1C3QzaMtic84Lm2vJ+tD0CgqF9AuzqcuP25hSNwPSs787/SN+63JsIqvca727kiuBy7gbzatAE0fjpMOaExF7WjtCc/GjbrOxcsCi6cKZ49Wbdctjy7ITlyNXW2D71Vv/676L+VskY1dzqkP9y0nD2Uq2i1+K59Oyi9UjB2uOe2kzgANl0orTcpNT6x5L0yOZ+8Ebp+uXC6SrSyMb00nbwGPD+6mLuxPHW1brVDPwu6Bro2uUQ9z6AVJa+9K7seNOe5nTl6sJy2L7p1tzXBUi0NPnA8+zvpOpW7rK5+trg+vjg/OkG9qjH/N6i2oLxJODe6ILq0u6e9PL1UqiM/NDvftGy+PqlCtN8uTr+xuQU2AC3mPV28gD4lvUG5vCwrsGi+irjAOkC907y4OovAQjdAOGo4HT09vNw6zbjHPH08Y7y3uXk+MTTmNbO69T2cuiS2pyhzOLM7ea00u307ProbtB/A/77GNCi9UTBTNQ8197z2NiooCTZsN3o7CDi8LUS2ob47skq31zlfOUAsOjRdvHq32z/7sjU8/518r/I5lDkzMSK9Kz7TPB+jTStztv4xGr2GN7A5urhzOL41oSAYtdewyS7lvP88fjidPBOgdz+zuAEzAD9iPs84fKWbPAwqXLLtuJE0+DuWuFI8XjRXuGY1erw/lII5UTKgwIW28zpjNo6h9rznuESwLzjqt/K2KcCJrr25iDgIthA4Djs1tQMuIb6cu4isChw0LYq4+rM1P0G4cjsIuM49Dy/CsFi2/T3+vbK8ZEANOCy5Sq+EtfE5RTUVuhO95b8Tthk4gLzlvMS+Nj2hsgu9KySlPSQ5ui/Uupg9XTAMtUM+zbmir3K/7Tz2OGqzpLl8Pog2fTpePOu3OT4MtzY2READtlO9nLwdudm90TtTv8GnYTM9Mh+4ODgOttC73j34M/i5WDsyNvM8PEB+NIq1W7vjtSa4GD6DPFk1O0BQPYE/PTsTN6O1hL4Xtnqy0KpbNDWxwTznO6+8Rr2WOBIj7LiwunKsiLh9PXc9ebkfvF03l7m8BjC4Qzt3tLisnbVuNPgrqjGKM425MbShPIa/YLnJuMA6oLmdsa0kEDY7uRG8Wjl6vqMvn7p8NTA90C2Ys0M3DbS2PDExECsku5I+XLxZNmKt0i4hOw44jbhSOVW5+zxUuOC4RDhNPGrAMj+pvgM9n7qsu2u4Wjworg24oTTqvKmxb75EwJAxTjrBPpq+xDvXtK++77+2PN2QzbOXr/m0FKRjOzs47j7LMWy8mDpTv3s5mD1/PK62oji0uvK/1DR6OBG66bglvjw3jLiduAc+ozr6rVI4Lbp4tN64ILsUQK20Ujk7uhy8/qqNOMqySq97IDMujTpctOasXEAMOjDA1y3UNlK7eTMPuEy767nloHo7+DRyOe++Zix9wBa/7by2rBg64LxbvdM4bzOUOMS0VDi0sWm7S7xit7Q1E7e1uNq44DU1uoUyjLisuRqz270rOHo8CbE9vLc4FD/svjO0YakuOpE4u7hBPoq8E7S/rZQ1v749OLO8jbZps+e5iDjsPni9cDuKOQ8qO6uUuduw2bpbtqA8Mbi6uHs+o7gTuIEsI6q7NCc+NjufO0M8PDPmKrG64LPKvos8dri9OsMwDj4yM1y/CTy5uh06MkDntsk4iLhhLXW0VLvoOR08XjzjqKk7TzmguzU5JTfSpEU8vLTbwNm7lLwptXW9k7CluJ08LL7UMGs0Mj1iPGLAZ7XWNZY9DjqONvqx8rfnvbk0IT6wMUK7Dq0QtnC66a8+vXqybz+1qaQ89LgXsT006Dt8Md+7kCTzNHO2UrpZPZC0lDcqawgwEApCA2IxaEpgVTRMPt46DbmXOLy7bkCqOXSuOzgXN+S2o77KPXW8pjjAO940kbXvumQ/SDW/Pp829LJKspU4mbifNcY3uz0auTe8Gby6rq+8GTnguuO2prbTPc66+jh4PNuzCz5Tvb64Ko4kCDAIMBAKQgNXMmhKgCQCwPAoTrjOrNO4wznkvK60PjiQtqW8xSwQN4M8nTkIuj4wgzrFPWg1bbpQPL6nG73cO9E0BTthPZ20BUAGNIDARbG2NYy6DbCdPUmhC61iM8C+SrsVOLW/nze/NtWzbCtTvMu2kTUzsWk5pbk5N7iwRDxCNaq07LiVNfA5lzqnNhE8Qb6vObGzsT6ssOE9DztZpRUsIDo4PTs58D3Qu14+ib5QOC48w7y1Luk4zzswuXG5LLDgOKivC7gQwDy4YMDDOtA5ajN9NVo9DcL0Nxi+Fj0DO+4pjLiHQCY8M0BRMKm2fr+1NWm7WLhqt6M3lzoCNxO1CcDwsZ++wbyCPZEycrd1OXO0jbvZNJe42L9tuBEzELwuvPUx1rozvo+0OTAcv4i4PTTosfa2fTO9tCDATzbyusI9PjybO763PTm1vnwwkL00Ok+6YzUnMXk8Zrr5L8svpMAiuxa9TD9QOR0/HS/tObG//ixOJDKW0DwguxK4nbh5v+27yDuUs+g8QrB1LWe8NDgdN9a4fK8gPwo0vzoEODi/fbqmPEO8ejhWulhC+77Vmq29AzykMBW7HCfSu767sbwjN821jKy6PCS5PiovNLq6MSZ3PDG+2EDYur09zDXWvO60QruoPGg4fLfXwKg97y8fuYu5ujqDJyU/DyaxPAmy+zgmOxK+friiNzi3QjdJPDm6hroAM/csfjj0PT08dbmevNyzFDOxMR47TTe8vGi5174aOSa19TcuPA0vPbSTOTs2fzowMY46zsBmOOC/grxss+cwtLRgOeSyGzgKNAU6LDiLOYm8NLQUMkq6lz72P+Q0k7BEuRa9SbkUuUu6I7wNt6g6OrNbMz6vJ7KcPMs7STgjPIU1tq9csSxBFLKsPQk0vr08PfK0EMCcN2K5GzWMsSk/srzftaI8ALQ8rvY4hLAUPBa+j7UFvSS2gj0hKDizHLj+NAcZvrHqO9W4w69eO/I2Dy/qPoytBLipOfu4hbgKPdSckjjItUk4dzhWuCW09DpqLQS6MTooOuszIDw0ORs/7ygHPTw7iTPGu3y0FaxMM485IrgRO+a8BrxMLDA4rTxTt36zFL38v7w+VjBKt8ywVECHPZk7bjgIO5c5yzkMspO+FbXxOZA6oDTKuB2yW78guhg4mDquuTa5gzs6OES7AyyqrGC9Xjm7tHE8/jcUvEi7r70+t0ezlL7jvHc6MT2TOCE4WiseOpc4l73VqG04FriwOPw8KLmHutq1GbVwrkUnX7SGvJM7rrgmvuG+ST2wsR6sEz9pu5W9fT9LoJc9UTAOtfOr2abNMiRBLTQtNRO+/j4NOGU737miuyFAq7YCNyK3IjTWu1o9qzEjwHY14D4jPAK8mL1wON86NT3GPfU95LopN6a8Jz2Ws6E10DobNB29hLBpNT4zOr6kOHO4kjoKvKm9wj5xObe9Dr0MOYQ9zLf+OaUwLMCZuyMvFT0yvJqztTg1uQjAkLRbLq44Kbd+MPuuOD3Hu5y7lzyPPYw5FkDNMfu8hrnJtWU0QDr8tu09HMAWuKkp4Drju7G7/LeivRs+S74wOOkwkbnSPv+2vzANPQa7A6odutY6njvyv2O8MzeHOEI9q7LYwKA7tLTjrhE51z0dvQexC7+Mtne+2TlEutgzT7hKsoI7jzW0uTy1YzjcuVDAMzb0PPe4uS/uvM2+5sDnueU8arzopsm2bD8cQVk2kz5YOOM+aTtlOPy8FTlFve23TLu8v0Qwwy+UOno5mj5XuEK8UrkVP7g09zRXOmY0A7f1vWE9mjlQue87RLo3Lak+J7SQuak1CLkaNFI7nTbQtxU8vTbbOMmnB73jqq0/wS3jPYi8hjpHujM1Yz97v3moIzKlPWy+DzuNMTuwg8FJKgE0wjmasDqwGL2pNl0+T7sYMgm7Mranu5I2DDFKKkQxDzDKq0y6zLG4tM84QLygusA4eEBmPOkqYTgurNG4Lzo6vn40xrheNxq8kTgcua0dnMCwPMC3cDQWt05AUTSaOyq6nTQWQBI3SsD1O4AxI7iyuz88nbvLNya+47LEuKY44TnzuYS0YK9ztaC5SjyCvDa2vz8GOWy0ZruItgXBKbnaPK64oLMyQqW19zp9u0XAdTgQPWO5NbsTNtopKUAzPAiyZjmPvgs5IKkyPGm8mDY4vRGvtLRaP1EgC0BVwG696DlTu3S7Yz3SLzm8+LgoMK02BC8jwJQy772xsnI15ritPG+50j0ku/U8Mjk5u/07fD3DO161oj3TOmO/YTgCuW89kya5vQGvKDcKsHCx3jKfuSI24bgMLb4xd7vLPiQw27f2PCQ7AEBLuF28QcAduEq+WL38PJ85KkDotg0zczuvPdO5m77RLL4ufi3kPUC/ZTh8PHKxvzexDp40KS4uNIc5UjvYNpO4u7ZfNv+8cjybsfg3CzYFvbw72bisvG+8zLygrX08MzkdPNm+tTtbvPI837VIuwsltjE9ubg7NTfIukU7ILoavNi0mjbKMDa/BT5zLhwzeLwNtG49ETTnuRk7IT0ctLU8sDHPuMNA97Y6sBi+7Dkcwu22ALQevNbAlLoBvnhATrHuL9A5kTqyvFc7dDY4NBAyfCtJvek2UrjIObG8kDlKOwI40UB4PAE4XL+YuD8y2rjyuWu9JjyYsQ04IjarvBaokDyVM2g5W7lgPHe8brUrt7Y7ZbrotTy/RL4VtHM2wLCjNxe0C8CfqHS3grctKIywdDjoth480TSxo5e4Ojlvv5i6gz7bLkW5IS+dt2M13qvdNWe8gLjAPQm+8T0hO0g+kb0qudu0U7m4OtS8jMCZOi455zVHPHY3rDjoO5Q8J7irvLG4fDcXOGE5v7QCOrc226z/vaM74j0QPsiy4EHns9G9kz99vgk45LtzKYY46Dz8OJI1EjnzuCI8hjuwuBlAJ7TtOR8xq7UzOG+3hj0dPZk8Ej0jNOq/kEDyPjo9b7oNNsO0HL5/OAOpDz9aOWW89LTxPhW3dbepvJW8DkGgOfO88DdjMf8/pDTGPMU+JL6EuBW9hLChuVwfMD0hMsE7CzFjuF45cjHTsB23WDrDMwK4779FwPu55bgkvnA0PLddOCi9lL0ks6Sxsjwptq460bl3ttciB0CZNemyrbECuAe3si7uuz4+nzwBuPI79zMXpca1zJUbu4y4QTPxtjG8WL93vUq99b3CvaGzc7jIuYu7wLGlQFc8XTwpOFa8qDYoHtq62zhPutQ6P6ooJ9U8DDGTM8051S2uvh+yRzidPMY7o7jCu9C36TmXO4e7J7wvuPw88b65uuev+DgavxW/6bnjOME5SD0fvbmvZL4sODhAHDbKPjE8n7g1N0Q3ZTlGOdQtwLZhtL4/ATgkvAm9lz5ZOx22yTpsPJM/dbvnt9A4aTwQt9er3znqNqu49CnKLNUtirlwLARA2L28Mx88D70vMC61fSXWPVXACbxXOIi5lr1gP4e59ToAPam7Nbn9u+i+bL92N0K5yK3fMSu6IrjyPFc/DjowrFM7/zLXowQ2cDr4urCzGi3Kuu88gz1mt+orGjxZNHIt0Ds9PvS31DheKSi4PDm7J8G/BLzhONcyhr7NtJi70sElumEsxa20PdW3Gb6GOBI1QzbLMUkvFrnJNPQ9eTYUvBtABDgqOMg8/zYWO2apby33s6u1ZrsNOJk3a7Wgu8u97TpbOVY4BTpGtWi0MDxZO8M4lz1UMp27JLpDOPYk5zuVvOo8Q78WOk45sSRrPj0yDsAFPi04G73duS25NK78N4hCm7ShMuqsBLrOtb64I7jRrXG9g56JLc9AfrDLvMq0fbwbvDa4LrCwsvG9CD+oPNM8TjR2tNs0hrnjtkA6kEBfNrM2xjyMPi28qLclKty8+bn4vyK5bj0GtKExuj6FPAM3T7pgP/68CbpGPVE4N7hJvCI6tTxYtsq+W7xjNPU/XK9FN+65q7BAOAEzNb87Pj85JzrbuZE7vjiDOOI6zbpaPCe2JTqtM/g+qjjJv402zLxPvXy+EDs8NQu8Zzj8uySxPzADrDw4wrhkrNK9a0GLOiI7tq2KvWg6GqI/ueorvS2HMzc3w77mo/w0mC6LOKC1kDalOUE7972yvG25D7qRNTY837eSvRg66qn/sDQ+JrihuCkrzUAGugo1IDvbMhQ117kEODQ+lLaJL1k4fzo0M3A/XzAWPtK9ubiBNoM5XEBVs8w4VcAGOWq8CzxBOc+tgLmHNyc3ODSOvP68CbtOOCQ85T2Ct6i8Qq6Fu7i8d7uRqjo0TrZKtz60nDn+ND48IqjvvEM+3jVrvdE8Iz8sO7Wy8zelNyM5krgFvla4czyrtWRALkGtwK0+GL6pNz22uzzjuTIlATQquse4VTzxvFyxMTrFvLkvOzqltDG++7qJusk5o7nJMgC+zq2+tr28qqpUNUQ6zz3rOAG3YLR/PLM4AK+lOuazbrjxOds4ormGG2y3qjVVNOS7bqv3NE4tnT3CrqRA3zoAOD0+1LoUPfk9XrNyK/4oRjpfLjU7wq7zvRw9PTcqNrY4SrUbOto+b7ETN9s3gT+otCmdUj24ug84srRbtmq0rzY/tza+iL2AOMQzTaxlN6a3u7jssWY7crjYsd69I7nBN7Y6ni21Mo0/TDwqto21WTuiKW+6P7EDPiAyXTTOLF2l5jypOoSxsz1Bv96r5zWQuXo6mMDqOpg63Dk1MHQ+hz6HOB08bLghNnS8Nr5mMiu9b7xhqzU45a5uu5evsriQNAY5HSDXNry7jjgMN0q2ILcBOvavkTmUvwy8ZjzLM/68n7xtto+s57U0NxnBXTVTN3yyfr6PuO2/c7MYutO0mz1MtXY3xjV3PYkxmb1ju9w9fqo1Pmq6rTkzsDLAN7jFPHC5cSw1PQa73bySPJC4xjhFLVy0Jry2Nme54rxDu9s1srphuNOmQilstsa7SDyKQTk6RTyLPiw6iS8pOqAqZTTnPJE3DjZUsBU9lrWUOmI8ujhQOkQ1ZMEoOQOpJb/JtUO5Yy4KNySuLjoisCU4/jkZrp07Wza/wAC3/6y9Oq07ezKlvRo1YzkoNs+61r1TOdQ9GBsHOgk8Lr+5O8syWDPQNw857bvQsVAyE0BhL5C4uDvYOfA+dzl4tlarED3zLjCuF74CNvm+WDtUvNewrLwhsDW9uL0/vTk3x6wiuYw0MzaatM87gbrGvgC4tbi6PNS1YTaAulg6JrxItmM8Bz14OHk9nL17P7s7YDpFuFk2RrEVOEw0rsHrKra5orqEPMy5ALauwaa487bYuV8/rDhlvdM5GyqCvqq9bS2nPAIrUzxYKhE1orjuvGS+DLsBphQ26C5QOx40JDpLN5607jS6tvy34yxcOv+6SMEBwOU4kTr3NJe1pLz1sQY4djT8O4WtAytQtTdAyzgeODUxTDvUvXs9mDbhPby1Q7qbNSE9WT9FwJM5ezAPQZ268bZwua885TxhvPY2/TEbukk8xj+BPLEynbWcL3o4aDWFvoi9Eb/cPfiw80ErNma4dD2QsO2r/7lftNW95zSTujM9V7gDNkO+uDSKug08g72MORy3prI9u+y5F8LDt+c9wj5Ishs/OzYwwiC5gz7VQHIzajj7s0G0MDhcsje5L7hTvum8Gr1QopM4oK5vNYW7Yjg3uIM2KrvlNTdAYL4osRQ/dDsPsoQ7tzlJu00wozlILOC+JrvCvgPAZjhuu7Q/7L7MO7ywi7YLtus8QjkROx+8Z7lxvCm0Or36N1m+uS9pPZk1JTVPPh0+GCEZvHQwK7J/u1u1Yri8rbO9PzwaPe06yTgntuc0UbqfMyw9TzvZQJo0HT11Kng98r8eOJE7Jhr4tH88f7wRtOS7BjdUNis4k0CDvRW0ADpIOdQ4Sj05POC/KrTrPb8sgj2oOI61JbhTttaycLudPFk46bhsOrQ8RDD1tOO7IjRfthI+ebAPubC/9LiwO0Q1oDnKPtA4rbtAtRUx76gLsZK88zMoOZq9ea4GN9E/Eq4XPNS357gVvkw+YrdyNZI5zCRKsTcuVrJSvno1NypRv8k5xTj9Oz0rqjoXvHQ1zLHwPL86l7DNtPo3V6j2PSO2Ejf7uMe2STgsPPA6hTadtgK8dLx3KvC8sC0LvOy22b2CtkuwkLw2usa81jmTOI++bTjBvNiz3jkqawgwEApCA2IyaEpgCjiLPTS6HD43QDS4iz+AOFC6UDRyuwi7mbtnvbk7drI1sQEuZDSbNc448TKVtWw8oLa5OuM8mrIKNde9bblNqty8Vr2yLi46p6yPvIk+Dy3auoa48LIQuD4yKjrIvLM7Kq4CCAMIMBAKQgNXM2hKoALVvydC+7TvuAGtnLm1vSQ0xT56Kr+4CqlzPmu+TCpSOY4mCLg1vfGxorYqOSmohrpiuRo8VbmcOHC5/LHqvKG8FzscPvQyca7lPMq+HLZnv0u3eC5YN0W4szzptQO8rb27O6I3zKqTPMO2bzsqvnoxs7rWuOe5uqwXvH+6JDMKOqq0S731Lne497ivN2G0D7GDu2S8PDk7OPg4fUD9udm7ojw7uj+x3EHjs+I7C7LMNLG4WD2Juxw5tro+ubmuRDDVuAQ5j0DJvHm5cT06O1Y8vsAtvXm2Qz2OOaqpGLxsOGq1ZLaVvIS9Jiylwg4s1b1FOqs45EAHsvstV6vBPg6spT/5OMisGLxpuHu/+rg9umrAQcC0QG69LbeLQHS41LgqEQgDEApCA2IzaEoGVjhVt069WhsKA29icxIUChIIARIOCgcSBWJhdGNoCgMI4AFiHQoGYWN0aW9uEhMKEQgBEg0KBxIFYmF0Y2gKAggDQgQKABAR", "entry_sha256": "67d08611f9c0987bd72670d45da71a015e760cca3fb5489d5e7bcdd2a11afb39", "entry_bytes": 27117, "ops": ["Cast", "Gemm", "Relu"], "tried": ["All scores are board-formula score per match from the named model's side, random spawns, on seeds never used in training or selection unless stated. Harness check: current rule vs Starter 0.9926666666666667 over 3,000 (2,978 wins, 22 losses; its seat measured 0.9933333333333333).", "Starter vs the four built-ins: 0.965625 over 800 (harness reference).", "Behaviour cloning of Starter into 224-48-48-3: shoot sign agreement 0.92 on held-out decisions, yet only 0.200 vs Starter over 3,000, 0.005 vs current over 1,000, 0.928 vs built-ins over 800.", "ES self-play phase 1 (league 3/8 current, 2/8 Starter, 2/8 past selves, 1/8 built-ins; fitness score + 0.05 x HP diff / 5), 8.5 min, about 1,360 generations: 0.20 to 0.979 vs Starter over 3,000 (fp32 0.977, fp16 0.979, nothing lost in shrinking), but 0.0 vs current over 1,000 and built-ins fell to 0.867.", "Phase 2, more weight on current (7/16) plus survival-time shaping, 8.5 min: no gain vs current (about 0.005 on validation), Starter fell to about 0.93. Rejected.", "Phase 3, the current rule as the only opponent, 8 min: 0.1735 vs current over 1,000 paired, the only real dent in the rule, but 0.782 vs Starter and 0.12 vs built-ins; it learned one narrow trick. Rejected as final, kept as a start.", "Phase 4, balanced league from phase 3 (6 current, 3 Starter, 3 past selves, 4 built-ins per 16 games), 17 min: the gain against current washed out (0.007 over 1,000 held-out) while Starter and built-ins came back. Non-transitive results: phase-4-last beats phase-3 0.729 and phase-4-best 0.615, but loses to phase 1 0.454.", "Shrinking check on the final model: fp32 0.95433 and fp16 0.95708 vs Starter on the same 3,000 seeds; the ONNX file matches the native forward exactly.", "Exploiter: an ES-trained net from the Starter imitation, against the frozen final model, reaches 0.99 within 6 minutes. The final model scores 0.008 against it over 500.", "BEST (final) phase-4-last, 27,117 bytes: 0.9570833333333333 vs Starter over 3,000; 0.924375 vs built-ins over 800 (chaser 0.92, dogfighter 0.79, ace 0.945, brawler 0.989 over 200 each); 6 wins, 993 losses, 1 draw vs the current rule over 1,000 (0.0065); 0.008 vs its 10-minute exploiter over 500; 0.986 vs its own imitation start. Far weaker than the current entry: do not upload it."], "next": "Self-play from a Starter imitation does not reach the current rule in about an hour of ES on a 48-unit MLP: the rule beats it 993 of 1,000, and a 10-minute exploiter beats it 496 of 500. The only progress against the rule came from training against it alone (0.17), and the league then washed that out. The next seat on this method should start from a network distilled from the current rule (DAgger on the rule's actions), not from Starter, and run self-play from there with the rule and exploiters kept in the league. It should also add each exploiter to the league as it is found (PFSP weighting toward opponents it loses to), use more matches per candidate or a gradient method (PPO) instead of ES at this budget, and pick the final model by a round robin on fresh seeds, because results here were non-transitive (0.454/0.729 cycles). Harness note: this machine had no C compiler; pip's ziglang works as cc for cargo, and a native Rust port of Starter and of the rule graph runs about 3,000 matches/s, far faster than ONNX Runtime."}