{"assertions":[{"assertion_index":0,"assertion_text":"Latent Space Nate Jones Anthropic News That target is smaller than August's tens of thousands, and I cannot tell from the evidence whether the academy is a first installment or a narrower plan.","citation_check":null,"claim_ids":[1165,53485,71396,71397,71398],"link_only_urls":[]},{"assertion_index":1,"assertion_text":"Nate Jones The IPO numbers show why that would matter.","citation_check":null,"claim_ids":[10093],"link_only_urls":[]},{"assertion_index":2,"assertion_text":"Big Technology TechCrunch AI TechCrunch AI So my argument does not rest on revenue being small.","citation_check":null,"claim_ids":[47998,47999,69228,71604,71605],"link_only_urls":[]},{"assertion_index":3,"assertion_text":"TechCrunch AI OpenAI News So the gap between the two is one of scale and channel, not of kind.","citation_check":null,"claim_ids":[46478,71421,71422],"link_only_urls":[]},{"assertion_index":4,"assertion_text":"Ollama Releases Ollama Releases llama.cpp Releases If you are an engineer deciding whether to build on this format, the risk has shifted from a one-vendor dead end to ordinary early-adopter risk.","citation_check":null,"claim_ids":[69169,69170,69173,71625,71626,71627,71660,71661,71662],"link_only_urls":[]},{"assertion_index":5,"assertion_text":"TechCrunch AI I still would not standardize on it, but I would run a pilot.","citation_check":null,"claim_ids":[70544],"link_only_urls":[]},{"assertion_index":6,"assertion_text":"Allen Institute for AI Those averages work out to about 3.5 times faster.","citation_check":null,"claim_ids":[71447,71448,71449],"link_only_urls":[]},{"assertion_index":7,"assertion_text":"- Anthropic has set the end of 2027 for 10,000 trained engineers, and the first published cohort count will show whether the pace is real. Anthropic News","citation_check":null,"claim_ids":[71396,71397,71398],"link_only_urls":[]},{"assertion_index":8,"assertion_text":"- Palisade Research has released interviews with current and former lab employees about existential risk, and more exits like Robert O'Callahan's from Google would deepen the pattern. Don't Worry About the Vase","citation_check":null,"claim_ids":[71482,71483,71484],"link_only_urls":[]},{"assertion_index":9,"assertion_text":"- Google says Gboard now trains its next-word models under a system with externally verifiable privacy guarantees; the test is whether outside parties publish that verification. Google Research Blog","citation_check":null,"claim_ids":[71516,71517,71518],"link_only_urls":[]},{"assertion_index":10,"assertion_text":"- llama.cpp supports Clef only in text mode so far, and image support there would make image-capable Clef workloads runnable without Ollama. llama.cpp Releases","citation_check":null,"claim_ids":[71660,71661,71662],"link_only_urls":[]},{"assertion_index":11,"assertion_text":"- OpenAI's new guide tells startups how to pick GPT-6 models and tune reasoning effort, and the next customer case studies will show whether buyers follow it. OpenAI News","citation_check":null,"claim_ids":[71569],"link_only_urls":[]},{"assertion_index":12,"assertion_text":"OpenAI News AstaBrief's speed numbers are Ai2's own averages.","citation_check":null,"claim_ids":[71421,71422],"link_only_urls":[]},{"assertion_index":13,"assertion_text":"Allen Institute for AI","citation_check":null,"claim_ids":[71447,71448,71449],"link_only_urls":[]}],"brief_date":"2026-10-03","bundle_id":"evidence-bundle-v2-570a5921f9a09dcae316229b0ce01372f3b29b1a894c4f8fc32b54bc4f5c1ffd","claims":[{"assertion_indexes":[0],"citation_authority_claim_id":1165,"citation_checks":[],"claim_id":1165,"claim_text":"Anthropic launched a Forward Deployed Engineer track mirroring similar initiatives from OpenAI and itself.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-06-08T13:08:35.045000+00:00","source_name":"Latent Space","source_url":"https://www.latent.space/p/ainews-founders-and-forward-deployed"},{"assertion_indexes":[0],"citation_authority_claim_id":53485,"citation_checks":[],"claim_id":53485,"claim_text":"Anthropic stated it would train tens of thousands of engineers to install AI in banks, airlines, and insurers.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-30T11:42:23.314313+00:00","source_name":"Nate Jones","source_url":"https://www.youtube.com/watch?v=0bLI31EFDDs"},{"assertion_indexes":[0,7],"citation_authority_claim_id":71396,"citation_checks":[],"claim_id":71396,"claim_text":"Anthropic announced the launch of the Claude Frontier Academy, a program designed to train 10,000 Frontier Deployed Engineers.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:34:55.391918+00:00","source_name":"Anthropic News","source_url":"https://www.anthropic.com/news/claude-frontier-academy"},{"assertion_indexes":[0,7],"citation_authority_claim_id":71397,"citation_checks":[],"claim_id":71397,"claim_text":"Anthropic committed $100 million to fund the training of 10,000 Frontier Deployed Engineers.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:34:55.391918+00:00","source_name":"Anthropic News","source_url":"https://www.anthropic.com/news/claude-frontier-academy"},{"assertion_indexes":[0,7],"citation_authority_claim_id":71398,"citation_checks":[],"claim_id":71398,"claim_text":"Anthropic aims to complete the training of 10,000 Frontier Deployed Engineers by the end of 2027.","correction_role":null,"corroborates_eligible":true,"label":"forecast","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:34:55.391918+00:00","source_name":"Anthropic News","source_url":"https://www.anthropic.com/news/claude-frontier-academy"},{"assertion_indexes":[1],"citation_authority_claim_id":10093,"citation_checks":[],"claim_id":10093,"claim_text":"Anthropic is focusing on enterprise distribution and forward-deployed engineers to build resilient revenue streams.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-09T06:20:42.647876+00:00","source_name":"Nate Jones","source_url":"https://www.youtube.com/watch?v=oOpgmS88pLw"},{"assertion_indexes":[2],"citation_authority_claim_id":47998,"citation_checks":[],"claim_id":47998,"claim_text":"Anthropic's annualized revenue run rate was $47 billion in May 2026.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-18T07:35:18.103873+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/08/17/anthropics-annualized-revenue-surges-to-65b/"},{"assertion_indexes":[2],"citation_authority_claim_id":47999,"citation_checks":[],"claim_id":47999,"claim_text":"Anthropic's annualized revenue run rate was $9 billion at the end of 2025.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-18T07:35:18.103873+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/08/17/anthropics-annualized-revenue-surges-to-65b/"},{"assertion_indexes":[2],"citation_authority_claim_id":69228,"citation_checks":[],"claim_id":69228,"claim_text":"Anthropic's prospectus reveals plans to spend $518 billion on cloud, computing, and infrastructure in the coming years.","correction_role":null,"corroborates_eligible":true,"label":"forecast","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-29T07:17:43.879079+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/09/28/anthropics-prospectus-details-losses-growth-and-yes-a-warning-that-its-ai-could-end-humanity/"},{"assertion_indexes":[2],"citation_authority_claim_id":71604,"citation_checks":[],"claim_id":71604,"claim_text":"Anthropic's revenue grew from $400 million in 2024 to $4.6 billion in 2025.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:44:41.784343+00:00","source_name":"Big Technology","source_url":"https://www.bigtechnology.com/p/anthropics-ipo-and-our-collective"},{"assertion_indexes":[2],"citation_authority_claim_id":71605,"citation_checks":[],"claim_id":71605,"claim_text":"Anthropic spent $7.3 billion on infrastructure in 2025 and expects to spend $518 billion on infrastructure over the next decade, with approximately 80% of that amount being non-cancellable.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:44:41.784343+00:00","source_name":"Big Technology","source_url":"https://www.bigtechnology.com/p/anthropics-ipo-and-our-collective"},{"assertion_indexes":[3],"citation_authority_claim_id":46478,"citation_checks":[],"claim_id":46478,"claim_text":"IBM will integrate OpenAI\u2019s latest models, including GPT-5.6, Codex, and ChatGPT Work, into IBM Consulting Advantage.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-14T06:26:48.917781+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/08/13/ibm-partners-with-openai-to-bolster-enterprise-ai-push/"},{"assertion_indexes":[3,12],"citation_authority_claim_id":71421,"citation_checks":[],"claim_id":71421,"claim_text":"Chatham Financial uses OpenAI's Codex and GPT-5.6 models to build technology and redesign workflows.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:35:38.411969+00:00","source_name":"OpenAI News","source_url":"https://openai.com/index/chatham-financial"},{"assertion_indexes":[3,12],"citation_authority_claim_id":71422,"citation_checks":[],"claim_id":71422,"claim_text":"Chatham Financial reduced its trade validation time from 30 minutes to under 4 minutes.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:35:38.411969+00:00","source_name":"OpenAI News","source_url":"https://openai.com/index/chatham-financial"},{"assertion_indexes":[4],"citation_authority_claim_id":69169,"citation_checks":[],"claim_id":69169,"claim_text":"Ollama version 0.35.0 added support for decision models via the /v1/systemone endpoint, which is based on TypeSafe's Jev API.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-29T07:14:44.219913+00:00","source_name":"Ollama Releases","source_url":"https://github.com/ollama/ollama/releases/tag/v0.35.0"},{"assertion_indexes":[4],"citation_authority_claim_id":69170,"citation_checks":[],"claim_id":69170,"claim_text":"The decision model feature in Ollama 0.35.0 returns choices, probabilities, and scores rather than text.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-29T07:14:44.219913+00:00","source_name":"Ollama Releases","source_url":"https://github.com/ollama/ollama/releases/tag/v0.35.0"},{"assertion_indexes":[4],"citation_authority_claim_id":69173,"citation_checks":[],"claim_id":69173,"claim_text":"The /v1/systemone API in Ollama 0.35.0 supports three question types: choice, noul, and score.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-09-29T07:14:44.219913+00:00","source_name":"Ollama Releases","source_url":"https://github.com/ollama/ollama/releases/tag/v0.35.0"},{"assertion_indexes":[4],"citation_authority_claim_id":71625,"citation_checks":[],"claim_id":71625,"claim_text":"Ollama version 0.35.1 adds support for Cloudflare's Clef and Clef Flash decision models via the /v1/systemone endpoint.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:45:09.845994+00:00","source_name":"Ollama Releases","source_url":"https://github.com/ollama/ollama/releases/tag/v0.35.1"},{"assertion_indexes":[4],"citation_authority_claim_id":71626,"citation_checks":[],"claim_id":71626,"claim_text":"The Clef (27B) and Clef Flash (9B) models released by Cloudflare are multimodal, allowing image inputs alongside text state.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:45:09.845994+00:00","source_name":"Ollama Releases","source_url":"https://github.com/ollama/ollama/releases/tag/v0.35.1"},{"assertion_indexes":[4],"citation_authority_claim_id":71627,"citation_checks":[],"claim_id":71627,"claim_text":"Ollama version 0.35.1 increases the maximum number of web searches a model can perform per response from three to ten.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:45:09.845994+00:00","source_name":"Ollama Releases","source_url":"https://github.com/ollama/ollama/releases/tag/v0.35.1"},{"assertion_indexes":[4,10],"citation_authority_claim_id":71660,"citation_checks":[],"claim_id":71660,"claim_text":"The llama.cpp project released version b11371.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:46:30.380048+00:00","source_name":"llama.cpp Releases","source_url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11371"},{"assertion_indexes":[4,10],"citation_authority_claim_id":71661,"citation_checks":[],"claim_id":71661,"claim_text":"Version b11371 of llama.cpp adds initialization support for the clef decision model in text-only mode.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:46:30.380048+00:00","source_name":"llama.cpp Releases","source_url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11371"},{"assertion_indexes":[4,10],"citation_authority_claim_id":71662,"citation_checks":[],"claim_id":71662,"claim_text":"Version b11371 of llama.cpp includes updates related to the static graph.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:46:30.380048+00:00","source_name":"llama.cpp Releases","source_url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11371"},{"assertion_indexes":[5],"citation_authority_claim_id":70544,"citation_checks":[],"claim_id":70544,"claim_text":"OpenAI's Decisions API provides functionality similar to Jev, a model released by TypeSafe AI that is designed for software automation.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-01T06:54:08.106149+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/09/30/openais-jev-clone-could-help-the-frontier-lab-stop-its-swarming-agents/"},{"assertion_indexes":[6,13],"citation_authority_claim_id":71447,"citation_checks":[],"claim_id":71447,"claim_text":"Allen Institute for AI (Ai2) developed and open-sourced AstaBrief 8B, a small open-weight language model designed for generating cited scientific reports from research questions and retrieved literature.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:37:46.118942+00:00","source_name":"Allen Institute for AI","source_url":"https://allenai.org/blog/astabrief"},{"assertion_indexes":[6,13],"citation_authority_claim_id":71448,"citation_checks":[],"claim_id":71448,"claim_text":"AstaBrief 8B is available in the Asta platform as 'Fast mode' alongside the Claude-powered 'Thinking mode'.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:37:46.118942+00:00","source_name":"Allen Institute for AI","source_url":"https://allenai.org/blog/astabrief"},{"assertion_indexes":[6,13],"citation_authority_claim_id":71449,"citation_checks":[],"claim_id":71449,"claim_text":"The source asserts that AstaBrief achieves a nearly order-of-magnitude reduction in report generation time compared to proprietary models, with Fast mode averaging 51.1 seconds per report versus 178.5 seconds for Thinking mode.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:37:46.118942+00:00","source_name":"Allen Institute for AI","source_url":"https://allenai.org/blog/astabrief"},{"assertion_indexes":[8],"citation_authority_claim_id":71482,"citation_checks":[],"claim_id":71482,"claim_text":"Robert O'Callahan resigned from Google because his team was working on a new generation of chips to make AI faster and cheaper, and he believed AI was already progressing too fast.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:38:58.042293+00:00","source_name":"Don't Worry About the Vase","source_url":"https://thezvi.substack.com/p/the-ai-preference-cascade-reaches"},{"assertion_indexes":[8],"citation_authority_claim_id":71483,"citation_checks":[],"claim_id":71483,"claim_text":"Palisade Research released a batch of interviews with current and former AI lab employees regarding their personal views on existential risk.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:38:58.042293+00:00","source_name":"Don't Worry About the Vase","source_url":"https://thezvi.substack.com/p/the-ai-preference-cascade-reaches"},{"assertion_indexes":[8],"citation_authority_claim_id":71484,"citation_checks":[],"claim_id":71484,"claim_text":"Current AI lab employees are discouraged from discussing existential risk, primarily through cultural pressure and sometimes lack of permissions, especially at Anthropic.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:38:58.042293+00:00","source_name":"Don't Worry About the Vase","source_url":"https://thezvi.substack.com/p/the-ai-preference-cascade-reaches"},{"assertion_indexes":[9],"citation_authority_claim_id":71516,"citation_checks":[],"claim_id":71516,"claim_text":"Google Research announced a new Federated Learning system that uses Trusted Execution Environments (TEEs) to provide externally verifiable privacy guarantees while shifting computation to the server.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:40:24.187009+00:00","source_name":"Google Research Blog","source_url":"https://research.google/blog/toward-provably-private-learning-from-federated-data/"},{"assertion_indexes":[9],"citation_authority_claim_id":71517,"citation_checks":[],"claim_id":71517,"claim_text":"The Gboard application has adopted the new TEE-based Federated Learning system.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:40:24.187009+00:00","source_name":"Google Research Blog","source_url":"https://research.google/blog/toward-provably-private-learning-from-federated-data/"},{"assertion_indexes":[9],"citation_authority_claim_id":71518,"citation_checks":[],"claim_id":71518,"claim_text":"Gboard has launched English and Japanese next-word prediction models using the new system, which offer stronger privacy guarantees and improved accuracy compared to previous versions.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:40:24.187009+00:00","source_name":"Google Research Blog","source_url":"https://research.google/blog/toward-provably-private-learning-from-federated-data/"},{"assertion_indexes":[11],"citation_authority_claim_id":71569,"citation_checks":[],"claim_id":71569,"claim_text":"OpenAI has published a guide detailing how startups can select models from the GPT-6 family, tune reasoning effort, improve prompts and skills, coordinate tools, and prepare workflows for production.","correction_role":null,"corroborates_eligible":true,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-10-03T06:42:45.903105+00:00","source_name":"OpenAI News","source_url":"https://openai.com/index/practical-guide-building-gpt-6"}],"version":"evidence-bundle-v2"}