{"assertions":[{"assertion_index":0,"assertion_text":"Latent Space The consensus read writes itself: DeepSeek does more with less, Chinese labs keep pushing on cost, and the per-token price war rolls on.","citation_check":null,"claim_ids":[83607,83608,83609],"link_only_urls":[]},{"assertion_index":1,"assertion_text":"Fireworks AI Put those two facts together.","citation_check":null,"claim_ids":[61332,61333],"link_only_urls":[]},{"assertion_index":2,"assertion_text":"Don't Worry About the Vase TechCrunch AI Part II arrived this week.","citation_check":null,"claim_ids":[7050,7052,61764],"link_only_urls":[]},{"assertion_index":3,"assertion_text":"OpenAI News OpenAI News That is a lobbying document and a marketing document arriving in the same week.","citation_check":null,"claim_ids":[83543,83544,83545],"link_only_urls":[]},{"assertion_index":4,"assertion_text":"OpenAI News Vendor-published wins are a reason to look rather than a reason to buy.","citation_check":null,"claim_ids":[75666],"link_only_urls":[]},{"assertion_index":5,"assertion_text":"Stanford HAI A Harvard study of 62 million workers points the same way, finding that companies adopting generative AI saw junior employment fall about 8% relative to non-adopters within 18 months, and Anthropic's own junior share of technical hires fell from 27% to 9% since early 2025.","citation_check":null,"claim_ids":[83587,83588,83590],"link_only_urls":[]},{"assertion_index":6,"assertion_text":"Nate Jones TechCrunch AI Set that beside the AI as Normal Technology argument that AI is not causing mass layoffs among software engineers, where the number behind the argument is a Federal Reserve finding that software engineer employment is growing about 3 percentage points a year slower than it would without AI.","citation_check":null,"claim_ids":[14719,80998],"link_only_urls":[]},{"assertion_index":7,"assertion_text":"AI as Normal Technology The two reads do not conflict.","citation_check":null,"claim_ids":[7098],"link_only_urls":[]},{"assertion_index":8,"assertion_text":"TechCrunch AI Those are company statements, and they do not overturn the measured read, because a press release naming AI is a claim about motive while the dashboards count heads.","citation_check":null,"claim_ids":[37473,37489],"link_only_urls":[]},{"assertion_index":9,"assertion_text":"Anthropic News Anthropic News Twice now METR has signed off on Anthropic's verdict while faulting the proof behind it, and this week makes a third review.","citation_check":null,"claim_ids":[63616,81235],"link_only_urls":[]},{"assertion_index":10,"assertion_text":"METR METR METR The pattern is the point.","citation_check":null,"claim_ids":[21611,21613,21614,21826,21827,21828,21831,21832,83423],"link_only_urls":[]},{"assertion_index":11,"assertion_text":"- DeepInfer is now runnable by anyone, so the first outside team to run it on its own traffic and publish a cache-hit figure will settle whether the economics travel. Latent Space","citation_check":null,"claim_ids":[83607,83608,83609],"link_only_urls":[]},{"assertion_index":12,"assertion_text":"- Mistral and NVIDIA have promised public benchmarks and open datasets for engineering AI, and the first release will show whether \"AI-native engineering\" can be measured rather than marketed. Mistral AI","citation_check":null,"claim_ids":[83631,83632,83635],"link_only_urls":[]},{"assertion_index":13,"assertion_text":"- Stanford's dashboards track the labor market in real time, so the next read on 22-to-25-year-old employment will show whether the decline is steepening. Stanford HAI","citation_check":null,"claim_ids":[83587,83588,83590],"link_only_urls":[]},{"assertion_index":14,"assertion_text":"- Policymakers' response to OpenAI's part II cyber blueprint, especially any move on the mandatory testing agency proposed in part I, is the signal that separates a lobbying document from a framework. OpenAI News Don't Worry About the Vase","citation_check":null,"claim_ids":[7050,7052,83545],"link_only_urls":[]},{"assertion_index":15,"assertion_text":"- Google's gestational-age ultrasound model, working from blind-sweep scans taken by novice operators, matched expert sonographers in a prospective multi-center study, and in a separate finding it held up in Nairobi with no local training data; the next evidence worth waiting for is a study where frontline workers act on the model's estimate in real care decisions. Google Research Blog","citation_check":null,"claim_ids":[83711,83712,83713],"link_only_urls":[]},{"assertion_index":16,"assertion_text":"Latent Space The Fireworks cache figures come from one vendor's single benchmark run on one model, so they illustrate the cost shape rather than establish a general rate.","citation_check":null,"claim_ids":[83607,83608,83609],"link_only_urls":[]},{"assertion_index":17,"assertion_text":"Fireworks AI The Sophos and Cloudflare results are published by OpenAI, not by the customers independently, and the labor evidence establishes an entry-level employment decline but cannot distinguish slower hiring from layoffs across the wider market.","citation_check":null,"claim_ids":[61332,61333],"link_only_urls":[]}],"brief_date":"2026-10-22","bundle_id":"evidence-bundle-v2-eac255a31efb3961e15841d234a2dcfbd20d0913d1fef56caee57a69bddf0c2a","claims":[{"assertion_indexes":[0,11,16],"citation_authority_claim_id":83607,"citation_checks":[],"claim_id":83607,"claim_text":"DeepSeek open-sourced its production inference stack, DeepInfer, during PyTorch Conf.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-22T06:53:04.004799+00:00","source_name":"Latent Space","source_url":"https://www.latent.space/p/ainews-deepseek-deepinfer-open-source"},{"assertion_indexes":[0,11,16],"citation_authority_claim_id":83608,"citation_checks":[],"claim_id":83608,"claim_text":"DeepSeek claims its production cluster serves 155 trillion tokens per day.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-22T06:53:04.004799+00:00","source_name":"Latent Space","source_url":"https://www.latent.space/p/ainews-deepseek-deepinfer-open-source"},{"assertion_indexes":[0,11,16],"citation_authority_claim_id":83609,"citation_checks":[],"claim_id":83609,"claim_text":"DeepSeek claims its production service achieves a 96% input-cache hit rate.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-22T06:53:04.004799+00:00","source_name":"Latent Space","source_url":"https://www.latent.space/p/ainews-deepseek-deepinfer-open-source"},{"assertion_indexes":[1,17],"citation_authority_claim_id":61332,"citation_checks":[],"claim_id":61332,"claim_text":"In the DeepSWE benchmark run, 99.6% of input tokens were cache hits.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-15T06:32:33.772675+00:00","source_name":"Fireworks AI","source_url":"https://fireworks.ai/blog/DeepSeek-V4.1-Flash-Astra"},{"assertion_indexes":[1,17],"citation_authority_claim_id":61333,"citation_checks":[],"claim_id":61333,"claim_text":"In the DeepSWE benchmark run, cache-hit charges accounted for 59.9% of the total cost.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-15T06:32:33.772675+00:00","source_name":"Fireworks AI","source_url":"https://fireworks.ai/blog/DeepSeek-V4.1-Flash-Astra"},{"assertion_indexes":[2,14],"citation_authority_claim_id":7050,"citation_checks":[],"claim_id":7050,"claim_text":"OpenAI released a document titled 'Democratic Governance of Frontier AI: A Blueprint For A Federal Framework'.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-02T20:07:03.401854+00:00","source_name":"Don't Worry About the Vase","source_url":"https://thezvi.substack.com/p/openai-offers-a-new-policy-blueprint"},{"assertion_indexes":[2,14],"citation_authority_claim_id":7052,"citation_checks":[],"claim_id":7052,"claim_text":"OpenAI's blueprint calls for the creation of a new agency or empowerment of an existing civilian agency (CAISI) to conduct mandatory evaluations of frontier models.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-02T20:07:03.401854+00:00","source_name":"Don't Worry About the Vase","source_url":"https://thezvi.substack.com/p/openai-offers-a-new-policy-blueprint"},{"assertion_indexes":[2],"citation_authority_claim_id":61764,"citation_checks":[],"claim_id":61764,"claim_text":"OpenAI supports a provision in the FRONTIER Act that would require top frontier labs to allow independent verification organizations to monitor model development.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-16T06:36:18.365142+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/09/15/openai-anthropic-google-have-been-in-talks-on-ai-safety-for-weeks/"},{"assertion_indexes":[3],"citation_authority_claim_id":83543,"citation_checks":[],"claim_id":83543,"claim_text":"Cloudflare utilizes the OpenAI Daybreak model to investigate vulnerabilities and propose fixes for review.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-22T06:50:29.266617+00:00","source_name":"OpenAI News","source_url":"https://openai.com/index/cloudflare-cyber"},{"assertion_indexes":[3],"citation_authority_claim_id":83544,"citation_checks":[],"claim_id":83544,"claim_text":"Cloudflare developers use Codex to build faster and improve code quality.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-22T06:50:29.266617+00:00","source_name":"OpenAI News","source_url":"https://openai.com/index/cloudflare-cyber"},{"assertion_indexes":[3,14],"citation_authority_claim_id":83545,"citation_checks":[],"claim_id":83545,"claim_text":"OpenAI has published a document titled 'Democratic governance of frontier AI, part II' which outlines a blueprint for empowering cyber defenders using frontier AI and strengthening safeguards.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-22T06:50:38.332479+00:00","source_name":"OpenAI News","source_url":"https://openai.com/index/empowering-cyber-defenders"},{"assertion_indexes":[4],"citation_authority_claim_id":75666,"citation_checks":[],"claim_id":75666,"claim_text":"Sophos uses OpenAI's Daybreak to cut cyber-threat investigation time by 96%.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-10T06:40:06.111965+00:00","source_name":"OpenAI News","source_url":"https://openai.com/index/sophos"},{"assertion_indexes":[5,13],"citation_authority_claim_id":83587,"citation_checks":[],"claim_id":83587,"claim_text":"Employment in AI-exposed jobs such as software development and customer service decreased by more than 4% in the year preceding the report.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-22T06:52:16.792633+00:00","source_name":"Stanford HAI","source_url":"https://hai.stanford.edu/news/tracking-ais-real-time-footprint-on-the-labor-market"},{"assertion_indexes":[5,13],"citation_authority_claim_id":83588,"citation_checks":[],"claim_id":83588,"claim_text":"Stanford HAI\u2019s Digital Economy Lab launched the AI Economic Indicators dashboards in June 2026 to track AI's real-time economic footprint.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-22T06:52:16.792633+00:00","source_name":"Stanford HAI","source_url":"https://hai.stanford.edu/news/tracking-ais-real-time-footprint-on-the-labor-market"},{"assertion_indexes":[5,13],"citation_authority_claim_id":83590,"citation_checks":[],"claim_id":83590,"claim_text":"Data from the Canaries Dashboard indicates that employment for 22- to 25-year-olds in software development and customer service has declined over 3% annually since ChatGPT's release in November 2022.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-22T06:52:16.792633+00:00","source_name":"Stanford HAI","source_url":"https://hai.stanford.edu/news/tracking-ais-real-time-footprint-on-the-labor-market"},{"assertion_indexes":[6],"citation_authority_claim_id":14719,"citation_checks":[],"claim_id":14719,"claim_text":"A Harvard study analyzing 62 million workers found that companies adopting generative AI saw junior employment drop by roughly 8% relative to non-adopters within 18 months.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-10T03:39:58.495413+00:00","source_name":"Nate Jones","source_url":"https://www.youtube.com/watch?v=awV2kJzh8zk"},{"assertion_indexes":[6],"citation_authority_claim_id":80998,"citation_checks":[],"claim_id":80998,"claim_text":"Anthropic reported that the proportion of junior technical hires dropped from 27% to 9% since early 2025.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-18T06:34:19.337100+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/10/17/anthropic-says-its-workforce-has-become-less-junior-and-is-spending-more-time-in-meetings/"},{"assertion_indexes":[7],"citation_authority_claim_id":7098,"citation_checks":[],"claim_id":7098,"claim_text":"Federal Reserve economists found that software engineer employment in the U.S. is growing slower post-ChatGPT by about 3 percentage points per year compared to a no-AI counterfactual.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-02T20:08:00.560720+00:00","source_name":"AI as Normal Technology","source_url":"https://www.normaltech.ai/p/why-ai-hasnt-replaced-software-engineers"},{"assertion_indexes":[8],"citation_authority_claim_id":37473,"citation_checks":[],"claim_id":37473,"claim_text":"Companies that have cited AI as a factor in job cuts have underperformed the Nasdaq by approximately 10% in the 30 trading days following their layoff announcements.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-26T06:28:16.544630+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/07/25/the-running-list-major-tech-layoffs-in-2026-where-employers-cited-ai/"},{"assertion_indexes":[8],"citation_authority_claim_id":37489,"citation_checks":[],"claim_id":37489,"claim_text":"Block cut 4,000 jobs, reducing its workforce to under 6,000, citing that AI intelligence tools and smaller teams are fundamentally changing how companies are built and run.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-26T06:28:16.544630+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/07/25/the-running-list-major-tech-layoffs-in-2026-where-employers-cited-ai/"},{"assertion_indexes":[9],"citation_authority_claim_id":63616,"citation_checks":[],"claim_id":63616,"claim_text":"Anthropic is in dialogue with METR and other nonprofit evaluators to pilot elements of embedded evaluation using their own funding.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-19T06:46:09.827182+00:00","source_name":"Anthropic News","source_url":"https://www.anthropic.com/news/accenture-embedded-evaluation"},{"assertion_indexes":[9],"citation_authority_claim_id":81235,"citation_checks":[],"claim_id":81235,"claim_text":"Anthropic is giving employee-like access to a team from METR to evaluate the safety of its systems.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-19T06:37:37.406875+00:00","source_name":"Anthropic News","source_url":"https://www.anthropic.com/news/public-benefit-fund"},{"assertion_indexes":[10],"citation_authority_claim_id":21611,"citation_checks":[],"claim_id":21611,"claim_text":"METR considers the evidence in Anthropic's February 11 Sabotage Risk Report to be generally clearly represented, though additional analysis could improve it.","correction_role":null,"corroborates_eligible":true,"label":"opinion","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-14T06:50:46.024825+00:00","source_name":"METR","source_url":"https://metr.substack.com/p/2026-03-12-sabotage-risk-report-opus-4-6-review"},{"assertion_indexes":[10],"citation_authority_claim_id":21613,"citation_checks":[],"claim_id":21613,"claim_text":"METR disagrees with Anthropic regarding the sensitivity of the alignment assessment for Claude Opus 4.6, citing risks from evaluation awareness and undetected low-severity misaligned behaviors.","correction_role":null,"corroborates_eligible":true,"label":"opinion","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-14T06:50:46.024825+00:00","source_name":"METR","source_url":"https://metr.substack.com/p/2026-03-12-sabotage-risk-report-opus-4-6-review"},{"assertion_indexes":[10],"citation_authority_claim_id":21614,"citation_checks":[],"claim_id":21614,"claim_text":"METR agrees with Anthropic that the risk of catastrophic outcomes substantially enabled by Claude Opus 4.6's misaligned actions is very low but not negligible.","correction_role":null,"corroborates_eligible":true,"label":"opinion","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-14T06:50:46.024825+00:00","source_name":"METR","source_url":"https://metr.substack.com/p/2026-03-12-sabotage-risk-report-opus-4-6-review"},{"assertion_indexes":[10],"citation_authority_claim_id":21826,"citation_checks":[],"claim_id":21826,"claim_text":"METR reviewed the \"Risks from automated R&D\" section of Anthropic's February 2026 Risk Report.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-14T06:55:37.204693+00:00","source_name":"METR","source_url":"https://metr.substack.com/p/2026-05-08-rd-section-anthropic-risk-report-feb-2026-review"},{"assertion_indexes":[10],"citation_authority_claim_id":21827,"citation_checks":[],"claim_id":21827,"claim_text":"Anthropic's February 2026 Risk Report argues that the catastrophic risk from Claude Opus 4.6 or a less capable Anthropic model automating R&D in any domain is very low.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-14T06:55:37.204693+00:00","source_name":"METR","source_url":"https://metr.substack.com/p/2026-05-08-rd-section-anthropic-risk-report-feb-2026-review"},{"assertion_indexes":[10],"citation_authority_claim_id":21828,"citation_checks":[],"claim_id":21828,"claim_text":"METR does not believe Anthropic's report adequately supports its conclusion that catastrophic risk from R&D automation is very low.","correction_role":null,"corroborates_eligible":true,"label":"opinion","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-14T06:55:37.204693+00:00","source_name":"METR","source_url":"https://metr.substack.com/p/2026-05-08-rd-section-anthropic-risk-report-feb-2026-review"},{"assertion_indexes":[10],"citation_authority_claim_id":21831,"citation_checks":[],"claim_id":21831,"claim_text":"METR agrees with the bottom-line conclusion that the risk of catastrophe from Opus 4.6 or a less capable model automating R&D is very low.","correction_role":null,"corroborates_eligible":true,"label":"opinion","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-14T06:55:37.204693+00:00","source_name":"METR","source_url":"https://metr.substack.com/p/2026-05-08-rd-section-anthropic-risk-report-feb-2026-review"},{"assertion_indexes":[10],"citation_authority_claim_id":21832,"citation_checks":[],"claim_id":21832,"claim_text":"METR believes the evidence presented in Anthropic's report is inadequate to establish the low risk of catastrophe from R&D automation.","correction_role":null,"corroborates_eligible":true,"label":"opinion","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-14T06:55:37.204693+00:00","source_name":"METR","source_url":"https://metr.substack.com/p/2026-05-08-rd-section-anthropic-risk-report-feb-2026-review"},{"assertion_indexes":[10],"citation_authority_claim_id":83423,"citation_checks":[],"claim_id":83423,"claim_text":"METR agrees with Anthropic that the risk of catastrophic harm from misalignment is low for the models covered in Anthropic's August 2026 Risk Report.","correction_role":null,"corroborates_eligible":true,"label":"opinion","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-22T06:45:28.195721+00:00","source_name":"METR","source_url":"https://metr.substack.com/p/2026-10-20-misalignment-risk-report-august-2026-review"},{"assertion_indexes":[12],"citation_authority_claim_id":83631,"citation_checks":[],"claim_id":83631,"claim_text":"Mistral and NVIDIA are collaborating to integrate Mistral's AI models with NVIDIA's accelerated computing and physics modeling tools to create AI-native engineering workflows.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-22T06:54:16.744202+00:00","source_name":"Mistral AI","source_url":"https://mistral.ai/news/mistral-nvidia-ai-for-industrial-engineering/"},{"assertion_indexes":[12],"citation_authority_claim_id":83632,"citation_checks":[],"claim_id":83632,"claim_text":"Mistral and NVIDIA will develop public benchmarks and open datasets to evaluate physics models and agentic AI in computer-aided engineering (CAE) and electronic design automation (EDA).","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-22T06:54:16.744202+00:00","source_name":"Mistral AI","source_url":"https://mistral.ai/news/mistral-nvidia-ai-for-industrial-engineering/"},{"assertion_indexes":[12],"citation_authority_claim_id":83635,"citation_checks":[],"claim_id":83635,"claim_text":"Mistral Large 4, also known as Le Chonk, is a multimodal model with approximately 1 trillion total parameters and 49 billion active parameters, trained on 3,800 NVIDIA Grace Blackwell GPUs.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-22T06:54:16.744202+00:00","source_name":"Mistral AI","source_url":"https://mistral.ai/news/mistral-nvidia-ai-for-industrial-engineering/"},{"assertion_indexes":[15],"citation_authority_claim_id":83711,"citation_checks":[],"claim_id":83711,"claim_text":"A prospective multi-center study found that an AI model for gestational age assessment achieved non-inferior accuracy compared to expert sonographers when using blind sweep ultrasounds from novice operators. The AI model had a mean absolute error (MAE) of 4.18 days, while the expert sonographers had an MAE of 4.48 days.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-22T06:58:15.545708+00:00","source_name":"Google Research Blog","source_url":"https://research.google/blog/empowering-frontline-maternal-care-with-ai-assisted-ultrasound/"},{"assertion_indexes":[15],"citation_authority_claim_id":83712,"citation_checks":[],"claim_id":83712,"claim_text":"The AI system demonstrated cross-setting generalization, achieving a mean absolute error (MAE) of 4.10 days in Chicago and 4.27 days in Nairobi, where no local data was used for fine-tuning.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-22T06:58:15.545708+00:00","source_name":"Google Research Blog","source_url":"https://research.google/blog/empowering-frontline-maternal-care-with-ai-assisted-ultrasound/"},{"assertion_indexes":[15],"citation_authority_claim_id":83713,"citation_checks":[],"claim_id":83713,"claim_text":"In cases of fetal weight extremes (above the 90th percentile), the research AI system outperformed traditional biometry formulas.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-22T06:58:15.545708+00:00","source_name":"Google Research Blog","source_url":"https://research.google/blog/empowering-frontline-maternal-care-with-ai-assisted-ultrasound/"}],"version":"evidence-bundle-v2"}