Sources
Where each section looks: Tier 1 sources fetched every run, Tier 2 covered by window-bounded search.
Two tiers. Tier 1: fetch directly every run; they carry primary announcements. If a page renders empty or stale, fall back to an Exa query scoped to that source. Tier 2: cover with at least one Exa query bounded to the window, and fetch an item only once it clears the Inclusion Bar. Treat the list as a coverage checklist, then add queries for anything inside the window that none of these surfaced.
model-releases: Model Releases
New open-weight and closed models (LLM, multimodal, vision, audio, code): provider, name, parameter count if known, key capabilities, benchmark highlights, license.
- Tier 1: OpenAI https://openai.com/news/, Anthropic https://www.anthropic.com/news, Google DeepMind https://deepmind.google/discover/blog/, Meta AI https://ai.meta.com/blog/
- Tier 2: Mistral AI https://mistral.ai/news/, xAI https://x.ai/blog, Cohere https://cohere.com/blog, AI21 Labs https://www.ai21.com/blog, Hugging Face https://huggingface.co/blog, Together AI https://www.together.ai/blog, Stability AI https://stability.ai/news, Chinese labs (DeepSeek, Qwen, Doubao, Kimi, GLM, MiniMax)
tools-products: AI Tools & Products
Developer tools, APIs, platforms, and consumer products. Technical differentiation over marketing claims.
- Tier 1: OpenAI https://openai.com/news/, Anthropic https://www.anthropic.com/news, Google DeepMind https://deepmind.google/discover/blog/
- Tier 2: Hugging Face https://huggingface.co/blog, Mistral AI https://mistral.ai/news/, Cohere https://cohere.com/blog, TechCrunch AI https://techcrunch.com/category/artificial-intelligence/, VentureBeat AI https://venturebeat.com/category/ai/
research: Research Breakthroughs
Frontier papers and lab research: novel architectures, training techniques, alignment findings, capability evaluations, scaling results. Include arXiv IDs or DOIs.
- Tier 1: Hugging Face Daily Papers https://huggingface.co/papers, arXiv https://arxiv.org/list/cs.AI/recent https://arxiv.org/list/cs.LG/recent https://arxiv.org/list/cs.CL/recent, OpenAI https://openai.com/research/, Anthropic https://www.anthropic.com/research, Google DeepMind https://deepmind.google/discover/blog/
- Tier 2: Meta AI https://ai.meta.com/blog/, EleutherAI https://blog.eleuther.ai/, Papers With Code https://paperswithcode.com/
workflows: Workflows & Techniques
Prompting strategies, inference optimizations, fine-tuning methods, agent frameworks, RAG improvements, production deployment patterns.
- Tier 1: Hugging Face https://huggingface.co/blog, Hugging Face Daily Papers https://huggingface.co/papers
- Tier 2: Anthropic engineering and OpenAI developer posts, Together AI https://www.together.ai/blog, framework release notes (vLLM, SGLang, LangChain, LlamaIndex, DSPy)
industry: Industry & Ecosystem
Significant funding, acquisitions, regulation, open-source milestones, notable benchmark results.
- Tier 2: TechCrunch AI https://techcrunch.com/category/artificial-intelligence/, VentureBeat AI https://venturebeat.com/category/ai/, The Information https://www.theinformation.com/, Crunchbase News https://news.crunchbase.com/, Reuters and Bloomberg technology coverage, government and regulator newsrooms
startups: AI Startups
Early- to growth-stage AI startups that launched, shipped a significant product, raised, or hit a milestone: name, stage or amount, core technology, why it matters. Genuine technical differentiation over AI wrappers. Not startups already in Industry & Ecosystem; healthcare startups belong in Healthcare AI.
- Tier 2: TechCrunch AI https://techcrunch.com/category/artificial-intelligence/, Crunchbase News https://news.crunchbase.com/, Y Combinator AI companies https://www.ycombinator.com/companies?batch=&tags=Artificial+Intelligence, a16z AI https://a16z.com/ai/, VentureBeat AI https://venturebeat.com/category/ai/
healthcare: Healthcare AI
Clinical tools and decision support, healthcare workflows and deployments, peer-reviewed clinical and health-services research, healthcare AI startups, and regulation (clearances and authorizations, reimbursement, interoperability rules, model governance). Items whose primary significance is general stay in their own section.
- Clinical research: NEJM AI https://ai.nejm.org/, JAMA Network Open https://jamanetwork.com/journals/jamanetworkopen
- Trade press: Healthcare IT News https://www.healthcareitnews.com/, HealthITAnalytics https://healthitanalytics.com/, STAT Health Tech https://www.statnews.com/category/health-tech/, Healthcare Dive https://www.healthcaredive.com/
- Regulation: FDA AI-enabled devices https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-and-machine-learning-aiml-enabled-medical-devices, CMS Newsroom https://www.cms.gov/newsroom, ONC https://www.healthit.gov/buzz-blog/
- Health IT platforms: track the dominant EHR vendor and the leading clinical-data and interoperability vendor by role and capability (newsrooms, app marketplaces, FHIR/HL7 API documentation, clinical partnerships). Name a vendor only in a verified, sourced item.