scan-v5.json (26980B)
1 { 2 "scan_version": 5, 3 "paper_type": "empirical", 4 "paper": { 5 "title": "eSapiens: A Platform for Secure and Auditable Retrieval-Augmented Generation", 6 "authors": [ 7 "Isaac Shi", 8 "Zeyuan Li", 9 "Fan Liu", 10 "Wenli Wang", 11 "Lewei He" 12 ], 13 "year": 2025, 14 "venue": "arXiv.org", 15 "arxiv_id": "2507.09588", 16 "doi": "10.48550/arXiv.2507.09588" 17 }, 18 "checklist": { 19 "claims_and_evidence": { 20 "abstract_claims_supported": { 21 "applies": true, 22 "answer": false, 23 "justification": "The abstract states 'chunk size of 512 tokens yields the highest retrieval precision (Top-3 accuracy: 91.3%)' but the experiments use 500 and 1000 token chunks exclusively; no 91.3% figure appears anywhere in Tables 3 or 4.", 24 "source": "haiku" 25 }, 26 "causal_claims_justified": { 27 "applies": true, 28 "answer": false, 29 "justification": "The paper claims '23% improvement in factual alignment' and various business outcome improvements without controls, randomization, or ablation design sufficient for causal attribution.", 30 "source": "haiku" 31 }, 32 "generalization_bounded": { 33 "applies": true, 34 "answer": false, 35 "justification": "The paper broadly claims suitability for 'high-stakes domains like legal and finance' and enterprise deployment based only on 100 questions from RAGtruth and four LegalBench subsets, without bounding the scope.", 36 "source": "haiku" 37 }, 38 "alternative_explanations_discussed": { 39 "applies": true, 40 "answer": false, 41 "justification": "No alternative explanations are discussed; the higher hallucination rate of eSapiens vs FAISS is attributed solely to prompt flexibility without considering other causes.", 42 "source": "haiku" 43 }, 44 "proxy_outcome_distinction": { 45 "applies": true, 46 "answer": false, 47 "justification": "The abstract claims '23% improvement in factual alignment' but measures 'Context Relevance' (a retrieval proxy); business outcome claims like '10x speedup in reporting' have no measured basis.", 48 "source": "haiku" 49 } 50 }, 51 "limitations_and_scope": { 52 "limitations_section_present": { 53 "applies": true, 54 "answer": false, 55 "justification": "There is no dedicated limitations or threats-to-validity section anywhere in the paper.", 56 "source": "haiku" 57 }, 58 "threats_to_validity_specific": { 59 "applies": true, 60 "answer": false, 61 "justification": "No threats to validity are discussed; the small sample size (100 questions), self-evaluation bias, and limited baselines are not acknowledged.", 62 "source": "haiku" 63 }, 64 "scope_boundaries_stated": { 65 "applies": true, 66 "answer": false, 67 "justification": "No explicit scope boundaries are stated; the paper does not articulate what its results do not show or under what conditions the system would fail.", 68 "source": "haiku" 69 } 70 }, 71 "conflicts_of_interest": { 72 "funding_disclosed": { 73 "applies": true, 74 "answer": false, 75 "justification": "No funding source is disclosed anywhere in the paper.", 76 "source": "haiku" 77 }, 78 "affiliations_disclosed": { 79 "applies": true, 80 "answer": true, 81 "justification": "All authors are listed as 'eSapiens Team' with the company URL https://www.esapiens.ai/, making clear they are employees evaluating their own commercial product.", 82 "source": "haiku" 83 }, 84 "funder_independent_of_outcome": { 85 "applies": true, 86 "answer": false, 87 "justification": "The authors are the eSapiens Team directly evaluating their own commercial platform; there is no independent funder.", 88 "source": "haiku" 89 }, 90 "financial_interests_declared": { 91 "applies": true, 92 "answer": false, 93 "justification": "No competing interests statement, patent disclosure, or financial interest declaration appears anywhere in the paper.", 94 "source": "haiku" 95 } 96 }, 97 "scope_and_framing": { 98 "key_terms_defined": { 99 "applies": true, 100 "answer": false, 101 "justification": "Key claims like 'trustworthy', 'auditable', and 'secure' are used throughout without operational definitions; 'auditability' is never defined in terms of what can actually be audited or verified.", 102 "source": "haiku" 103 }, 104 "intended_contribution_clear": { 105 "applies": true, 106 "answer": true, 107 "justification": "The paper clearly states it introduces eSapiens as an enterprise RAG platform combining document ingestion, hybrid vector retrieval, and no-code workflow orchestration for business use.", 108 "source": "haiku" 109 }, 110 "engagement_with_prior_work": { 111 "applies": true, 112 "answer": true, 113 "justification": "Section 3 engages with prior RAG frameworks (Lewis et al. 2020, FiD), orchestration tools (LangChain, LlamaIndex), agent systems (Gorilla, Toolformer), and domain-specific systems (ChatLaw, Lawyer-LLM), explaining how eSapiens differs.", 114 "source": "haiku" 115 } 116 } 117 }, 118 "type_checklist": { 119 "empirical": { 120 "artifacts": { 121 "code_released": { 122 "applies": true, 123 "answer": false, 124 "justification": "No source code is released; the paper only links to the commercial website https://www.esapiens.ai/.", 125 "source": "haiku" 126 }, 127 "data_released": { 128 "applies": true, 129 "answer": true, 130 "justification": "The evaluation uses publicly available benchmarks: LegalBench subsets (PrivacyQA, CUAD, MAUD, ContractNLI) and RAGtruth, all standard public benchmarks used unmodified.", 131 "source": "haiku" 132 }, 133 "environment_specified": { 134 "applies": true, 135 "answer": false, 136 "justification": "No requirements.txt, Dockerfile, or reproducible environment specs are provided; the tech stack is described at a high level (LangChain, Elasticsearch 8.x) without version-pinned dependencies.", 137 "source": "haiku" 138 }, 139 "reproduction_instructions": { 140 "applies": true, 141 "answer": false, 142 "justification": "No step-by-step reproduction instructions are provided for either experiment; the platform is a commercial SaaS product without public API access or self-hosted option described.", 143 "source": "haiku" 144 } 145 }, 146 "statistical_methodology": { 147 "confidence_intervals_or_error_bars": { 148 "applies": true, 149 "answer": false, 150 "justification": "No confidence intervals or error bars are reported in Tables 3, 4, or 5.", 151 "source": "haiku" 152 }, 153 "significance_tests": { 154 "applies": true, 155 "answer": false, 156 "justification": "No statistical significance tests are applied despite multiple comparative claims between eSapiens and the FAISS baseline across five LLMs.", 157 "source": "haiku" 158 }, 159 "effect_sizes_reported": { 160 "applies": true, 161 "answer": false, 162 "justification": "Raw scores are reported but effect sizes with baseline context are not formally reported; the abstract's '23% improvement' figure is not labeled as a relative effect in the results tables.", 163 "source": "haiku" 164 }, 165 "sample_size_justified": { 166 "applies": true, 167 "answer": false, 168 "justification": "The generation quality evaluation uses '100 random questions from RAGtruth' with no justification for this choice and no power analysis.", 169 "source": "haiku" 170 }, 171 "variance_reported": { 172 "applies": true, 173 "answer": false, 174 "justification": "No variance, standard deviation, or confidence spread is reported for any metric across Tables 3, 4, or 5.", 175 "source": "haiku" 176 } 177 }, 178 "evaluation_design": { 179 "baselines_included": { 180 "applies": true, 181 "answer": true, 182 "justification": "A FAISS-based DEREK pipeline is used as a baseline for the generation quality evaluation in Appendix B across all five LLMs.", 183 "source": "haiku" 184 }, 185 "baselines_contemporary": { 186 "applies": true, 187 "answer": false, 188 "justification": "The only baseline is the authors' own FAISS implementation; no established competitive RAG systems (LlamaIndex, LangChain RAG, commercial platforms) are included as comparators.", 189 "source": "haiku" 190 }, 191 "ablation_study": { 192 "applies": true, 193 "answer": true, 194 "justification": "A chunk size comparison (500 vs 1000 tokens) is conducted in Appendix A across four datasets, functioning as a limited ablation of a key hyperparameter.", 195 "source": "haiku" 196 }, 197 "multiple_metrics": { 198 "applies": true, 199 "answer": true, 200 "justification": "Multiple metrics are used: Recall@k and Precision@k for retrieval; Completeness, Utilization, Context Relevance, pc_hallucinated, and Accuracy for generation quality.", 201 "source": "haiku" 202 }, 203 "human_evaluation": { 204 "applies": true, 205 "answer": true, 206 "justification": "Table 5 includes 'Accuracy: Human-graded alignment with ground truth' as an evaluation dimension, constituting human evaluation of system outputs.", 207 "source": "haiku" 208 }, 209 "held_out_test_set": { 210 "applies": true, 211 "answer": true, 212 "justification": "Evaluation uses standard benchmark questions from LegalBench subsets and RAGtruth (100 random questions) not used for system training or development.", 213 "source": "haiku" 214 }, 215 "per_category_breakdown": { 216 "applies": true, 217 "answer": true, 218 "justification": "Retrieval results are broken down per dataset (PrivacyQA, CUAD, MAUD, ContractNLI) and generation quality is broken down per LLM model.", 219 "source": "haiku" 220 }, 221 "failure_cases_discussed": { 222 "applies": true, 223 "answer": true, 224 "justification": "The paper acknowledges that eSapiens has higher hallucination rates than FAISS and lower completeness, offering brief explanations for each failure mode.", 225 "source": "haiku" 226 }, 227 "negative_results_reported": { 228 "applies": true, 229 "answer": true, 230 "justification": "Table 5 clearly shows FAISS outperforms eSapiens on pc_hallucinated and completeness across all models; these negative results are reported and briefly analyzed.", 231 "source": "haiku" 232 } 233 }, 234 "setup_transparency": { 235 "model_versions_specified": { 236 "applies": true, 237 "answer": false, 238 "justification": "Model names like 'GPT-4o', 'Claude 3.7', 'Gemini 1.5 Pro' are used without snapshot dates or exact API version IDs (e.g., no 'gpt-4o-2024-05-13' or 'claude-3-7-sonnet-20250219').", 239 "source": "haiku" 240 }, 241 "prompts_provided": { 242 "applies": true, 243 "answer": false, 244 "justification": "Appendix C provides example SQL outputs but not the prompts used for TRACe evaluation or retrieval experiments; the CO-STAR format is mentioned but actual evaluation prompts are not shown.", 245 "source": "haiku" 246 }, 247 "hyperparameters_reported": { 248 "applies": true, 249 "answer": false, 250 "justification": "No temperature, top-p, max tokens, or other LLM hyperparameters are reported for any model in either experiment.", 251 "source": "haiku" 252 }, 253 "scaffolding_described": { 254 "applies": true, 255 "answer": true, 256 "justification": "The DEREK and THOR architectures, multi-agent workflows, LangGraph orchestration, query refinement, and hybrid retrieval pipeline are described with sufficient detail to understand the agentic scaffolding.", 257 "source": "haiku" 258 }, 259 "data_preprocessing_documented": { 260 "applies": true, 261 "answer": true, 262 "justification": "Document preprocessing is described: RecursiveCharacterTextSplitter for chunking (1000 tokens, 150 overlap), OpenAIEmbeddings for vectorization, and Elasticsearch for hybrid BM25 + dense vector retrieval.", 263 "source": "haiku" 264 } 265 }, 266 "data_integrity": { 267 "raw_data_available": { 268 "applies": true, 269 "answer": false, 270 "justification": "No raw evaluation data, question-answer pairs, or retrieval logs are made available for independent verification.", 271 "source": "haiku" 272 }, 273 "data_collection_described": { 274 "applies": true, 275 "answer": false, 276 "justification": "The 100 questions from RAGtruth are described only as 'random' without selection criteria or sampling procedure; no description of how the 100 questions were drawn.", 277 "source": "haiku" 278 }, 279 "recruitment_methods_described": { 280 "applies": false, 281 "answer": false, 282 "justification": "NA — standard public benchmarks were used; no participant recruitment was involved.", 283 "source": "haiku" 284 }, 285 "data_pipeline_documented": { 286 "applies": true, 287 "answer": false, 288 "justification": "The general pipeline (ingest → chunk → embed → index → retrieve → generate) is described at a high level but lacks sufficient detail to independently replicate the exact evaluation pipeline.", 289 "source": "haiku" 290 } 291 }, 292 "contamination": { 293 "training_cutoff_stated": { 294 "applies": true, 295 "answer": false, 296 "justification": "The paper evaluates multiple LLMs (GPT-4o, Claude 3.7, Gemini 1.5 Pro, DeepSeek R1) on legal benchmarks but states no model's training data cutoff.", 297 "source": "haiku" 298 }, 299 "train_test_overlap_discussed": { 300 "applies": true, 301 "answer": false, 302 "justification": "LegalBench subsets (CUAD, ContractNLI, etc.) and RAGtruth may overlap with LLM pre-training data; this is never mentioned.", 303 "source": "haiku" 304 }, 305 "benchmark_contamination_addressed": { 306 "applies": true, 307 "answer": false, 308 "justification": "The paper does not address whether benchmark examples from LegalBench or RAGtruth were available before the training cutoffs of the evaluated LLMs.", 309 "source": "haiku" 310 } 311 }, 312 "human_studies": { 313 "pre_registered": { 314 "applies": false, 315 "answer": false, 316 "justification": "NA — no human participant study.", 317 "source": "haiku" 318 }, 319 "irb_or_ethics_approval": { 320 "applies": false, 321 "answer": false, 322 "justification": "NA — no human participant study.", 323 "source": "haiku" 324 }, 325 "demographics_reported": { 326 "applies": false, 327 "answer": false, 328 "justification": "NA — no human participant study.", 329 "source": "haiku" 330 }, 331 "inclusion_exclusion_criteria": { 332 "applies": false, 333 "answer": false, 334 "justification": "NA — no human participant study.", 335 "source": "haiku" 336 }, 337 "randomization_described": { 338 "applies": false, 339 "answer": false, 340 "justification": "NA — no human participant study.", 341 "source": "haiku" 342 }, 343 "blinding_described": { 344 "applies": false, 345 "answer": false, 346 "justification": "NA — no human participant study.", 347 "source": "haiku" 348 }, 349 "attrition_reported": { 350 "applies": false, 351 "answer": false, 352 "justification": "NA — no human participant study.", 353 "source": "haiku" 354 } 355 }, 356 "cost_and_practicality": { 357 "inference_cost_reported": { 358 "applies": true, 359 "answer": false, 360 "justification": "The admin dashboard is described as tracking token spend, but no actual inference cost or latency figures are reported in the experiments.", 361 "source": "haiku" 362 }, 363 "compute_budget_stated": { 364 "applies": true, 365 "answer": false, 366 "justification": "No compute budget for either experiment is stated anywhere in the paper.", 367 "source": "haiku" 368 } 369 } 370 } 371 }, 372 "claims": [ 373 { 374 "claim": "Chunk size of 512 tokens yields the highest retrieval precision with Top-3 accuracy of 91.3%", 375 "evidence": "Experiments in Appendix A use chunk sizes of 500 and 1000 tokens only; no 512-token condition exists and no 91.3% figure appears in Tables 3 or 4", 376 "supported": "unsupported" 377 }, 378 { 379 "claim": "eSapiens delivers up to 23% improvement in factual alignment over FAISS baseline", 380 "evidence": "Table 5 shows Context Relevance for eSapiens-gpt4o-mini (0.3785) vs FAISS+gpt4o-mini (0.3090), yielding ~22.5% relative improvement on one metric for one model; other metrics favor FAISS; 'factual alignment' overstates what Context Relevance measures", 381 "supported": "weak" 382 }, 383 { 384 "claim": "Monthly financial reporting time fell from two hours to twelve minutes with eSapiens", 385 "evidence": "Cited only as an 'early adopter' report with no methodology, sample size, baseline, or controlled evaluation", 386 "supported": "unsupported" 387 }, 388 { 389 "claim": "Automatic ticket categorization accuracy rose by 40 percent", 390 "evidence": "Anecdotal early adopter claim with no supporting data, methodology, or baseline described", 391 "supported": "unsupported" 392 }, 393 { 394 "claim": "eSapiens shows higher context relevance than FAISS baseline", 395 "evidence": "Table 5 shows Context Relevance consistently higher for eSapiens across most models (0.26–0.50 vs FAISS 0.31–0.34), though FAISS outperforms on completeness and hallucination", 396 "supported": "moderate" 397 }, 398 { 399 "claim": "Chunk size 1000 is better for recall than 500 on most legal datasets", 400 "evidence": "Tables 3 and 4 show chunk=1000 outperforming chunk=500 at Recall@50 for CUAD (62.30% vs 55.66%), MAUD (13.60% vs 22.60% — reversed here), and ContractNLI (39.78% vs 46.90% — reversed); claim holds for CUAD but not PrivacyQA or ContractNLI", 401 "supported": "weak" 402 }, 403 { 404 "claim": "FAISS baseline achieves lower hallucination rates than eSapiens", 405 "evidence": "Table 5 shows pc_hallucinated consistently lower for FAISS (0.086–0.152) vs eSapiens (0.140–0.273) across all five models tested", 406 "supported": "strong" 407 } 408 ], 409 "methodology_tags": [ 410 "benchmark-eval", 411 "case-study" 412 ], 413 "key_findings": "eSapiens is an enterprise RAG platform that shows higher context relevance than a FAISS-based baseline on a 100-question subset of RAGtruth (up to ~22.5% relative improvement for GPT-4o-mini), but consistently has higher hallucination rates and lower completeness than the FAISS baseline. The paper contains a factual inconsistency between the abstract (claiming 512-token chunks and 91.3% Top-3 accuracy) and the actual experiments (using 500/1000 token chunks, with no 91.3% value appearing). Business outcome claims (60% cost reduction, 10x reporting speedup, 40% ticket accuracy gain) are anecdotal with no controlled evaluation methodology.", 414 "red_flags": [ 415 { 416 "flag": "Abstract-data inconsistency", 417 "detail": "Abstract claims 'chunk size of 512 tokens yields the highest retrieval precision (Top-3 accuracy: 91.3%)' but experiments use 500 and 1000 token chunks exclusively; no 91.3% figure appears anywhere in Tables 3 or 4." 418 }, 419 { 420 "flag": "Self-evaluation conflict of interest", 421 "detail": "All authors are eSapiens employees evaluating their own commercial platform; no independent validation, no competing interests statement, no external reviewers of the evaluation methodology." 422 }, 423 { 424 "flag": "Anecdotal business outcome claims", 425 "detail": "Marketing claims ('2 hours to 12 minutes reporting', '40% ticket accuracy improvement', '60% cost reduction', 'double-digit lead-to-deal velocity') are presented in the executive summary without any methodology, sample sizes, or controlled evaluation." 426 }, 427 { 428 "flag": "Weak baseline choice", 429 "detail": "The sole comparison baseline is the authors' own FAISS implementation; no established competitive RAG systems (LlamaIndex, Azure AI Search, commercial RAG APIs) are included, making relative performance claims uninterpretable." 430 }, 431 { 432 "flag": "No statistical rigor", 433 "detail": "No confidence intervals, significance tests, variance, or power analysis reported for any metric; single-run point estimates presented as definitive performance results." 434 }, 435 { 436 "flag": "Product paper masquerading as research paper", 437 "detail": "The paper is primarily a product description/marketing document with two limited appendix experiments; the empirical framing and arXiv submission context overstate the research contribution and evaluation depth." 438 } 439 ], 440 "cited_papers": [ 441 { 442 "title": "Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks", 443 "relevance": "Foundational RAG paper (Lewis et al. 2020) that eSapiens' DEREK engine builds upon for retrieval-augmented generation" 444 }, 445 { 446 "title": "Distilling Knowledge from Reader to Retriever for Question Answering (FiD)", 447 "relevance": "Early dense retrieval framework contextualizing eSapiens' hybrid retrieval approach" 448 }, 449 { 450 "title": "LangChain: Language Models in Chains", 451 "relevance": "Core orchestration framework used throughout eSapiens for prompt templating, tool calls, and agent coordination" 452 }, 453 { 454 "title": "LlamaIndex (GPT Index)", 455 "relevance": "Competing modular LLM application framework that eSapiens is positioned against as requiring more engineering effort" 456 }, 457 { 458 "title": "Gorilla: Large Language Model Connected with Massive APIs", 459 "relevance": "Prior work on autonomous agent tool use that contextualizes eSapiens' THOR module for structured data queries" 460 }, 461 { 462 "title": "Toolformer: Language Models Can Teach Themselves to Use Tools", 463 "relevance": "Prior work on LLM tool use that eSapiens contrasts against as lacking enterprise governance features" 464 }, 465 { 466 "title": "ChatLaw: Open-Source Legal Large Language Model", 467 "relevance": "Domain-specific legal LLM representing the vertical systems that eSapiens aims to generalize beyond" 468 }, 469 { 470 "title": "Lawyer LLM: An Expert-Level Chinese Legal Large Language Model", 471 "relevance": "Vertical legal AI representing the category of single-purpose systems with limited cross-domain applicability" 472 } 473 ], 474 "engagement_factors": { 475 "practical_relevance": { 476 "score": 2, 477 "justification": "Addresses real enterprise deployment needs (security, audit trails, no-code workflows, hybrid retrieval) but is a closed commercial product limiting practitioner adoption." 478 }, 479 "surprise_contrarian": { 480 "score": 0, 481 "justification": "No surprising findings; results confirm expected trade-offs between precision-oriented (FAISS) and fluency-oriented (eSapiens) RAG approaches." 482 }, 483 "fear_safety": { 484 "score": 1, 485 "justification": "Addresses enterprise data security, prompt injection mitigation, and regulatory compliance (SOC 2, GDPR, HIPAA) but framed as product features rather than research findings." 486 }, 487 "drama_conflict": { 488 "score": 0, 489 "justification": "No controversy or conflict; paper is a product description with minor empirical appendices." 490 }, 491 "demo_ability": { 492 "score": 2, 493 "justification": "A live commercial website with demos is referenced (esapiens.ai), though access terms and trial availability are not described." 494 }, 495 "brand_recognition": { 496 "score": 0, 497 "justification": "eSapiens is an unknown startup with no brand recognition in the AI research community." 498 } 499 }, 500 "hn_data": { 501 "threads": [ 502 { 503 "hn_id": "41039213", 504 "title": "Planck stars, White Holes, Remnants and Planck-mass quasi-particles", 505 "points": 62, 506 "comments": 32, 507 "url": "https://news.ycombinator.com/item?id=41039213" 508 }, 509 { 510 "hn_id": "43708789", 511 "title": "Eccfrog512ck2: An Enhanced 512-Bit Weierstrass Elliptic Curve [pdf]", 512 "points": 45, 513 "comments": 16, 514 "url": "https://news.ycombinator.com/item?id=43708789" 515 }, 516 { 517 "hn_id": "43701195", 518 "title": "Reasoning Models Can Be Effective Without Thinking", 519 "points": 21, 520 "comments": 2, 521 "url": "https://news.ycombinator.com/item?id=43701195" 522 }, 523 { 524 "hn_id": "32097013", 525 "title": "A Study of HTTP/2’s Server Push Performance Potential", 526 "points": 21, 527 "comments": 2, 528 "url": "https://news.ycombinator.com/item?id=32097013" 529 }, 530 { 531 "hn_id": "44607842", 532 "title": "BeePL: Correct-by-Compilation Kernel Extensions", 533 "points": 4, 534 "comments": 0, 535 "url": "https://news.ycombinator.com/item?id=44607842" 536 }, 537 { 538 "hn_id": "44755879", 539 "title": "TinyTroupe: An LLM-Powered Multiagent Persona Simulation Toolkit (OSS Paper)", 540 "points": 3, 541 "comments": 1, 542 "url": "https://news.ycombinator.com/item?id=44755879" 543 }, 544 { 545 "hn_id": "44639814", 546 "title": "Automated Hypothesis Validation with Agentic Sequential Falsifications", 547 "points": 3, 548 "comments": 0, 549 "url": "https://news.ycombinator.com/item?id=44639814" 550 }, 551 { 552 "hn_id": "43935110", 553 "title": "ZeroSearch: Incentivize the Search Capability of LLMs Without Searching", 554 "points": 2, 555 "comments": 0, 556 "url": "https://news.ycombinator.com/item?id=43935110" 557 }, 558 { 559 "hn_id": "43175116", 560 "title": "Maximizing Energy Efficiency in Subthreshold RISC-V Cores", 561 "points": 2, 562 "comments": 0, 563 "url": "https://news.ycombinator.com/item?id=43175116" 564 }, 565 { 566 "hn_id": "44583158", 567 "title": "TinyTroupe: An LLM-Powered Multiagent Persona Simulation Toolkit", 568 "points": 1, 569 "comments": 0, 570 "url": "https://news.ycombinator.com/item?id=44583158" 571 } 572 ], 573 "top_points": 62, 574 "total_points": 164, 575 "total_comments": 53 576 } 577 }