• Databricks hits $188B valuation, extending its run as AI's favorite second act — Databricks announced a new funding round valuing the company at $188 billion, cementing its transformation from a data lakehouse company into a core AI infrastructure player with products like Lakebase (database for AI agents), Unity (AI gateway), and Omnigent (multi-agent meta-harness). 🔗 Graph: Databricks, AI Strategy, Agentic AI 📅 Published: 2026-07-17 📰 https://techcrunch.com/2026/07/17/databricks-hits-188b-valuation-extending-its-run-as-ais-favorite-second-act/ 📌 Key takeaways: • Databricks raised at a $188B valuation on a year-and-a-half fundraising tear, successfully repositioning as an AI company rather than just a SaaS data platform • The company's product portfolio now includes Lakebase (a database built for AI agents), Unity (an AI gateway), and Omnigent (a meta-harness that manages multiple agents) — directly competing in the agentic AI infrastructure space • Databricks published research showing that open-weight models like GLM 5.2 can handle the highest level of coding task difficulty at lower total cost than proprietary models from Anthropic and OpenAI — validating the model-agnostic approach Brett's TritonAI gateway uses • The AI-halo effect on valuations is so strong that even enterprise data companies are leveraging AI positioning to command premium multiples, which matters for vendor evaluation and budget planning at UCSD

• New York bans data center construction for a year, rattling AI industry — New York became the first U.S. state to impose a moratorium on new data center construction, a move that could become a blueprint for the anti-AI movement and signals growing regulatory pressure on AI infrastructure expansion. 🔗 Graph: AI Governance, Enterprise Monitoring, Infrastructure & Migration 📅 Published: 2026-07-14 📰 https://arstechnica.com/tech-policy/2026/07/new-york-is-the-first-state-to-impose-a-data-center-moratorium/ 📌 Key takeaways: • New York signed a one-year moratorium on new data center construction on Tuesday, making it the first state to formally pause AI infrastructure buildout • The moratorium's greatest immediate impact may be the momentum it gives to the anti-AI movement by signaling that states can and will set higher standards for data center construction • This creates precedent for other states to follow, potentially constraining where AI infrastructure can be deployed — relevant to any institution planning on-prem AI expansions • For UCSD's on-prem TritonAI hosting at SDSC, this trend underscores the strategic value of having existing data center capacity and the importance of proactive engagement with regulatory developments

• Thinking Machines Lab Drops Its First Model — Mira Murati's Thinking Machines Lab released Inkling, a 975-billion-parameter open-weight model trained from scratch to understand video, audio, and text, aiming to compete with the best open-weight models from China and proprietary models from OpenAI and Anthropic. 🔗 Graph: OpenAI, AI Strategy, Model Agnosticism 📅 Published: 2026-07-15 📰 https://www.wired.com/story/thinking-machines-lab-releases-its-first-model-inkling/ 📌 Key takeaways: • Inkling is a 975-billion-parameter open-weight model — researchers and startups can download and modify it, unlike closed models from OpenAI and Anthropic • The model was trained from scratch on audio, video, and text (multimodal), and while not the best on benchmarks, it performs well at many tasks including advanced reasoning and coding • Thinking Machines used Inkling to fine-tune and improve itself, and observed that the model's chain-of-thought became more concise over training — dropping grammatical overhead while maintaining reasoning quality • Founded by former OpenAI CTO Mira Murati, John Schulman, and Lilian Weng with the largest seed round in history ($12B valuation), the company advocates for decentralized AI — a signal that the open-weight ecosystem continues to mature alongside proprietary frontier models

• Back to Campus? Higher Education's Great Work Debate — This EDUCAUSE Rising Voices podcast episode explores how remote, hybrid, and in-person work models are evolving in higher education, with a particular focus on how AI is reshaping jobs and elevating the importance of human skills like adaptability and communication. 🔗 Graph: Higher Ed AI, AI Adoption, AI Strategy 📅 Published: 2026-07-16 📰 https://er.educause.edu/podcasts/educause-rising-voices/2026/back-to-campus-higher-educations-great-work-debate 📌 Key takeaways: • The episode examines how higher education institutions are navigating the tension between remote, hybrid, and in-person work models — a debate that directly affects IT staffing and team cohesion • AI is reshaping higher ed jobs, increasing the premium on human skills such as adaptability and communication rather than purely technical competencies • The discussion highlights the difference between inclusion and genuine connection in distributed teams — relevant for Brett's hybrid ITS organization managing both in-person and remote direct reports • EDUCAUSE continues to be the leading voice framing how AI intersects with higher ed workforce strategy, making this useful context for Brett's cabinet-level AI presentations

• Our approach to bioresilience — Google DeepMind and Isomorphic Labs published their joint approach to bioresilience, outlining how AI models can support prevention, detection, and response for biological threats — a significant policy and research development at the intersection of AI safety and public health. 🔗 Graph: Google, AI Security, AI Governance 📅 Published: 2026-07-16 📰 https://deepmind.google/blog/our-approach-to-bioresilience/ 📌 Key takeaways: • Google DeepMind and Isomorphic Labs jointly outlined their bioresilience strategy — using AI models for prevention, detection, and response to biological threats • This represents a major AI lab formally engaging with biosecurity governance, an area of increasing regulatory attention as AI models gain capabilities in protein design and biological prediction • The approach is categorized under "Responsibility & Safety" at DeepMind, signaling that bioresilience is being treated as an AI safety issue, not just a public health one • For AI governance frameworks in higher ed, this sets a precedent for how institutions should think about dual-use AI capabilities — relevant to UCSD's AI compliance and governance priorities

• Fine-tune video and image models at scale with NVIDIA NeMo Automodel and 🤗 Diffusers — NVIDIA and Hugging Face released an integrated fine-tuning library that combines NeMo Automodel's distributed training capabilities with the Diffusers ecosystem, enabling scalable customization of open diffusion models like FLUX.1-dev (12B) and HunyuanVideo (13B). 🔗 Graph: AI Strategy, LiteLLM Enterprise, Model Agnosticism 📅 Published: 2026-07-17 📰 https://huggingface.co/blog/nvidia/scale-diffusers-finetuning-nemo-automodel 📌 Key takeaways: • NeMo Automodel is an open-source PyTorch DTensor-native training library that integrates directly with Hugging Face Diffusers model classes, lowering the barrier to fine-tuning large generative models • The integration ships with ready-to-use fine-tuning recipes for open diffusion models including FLUX.1-dev (12B) and HunyuanVideo (13B), with support for FSDP2, tensor, context, and pipeline parallelisms • Multi-node orchestration via SLURM is available today with Kubernetes support coming — making this viable for enterprise and research cluster deployments like SDSC • The approach uses small code additions (data preprocessing handler + model adapter) rather than full custom training scripts, with YAML-driven workflows — lowering the technical barrier for teams adopting model fine-tuning

💡 Signal: Open-weight models continue to close the gap with proprietary frontier models — Databricks validated GLM 5.2 for highest-difficulty coding tasks at lower cost, Thinking Machines released a 975B-parameter multimodal model, and NVIDIA/Hugging Face shipped tooling that makes fine-tuning at scale more accessible. Meanwhile, New York's data center moratorium signals that infrastructure regulation is becoming a real constraint. For Brett, the open-weight validation and fine-tuning tooling directly support the model-agnostic TritonAI gateway strategy, while the regulatory pressure reinforces the value of UCSD's existing on-prem capacity at SDSC.