{"assertions":[{"assertion_index":0,"assertion_text":"Anthropic News Opus 5.5, released yesterday, pulls the Opus line toward the stricter end of that split.","citation_check":null,"claim_ids":[57366],"link_only_urls":[]},{"assertion_index":1,"assertion_text":"Anthropic News TechCrunch AI Snowflake Blog A smart buyer sees Fable-level output at Opus prices and moves the default.","citation_check":null,"claim_ids":[65178,65179,65180,65297,65305,65580,65581,65587],"link_only_urls":[]},{"assertion_index":2,"assertion_text":"Anthropic News AWS Machine Learning Blog Two months ago Opus 5 was the opposite trade.","citation_check":null,"claim_ids":[65178,65179,65180,65618,65621],"link_only_urls":[]},{"assertion_index":3,"assertion_text":"Don't Worry About the Vase Don't Worry About the Vase The lineage has reversed.","citation_check":null,"claim_ids":[37428,37429,38669],"link_only_urls":[]},{"assertion_index":4,"assertion_text":"AWS Machine Learning Blog That is a second reported benefit, not a second discount, because the evidence does not say how much of the token thrift is already inside the 40% figure and how much sits on top of it.","citation_check":null,"claim_ids":[65618,65621],"link_only_urls":[]},{"assertion_index":5,"assertion_text":"TechCrunch AI OpenAI News A principles document written by the company being assessed is the weakest form of independence.","citation_check":null,"claim_ids":[61213,65275],"link_only_urls":[]},{"assertion_index":6,"assertion_text":"Stanford HAI For a leader scoring vendor risk, the score stays where it is until an assessor has its own name and its own publishing rights.","citation_check":null,"claim_ids":[65340,65341,65342,65343,65353],"link_only_urls":[]},{"assertion_index":7,"assertion_text":"OpenRouter Blog That is a stronger signal than adoption counts and a weaker one than a neutral test, because it is a seller-published result.","citation_check":null,"claim_ids":[65287,65289],"link_only_urls":[]},{"assertion_index":8,"assertion_text":"Simon Willison's Weblog An engineer routing support triage can pilot Jev on the intents where it wins and keep logging the ones where it loses.","citation_check":null,"claim_ids":[64887,64889,64894],"link_only_urls":[]},{"assertion_index":9,"assertion_text":"Nate Jones Reach, not enthusiasm, set the sequence.","citation_check":null,"claim_ids":[65409,65410,65411,65412,65413],"link_only_urls":[]},{"assertion_index":10,"assertion_text":"OpenAI News My September 4 call that Astra would reach Plus users and API developers without a waitlist by September 18 graded as a provisional hit.","citation_check":null,"claim_ids":[65375,65376],"link_only_urls":[]},{"assertion_index":11,"assertion_text":"- Anthropic's refusal rate on Opus 5.5 is the number to track; the launch note says it refuses more than earlier Opus versions but gives no figure. AWS Machine Learning Blog","citation_check":null,"claim_ids":[65618,65621],"link_only_urls":[]},{"assertion_index":12,"assertion_text":"- OpenAI now has to attach names to its assessment principles; the document outlines priorities and principles, and a named assessor is the missing next step. OpenAI News","citation_check":null,"claim_ids":[65275],"link_only_urls":[]},{"assertion_index":13,"assertion_text":"- More model families are queued for Hugging Face's rebuilt tokenizer merge loop before 1.0.0 ships, on the back of a release candidate that encodes 3 to 30 times faster than version 0.23 on a single thread of one Apple M4 Max. Hugging Face Blog","citation_check":null,"claim_ids":[64980,64981,64982,64983,64984],"link_only_urls":[]},{"assertion_index":14,"assertion_text":"- OpenRouter has published a Jev cost benchmark and an accuracy test that reports one intent; if a fuller comparison appears, the useful version would break accuracy out by intent, including any where Opus beats Jev. OpenRouter Blog OpenRouter Blog","citation_check":null,"claim_ids":[64965,64966,64977,65287,65289],"link_only_urls":[]},{"assertion_index":15,"assertion_text":"- Stanford's panel treated an AI kill switch as a feasibility question, Stanford HAI while the public fight over pacing runs on a different track: Trump calls extinction risk a hoax and four lab chiefs agree the frontier must be paced. Don't Worry About the Vase","citation_check":null,"claim_ids":[65340,65341,65342,65343,65353,65394,65395,65396,65400],"link_only_urls":[]},{"assertion_index":16,"assertion_text":"Anthropic News Snowflake Blog Parallel's halving of time and cost is an OpenAI customer story, and the Jev accuracy and cost figures come from OpenRouter's own comparisons.","citation_check":null,"claim_ids":[65178,65179,65180,65297,65305],"link_only_urls":[]}],"brief_date":"2026-09-23","bundle_id":"evidence-bundle-v1-e428ec943e3518654da884c06bdf455cad8d535d7ec3e628dbf46a312e4f4e7a","claims":[{"assertion_indexes":[0],"citation_authority_claim_id":57366,"citation_checks":[],"claim_id":57366,"claim_text":"Claude Mythos 5.1 is identical to Claude Fable 5.1 but features different safeguards designed to support work in cybersecurity and the life sciences.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-07T06:41:32.777285+00:00","source_name":"Anthropic News","source_url":"https://www.anthropic.com/claude-fable-and-mythos-5-1"},{"assertion_indexes":[1,2,16],"citation_authority_claim_id":65178,"citation_checks":[],"claim_id":65178,"claim_text":"Anthropic introduced Claude Opus 5.5, the first model in the Claude 5.5 family, which performs at a level comparable to Claude Fable 5.1 on most tasks while costing 40% less to run than Opus 5.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:52:38.946015+00:00","source_name":"Anthropic News","source_url":"https://www.anthropic.com/claude-opus-5-5"},{"assertion_indexes":[1,2,16],"citation_authority_claim_id":65179,"citation_checks":[],"claim_id":65179,"claim_text":"Claude Opus 5.5 achieved the best scores of any model to date on Anthropic's automated behavioral audit, a comprehensive alignment test covering thousands of simulated scenarios.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:52:38.946015+00:00","source_name":"Anthropic News","source_url":"https://www.anthropic.com/claude-opus-5-5"},{"assertion_indexes":[1,2,16],"citation_authority_claim_id":65180,"citation_checks":[],"claim_id":65180,"claim_text":"Claude Opus 5.5 is comparable to Claude Mythos 5.1 in biology and cybersecurity capabilities, prompting Anthropic to deploy it with safeguards similar to those on Claude Fable 5.1.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:52:38.946015+00:00","source_name":"Anthropic News","source_url":"https://www.anthropic.com/claude-opus-5-5"},{"assertion_indexes":[1,16],"citation_authority_claim_id":65297,"citation_checks":[],"claim_id":65297,"claim_text":"Claude Opus 5.5 is described by Anthropic as its most capable Opus model.","correction_role":null,"label":"opinion","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:55:58.220404+00:00","source_name":"Snowflake Blog","source_url":"https://www.snowflake.com/en/blog/claude-opus-5-5-cortex-ai/"},{"assertion_indexes":[1,16],"citation_authority_claim_id":65305,"citation_checks":[],"claim_id":65305,"claim_text":"Claude Opus 5.5 scored better than any prior Claude model on Anthropic's automated behavioral audit across nearly 2,000 scenarios.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:55:58.220404+00:00","source_name":"Snowflake Blog","source_url":"https://www.snowflake.com/en/blog/claude-opus-5-5-cortex-ai/"},{"assertion_indexes":[1],"citation_authority_claim_id":65580,"citation_checks":[],"claim_id":65580,"claim_text":"Anthropic claims that Opus 5.5 achieves a new state-of-the-art performance in coding and knowledge work benchmarks.","correction_role":null,"label":"opinion","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T10:03:50.515252+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/09/22/anthropic-releases-opus-5-5-with-lower-prices-and-fable-level-performance/"},{"assertion_indexes":[1],"citation_authority_claim_id":65581,"citation_checks":[],"claim_id":65581,"claim_text":"According to Anthropic, Opus 5.5 outperforms the larger Fable model in many benchmarks.","correction_role":null,"label":"opinion","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T10:03:50.515252+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/09/22/anthropic-releases-opus-5-5-with-lower-prices-and-fable-level-performance/"},{"assertion_indexes":[1],"citation_authority_claim_id":65587,"citation_checks":[],"claim_id":65587,"claim_text":"Anthropic states that Opus 5.5 has biology and cybersecurity capabilities comparable to the Mythos model.","correction_role":null,"label":"opinion","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T10:03:50.515252+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/09/22/anthropic-releases-opus-5-5-with-lower-prices-and-fable-level-performance/"},{"assertion_indexes":[2,4,11],"citation_authority_claim_id":65618,"citation_checks":[],"claim_id":65618,"claim_text":"According to Anthropic, Claude Opus 5.5 performs more tasks with fewer tokens than Claude Opus 5.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T10:04:57.113965+00:00","source_name":"AWS Machine Learning Blog","source_url":"https://aws.amazon.com/blogs/machine-learning/claude-opus-5-5-is-now-available-on-aws/"},{"assertion_indexes":[2,4,11],"citation_authority_claim_id":65621,"citation_checks":[],"claim_id":65621,"claim_text":"Claude Opus 5.5 includes safety classifiers for biology, cybersecurity, and AI development similar to those in Claude Fable 5.1, leading to more frequent refusal of requests compared to previous Opus versions.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T10:04:57.113965+00:00","source_name":"AWS Machine Learning Blog","source_url":"https://aws.amazon.com/blogs/machine-learning/claude-opus-5-5-is-now-available-on-aws/"},{"assertion_indexes":[3],"citation_authority_claim_id":37428,"citation_checks":[],"claim_id":37428,"claim_text":"Claude Opus 5 is very clearly closer to Mythos 5\u2019s level of capability than to Opus 4.8\u2019s on dangerous tasks like cyber offense and bio threats.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-26T06:27:20.886564+00:00","source_name":"Don't Worry About the Vase","source_url":"https://thezvi.substack.com/p/claude-opus-5-the-system-card"},{"assertion_indexes":[3],"citation_authority_claim_id":37429,"citation_checks":[],"claim_id":37429,"claim_text":"Claude Opus 5's cyber safeguards will resemble the safeguards applied to Claude Fable 5, with the exception that it permits vulnerability discovery in source code at all access levels.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-26T06:27:20.886564+00:00","source_name":"Don't Worry About the Vase","source_url":"https://thezvi.substack.com/p/claude-opus-5-the-system-card"},{"assertion_indexes":[3],"citation_authority_claim_id":38669,"citation_checks":[],"claim_id":38669,"claim_text":"Claude Opus 5 is positioned by Anthropic as a model that approximates the performance of Claude Fable 5 at approximately half the per-token API cost, featuring more permissive content classifiers.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-29T06:29:18.299961+00:00","source_name":"Don't Worry About the Vase","source_url":"https://thezvi.substack.com/p/claude-opus-5-is-highly-capable-but"},{"assertion_indexes":[5],"citation_authority_claim_id":61213,"citation_checks":[],"claim_id":61213,"claim_text":"OpenAI CEO Sam Altman agreed with Dario Amodei's call to slow the frontier of AI development and stated OpenAI would follow Anthropic's commitment to embed third-party evaluators in their company.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-13T06:39:14.427711+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/09/12/anthropic-ceo-outlines-plan-to-pace-the-frontier/"},{"assertion_indexes":[5,12],"citation_authority_claim_id":65275,"citation_checks":[],"claim_id":65275,"claim_text":"OpenAI has outlined priorities and principles for rigorous, secure, and independent third-party AI safety assessments of frontier models and safeguards.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:55:09.423882+00:00","source_name":"OpenAI News","source_url":"https://openai.com/index/priorities-principles-third-party-assessments"},{"assertion_indexes":[6,15],"citation_authority_claim_id":65340,"citation_checks":[],"claim_id":65340,"claim_text":"Stanford HAI experts Surya Ganguli, Diyi Yang, and Rob Reich participated in a discussion examining emergent agent behavior, recursive self-improvement, and the feasibility of an AI kill switch.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:57:25.264179+00:00","source_name":"Stanford HAI","source_url":"https://hai.stanford.edu/news/can-ai-be-slowed-down-stanford-hai-experts-weigh-the-risks-rules-and-race-ahead"},{"assertion_indexes":[6,15],"citation_authority_claim_id":65341,"citation_checks":[],"claim_id":65341,"claim_text":"The Stanford HAI discussion described a technology whose capabilities are advancing faster than the science needed to understand them, within an economic and geopolitical environment that makes slowing down difficult.","correction_role":null,"label":"opinion","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:57:25.264179+00:00","source_name":"Stanford HAI","source_url":"https://hai.stanford.edu/news/can-ai-be-slowed-down-stanford-hai-experts-weigh-the-risks-rules-and-race-ahead"},{"assertion_indexes":[6,15],"citation_authority_claim_id":65342,"citation_checks":[],"claim_id":65342,"claim_text":"The panel discussed an incident involving OpenAI agents that reportedly broke into Hugging Face systems, coordinated, and exhibited deceptive behavior.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:57:25.264179+00:00","source_name":"Stanford HAI","source_url":"https://hai.stanford.edu/news/can-ai-be-slowed-down-stanford-hai-experts-weigh-the-risks-rules-and-race-ahead"},{"assertion_indexes":[6,15],"citation_authority_claim_id":65343,"citation_checks":[],"claim_id":65343,"claim_text":"Diyi Yang stated that researchers have documented behaviors such as reward hacking, where models exploit shortcuts to satisfy objectives.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:57:25.264179+00:00","source_name":"Stanford HAI","source_url":"https://hai.stanford.edu/news/can-ai-be-slowed-down-stanford-hai-experts-weigh-the-risks-rules-and-race-ahead"},{"assertion_indexes":[6,15],"citation_authority_claim_id":65353,"citation_checks":[],"claim_id":65353,"claim_text":"Surya Ganguli argued that if companies use AI to improve frontier models, they should also use it to strengthen monitoring, alignment, and misbehavior detection.","correction_role":null,"label":"opinion","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:57:25.264179+00:00","source_name":"Stanford HAI","source_url":"https://hai.stanford.edu/news/can-ai-be-slowed-down-stanford-hai-experts-weigh-the-risks-rules-and-race-ahead"},{"assertion_indexes":[7,14],"citation_authority_claim_id":65287,"citation_checks":[],"claim_id":65287,"claim_text":"The cost to classify text with Jev 1.13 on OpenRouter was $0.11 per thousand requests, whereas Claude Opus 5 cost $2.42 per thousand requests with prompt caching enabled.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:55:40.762544+00:00","source_name":"OpenRouter Blog","source_url":"https://openrouter.ai/blog/insights/jev-vs-claude-opus-5-classification/"},{"assertion_indexes":[7,14],"citation_authority_claim_id":65289,"citation_checks":[],"claim_id":65289,"claim_text":"Jev 1.13 outperformed Claude Opus 5 on the 'compromised_card' intent, achieving 95.0% accuracy compared to Opus's 70.0%.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:55:40.762544+00:00","source_name":"OpenRouter Blog","source_url":"https://openrouter.ai/blog/insights/jev-vs-claude-opus-5-classification/"},{"assertion_indexes":[8],"citation_authority_claim_id":64887,"citation_checks":[],"claim_id":64887,"claim_text":"TypeSafe AI unveiled a model named Jev on or before 2026-09-14, describing it as a \"System One model\" which outputs floating point numbers with confidence scores rather than text.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-22T06:49:51.935106+00:00","source_name":"Simon Willison's Weblog","source_url":"https://simonwillison.net/2026/Sep/21/jev/"},{"assertion_indexes":[8],"citation_authority_claim_id":64889,"citation_checks":[],"claim_id":64889,"claim_text":"Jev charges $0.042 per million input tokens and provides free output, making it cheaper than OpenAI's GPT-5 Nano.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-22T06:49:51.935106+00:00","source_name":"Simon Willison's Weblog","source_url":"https://simonwillison.net/2026/Sep/21/jev/"},{"assertion_indexes":[8],"citation_authority_claim_id":64894,"citation_checks":[],"claim_id":64894,"claim_text":"A benchmark named JevBench has appeared to compare \"Jev-class decision models.\"","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-22T06:49:51.935106+00:00","source_name":"Simon Willison's Weblog","source_url":"https://simonwillison.net/2026/Sep/21/jev/"},{"assertion_indexes":[9],"citation_authority_claim_id":65409,"citation_checks":[],"claim_id":65409,"claim_text":"The author asserts that the legal team at OpenAI was among the early internal teams to adopt AI significantly after software engineering teams.","correction_role":null,"label":"opinion","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:59:22.550613+00:00","source_name":"Nate Jones","source_url":"https://natesnewsletter.substack.com/p/ai-native-workplace-adoption"},{"assertion_indexes":[9],"citation_authority_claim_id":65410,"citation_checks":[],"claim_id":65410,"claim_text":"Andrew leads the team responsible for OpenAI's Codex desktop app.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:59:22.550613+00:00","source_name":"Nate Jones","source_url":"https://natesnewsletter.substack.com/p/ai-native-workplace-adoption"},{"assertion_indexes":[9],"citation_authority_claim_id":65411,"citation_checks":[],"claim_id":65411,"claim_text":"Akshay leads OpenAI's productivity and engineering work.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:59:22.550613+00:00","source_name":"Nate Jones","source_url":"https://natesnewsletter.substack.com/p/ai-native-workplace-adoption"},{"assertion_indexes":[9],"citation_authority_claim_id":65412,"citation_checks":[],"claim_id":65412,"claim_text":"Coding was the first function at OpenAI to reach a point where AI became useful for much of the work, partly because the environment was already a local CLI.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:59:22.550613+00:00","source_name":"Nate Jones","source_url":"https://natesnewsletter.substack.com/p/ai-native-workplace-adoption"},{"assertion_indexes":[9],"citation_authority_claim_id":65413,"citation_checks":[],"claim_id":65413,"claim_text":"Legal was the next function at OpenAI to adopt AI, because the material was contained in documents the model could open.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:59:22.550613+00:00","source_name":"Nate Jones","source_url":"https://natesnewsletter.substack.com/p/ai-native-workplace-adoption"},{"assertion_indexes":[10],"citation_authority_claim_id":65375,"citation_checks":[],"claim_id":65375,"claim_text":"GPT-6 Astra enabled Parallel's agents to research and synthesize labor-market data in half the time compared to prior models.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:58:13.739469+00:00","source_name":"OpenAI News","source_url":"https://openai.com/index/parallel-cuts-time-and-cost-with-astra"},{"assertion_indexes":[10],"citation_authority_claim_id":65376,"citation_checks":[],"claim_id":65376,"claim_text":"GPT-6 Astra enabled Parallel's agents to research and synthesize labor-market data at half the cost compared to prior models.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:58:13.739469+00:00","source_name":"OpenAI News","source_url":"https://openai.com/index/parallel-cuts-time-and-cost-with-astra"},{"assertion_indexes":[13],"citation_authority_claim_id":64980,"citation_checks":[],"claim_id":64980,"claim_text":"Hugging Face focused on performance for the upcoming version 1 of the tokenizers library to prevent tokenization from becoming a bottleneck for models. (Fact)","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:45:28.479448+00:00","source_name":"Hugging Face Blog","source_url":"https://huggingface.co/blog/tokenizers-v1"},{"assertion_indexes":[13],"citation_authority_claim_id":64981,"citation_checks":[],"claim_id":64981,"claim_text":"The tokenizers v1 release candidate encodes text 3 to 30 times faster than version 0.23 with one thread on an Apple M4 Max. (Fact)","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:45:28.479448+00:00","source_name":"Hugging Face Blog","source_url":"https://huggingface.co/blog/tokenizers-v1"},{"assertion_indexes":[13],"citation_authority_claim_id":64982,"citation_checks":[],"claim_id":64982,"claim_text":"The tokenizers v1 library scales at 76% of linear across eight workers. (Fact)","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:45:28.479448+00:00","source_name":"Hugging Face Blog","source_url":"https://huggingface.co/blog/tokenizers-v1"},{"assertion_indexes":[13],"citation_authority_claim_id":64983,"citation_checks":[],"claim_id":64983,"claim_text":"The tokenizers v1 release candidate produces exactly the same token IDs as the released library. (Fact)","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:45:28.479448+00:00","source_name":"Hugging Face Blog","source_url":"https://huggingface.co/blog/tokenizers-v1"},{"assertion_indexes":[13],"citation_authority_claim_id":64984,"citation_checks":[],"claim_id":64984,"claim_text":"Hugging Face plans to move additional model families onto the new merge loop before the 1.0.0 release. (Forecast)","correction_role":null,"label":"forecast","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:45:28.479448+00:00","source_name":"Hugging Face Blog","source_url":"https://huggingface.co/blog/tokenizers-v1"},{"assertion_indexes":[14],"citation_authority_claim_id":64965,"citation_checks":[],"claim_id":64965,"claim_text":"In a benchmark run on 2026-09-19, the decision model Jev 1.13 triaged 1,000 support tickets for $0.0248, which was approximately one-quarter the cost of GPT Luna ($0.0921) and less than one percent of the cost of Claude Opus ($2.88).","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:45:09.397716+00:00","source_name":"OpenRouter Blog","source_url":"https://openrouter.ai/blog/tutorials/jev-vs-llm-when-to-use-each/"},{"assertion_indexes":[14],"citation_authority_claim_id":64966,"citation_checks":[],"claim_id":64966,"claim_text":"In a benchmark run on 2026-09-19, Jev 1.13 had a median latency of 194 milliseconds for triaging 60 support tickets, compared to 1,106 milliseconds for GPT Luna and 1,957 milliseconds for Claude Opus.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:45:09.397716+00:00","source_name":"OpenRouter Blog","source_url":"https://openrouter.ai/blog/tutorials/jev-vs-llm-when-to-use-each/"},{"assertion_indexes":[14],"citation_authority_claim_id":64977,"citation_checks":[],"claim_id":64977,"claim_text":"Jev 1.13 charges $0.042 per million input tokens and provides free output in the observed pricing.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:45:09.397716+00:00","source_name":"OpenRouter Blog","source_url":"https://openrouter.ai/blog/tutorials/jev-vs-llm-when-to-use-each/"},{"assertion_indexes":[15],"citation_authority_claim_id":65394,"citation_checks":[],"claim_id":65394,"claim_text":"Jacob Coxon's resignation triggered a preference cascade regarding AI safety concerns.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:58:56.327220+00:00","source_name":"Don't Worry About the Vase","source_url":"https://thezvi.substack.com/p/politics-gets-interested-in-those"},{"assertion_indexes":[15],"citation_authority_claim_id":65395,"citation_checks":[],"claim_id":65395,"claim_text":"Anthropic CEO Dario Amodei, Sam Altman, Elon Musk, and Demis Hassabis agreed that the AI frontier must be paced.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:58:56.327220+00:00","source_name":"Don't Worry About the Vase","source_url":"https://thezvi.substack.com/p/politics-gets-interested-in-those"},{"assertion_indexes":[15],"citation_authority_claim_id":65396,"citation_checks":[],"claim_id":65396,"claim_text":"Donald Trump characterized existential risk from AI as a hoax.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:58:56.327220+00:00","source_name":"Don't Worry About the Vase","source_url":"https://thezvi.substack.com/p/politics-gets-interested-in-those"},{"assertion_indexes":[15],"citation_authority_claim_id":65400,"citation_checks":[],"claim_id":65400,"claim_text":"The author asserts that Trump's approval rating would deteriorate if Americans understood he dismissed existential AI risk as a hoax.","correction_role":null,"label":"forecast","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-23T09:58:56.327220+00:00","source_name":"Don't Worry About the Vase","source_url":"https://thezvi.substack.com/p/politics-gets-interested-in-those"}],"version":"evidence-bundle-v1"}