{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:TK47WXRH5QF4P5ZQZ6RGUBBMRZ","short_pith_number":"pith:TK47WXRH","schema_version":"1.0","canonical_sha256":"9ab9fb5e27ec0bc7f730cfa26a042c8e4ea346e09a6a68b9827e480927492daa","source":{"kind":"arxiv","id":"2410.20399","version":1},"attestation_state":"computed","paper":{"title":"ThunderKittens: Simple, Fast, and Adorable AI Kernels","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Aaryan Singhal, Benjamin F. Spector, Christopher R\\'e, Daniel Y. Fu, Simran Arora","submitted_at":"2024-10-27T10:07:16Z","abstract_excerpt":"The challenge of mapping AI architectures to GPU hardware is creating a critical bottleneck in AI progress. Despite substantial efforts, hand-written custom kernels fail to meet their theoretical performance thresholds, even on well-established operations like linear attention. The diverse hardware capabilities of GPUs might suggest that we need a wide variety of techniques to achieve high performance. However, our work explores whether a small number of key abstractions can drastically simplify the process. We present ThunderKittens (TK), a framework for writing performant AI kernels while re"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2410.20399","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.LG","submitted_at":"2024-10-27T10:07:16Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"72ffddaebab8ff985612ccb3410de339c88cedb1a0c33db1e6e9a9fa92d9490f","abstract_canon_sha256":"7548f3555cf94b260c2e811073e64e97109f4ae5e13ee069e26c40d5ad0ad585"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:26:44.331660Z","signature_b64":"L47SBs06ReYZxXoRb7m+7TDRMmxMee1FYjS7Mhca+K2cBj3+RZKg2SDqgARPRTz30+TC5FCIyXzwR8pEw3RSDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9ab9fb5e27ec0bc7f730cfa26a042c8e4ea346e09a6a68b9827e480927492daa","last_reissued_at":"2026-07-05T09:26:44.331152Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:26:44.331152Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ThunderKittens: Simple, Fast, and Adorable AI Kernels","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Aaryan Singhal, Benjamin F. Spector, Christopher R\\'e, Daniel Y. Fu, Simran Arora","submitted_at":"2024-10-27T10:07:16Z","abstract_excerpt":"The challenge of mapping AI architectures to GPU hardware is creating a critical bottleneck in AI progress. Despite substantial efforts, hand-written custom kernels fail to meet their theoretical performance thresholds, even on well-established operations like linear attention. The diverse hardware capabilities of GPUs might suggest that we need a wide variety of techniques to achieve high performance. However, our work explores whether a small number of key abstractions can drastically simplify the process. We present ThunderKittens (TK), a framework for writing performant AI kernels while re"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.20399","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2410.20399/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2410.20399","created_at":"2026-07-05T09:26:44.331218+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.20399v1","created_at":"2026-07-05T09:26:44.331218+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.20399","created_at":"2026-07-05T09:26:44.331218+00:00"},{"alias_kind":"pith_short_12","alias_value":"TK47WXRH5QF4","created_at":"2026-07-05T09:26:44.331218+00:00"},{"alias_kind":"pith_short_16","alias_value":"TK47WXRH5QF4P5ZQ","created_at":"2026-07-05T09:26:44.331218+00:00"},{"alias_kind":"pith_short_8","alias_value":"TK47WXRH","created_at":"2026-07-05T09:26:44.331218+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":13,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.21228","citing_title":"Sakana Fugu Technical Report","ref_index":187,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09682","citing_title":"AutoMegaKernel: A Statically-Checked Agent Harness for Self-Retargeting Megakernel Synthesis","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30325","citing_title":"Veda: Scalable Video Diffusion via Distilled Sparse Attention","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2511.02043","citing_title":"Flashlight: PyTorch Compiler Extensions to Accelerate Attention Variants","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19269","citing_title":"CODA: Rewriting Transformer Blocks as GEMM-Epilogue Programs","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19269","citing_title":"CODA: Rewriting Transformer Blocks as GEMM-Epilogue Programs","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19652","citing_title":"Characterizing Real-World Bugs in Tile Programs for Automated Bug Detection","ref_index":65,"is_internal_anchor":false},{"citing_arxiv_id":"2509.19349","citing_title":"ShinkaEvolve: Towards Open-Ended And Sample-Efficient Program Evolution","ref_index":153,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10905","citing_title":"TLX: Hardware-Native, Evolvable MIMW GPU Compiler for Large-scale Production Environments","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10905","citing_title":"TLX: Hardware-Native, Evolvable MIMW GPU Compiler for Large-scale Production Environments","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2604.26074","citing_title":"DAK: Direct-Access-Enabled GPU Memory Offloading with Optimal Efficiency for LLM Inference","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14825","citing_title":"Nautilus: An Auto-Scheduling Tensor Compiler for Efficient Tiled GPU Kernels","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21221","citing_title":"Sparse Forcing: Native Trainable Sparse Attention for Real-time Autoregressive Diffusion Video Generation","ref_index":15,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TK47WXRH5QF4P5ZQZ6RGUBBMRZ","json":"https://pith.science/pith/TK47WXRH5QF4P5ZQZ6RGUBBMRZ.json","graph_json":"https://pith.science/api/pith-number/TK47WXRH5QF4P5ZQZ6RGUBBMRZ/graph.json","events_json":"https://pith.science/api/pith-number/TK47WXRH5QF4P5ZQZ6RGUBBMRZ/events.json","paper":"https://pith.science/paper/TK47WXRH"},"agent_actions":{"view_html":"https://pith.science/pith/TK47WXRH5QF4P5ZQZ6RGUBBMRZ","download_json":"https://pith.science/pith/TK47WXRH5QF4P5ZQZ6RGUBBMRZ.json","view_paper":"https://pith.science/paper/TK47WXRH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.20399&json=true","fetch_graph":"https://pith.science/api/pith-number/TK47WXRH5QF4P5ZQZ6RGUBBMRZ/graph.json","fetch_events":"https://pith.science/api/pith-number/TK47WXRH5QF4P5ZQZ6RGUBBMRZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TK47WXRH5QF4P5ZQZ6RGUBBMRZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TK47WXRH5QF4P5ZQZ6RGUBBMRZ/action/storage_attestation","attest_author":"https://pith.science/pith/TK47WXRH5QF4P5ZQZ6RGUBBMRZ/action/author_attestation","sign_citation":"https://pith.science/pith/TK47WXRH5QF4P5ZQZ6RGUBBMRZ/action/citation_signature","submit_replication":"https://pith.science/pith/TK47WXRH5QF4P5ZQZ6RGUBBMRZ/action/replication_record"}},"created_at":"2026-07-05T09:26:44.331218+00:00","updated_at":"2026-07-05T09:26:44.331218+00:00"}