The Hallway Track
Every signal tracked from the world's AI conferences — 4,665 talks, quotes & announcements, newest first.
AWS demonstrates pool-model multi-tenancy patterns for isolated, tiered AI agents using Amazon Bedrock AgentCore.
“This solution demonstrates how to use native capabilities of Amazon Bedrock AgentCore to achieve complete tenant isolation using AWS-managed services.”
NVIDIA's DFlash speculative decoding boosts LLM inference performance up to 15x on Blackwell GPUs.
“Speculative decoding helps mitigate this bottleneck by using a lightweight model to draft future tokens”
Microsoft is building AI agents into Visual Studio and VS Code, including skill creation and a multi-harness agents window.
“I'm using the agent's window to build the agent's window.”
TK Elevator built AI-ready data on Databricks Unity Catalog to harmonize 100+ systems across 50 countries for AI agents.
“If humans can't read it, how would agents read it?”
NVIDIA's BioNeMo Agent Toolkit enables building AI scientist agents for life science discovery.
“But science isn’t software engineering. There is no test suite that turns green when a hypothesis is correct; discovery is iterative, uncertain, and grounded in the physical world.”
OpenAI supports shared standards for advanced AI via the Appia Foundation.
Hugging Face showcases CUGA, a lightweight harness with two dozen working examples for building real agentic apps.
Companies lose their fearless engineering culture and stop taking risks as they scale to thousands of employees.
“we would much rather fail in pursuit of the extraordinary than succeed in the ordinary”
Microsoft launches SharePoint Copilot Apps, bringing interactive SPFx-built UX components into the Microsoft 365 Copilot canvas.
“your AI coding agents already know how to build this”
An AWS Tech Tales episode introduces a deep dive into Cloud Intelligence dashboards covering pricing and health.
Transformers.js uses one JavaScript API pattern, driven by task ID contracts, to run 27 ML tasks.
“One API pattern, many machine learning tasks.”
Unity Catalog enforces fine-grained access controls on external engines like open-source Spark, not just Databricks.
“Fine-grained access controls defined once and enforced everywhere, plus external engines that can read and write to UC managed tables.”
AWS security livestream recaps its season's fundamentals using a video-game speedrun metaphor.
“Any percent means you just got to complete it.”
SpaceX's GPU rental deals now annualize to ~$28B/year, roughly twice CoreWeave's revenue.
“In Summary, $2.32B / month, >$10 / hour for Blackwells (which is a very high rate)”
Telcos are adopting agentic AI for autonomous networks but most remain at TM Forum Level 2-3 autonomy.
“automation typically sits in the Level 2–3 band of TM Forum’s autonomous networks levels taxonomy, streamlining execution of predefined solutions in selective network domains”
AWS solutions architects describe building an enterprise voice AI agent for healthcare using Amazon Bedrock AgentCore.
“we recently work with a healthcare customer to help them build out their agentic platform using the voice agent as a use case”
AWS showcases an AI agent on Amazon EMR that troubleshoots and upgrades Apache Spark workloads.
Databricks unveils Omnigen, letting practitioners use any model, harness, or coding agent to build with frontier AI.
“now to get access to any model, any harness, any coding agent to build crazy cool stuff using all of this latest frontier AI capability.”
Databricks unveiled Genie ontology, an auto-learned knowledge graph, as its headline Data + AI Summit 2026 announcement.
“The real secret sauce that we've announced today is Genie ontology. It's an automatically learned, automatically created knowledge graph that does all of that hard work for you, connecting data domains, humans, agents, and all the knowledge insights in your environment.”
Omio uses OpenAI to power conversational travel and become an AI-native company.
Hugging Face ships huggingface_hub weekly using AI agents with a human reviewer in the loop.
Hugging Face is experimenting with the proposed Cross-Origin Storage API in Transformers.js.
Models confuse text style with role, making prompt injection a 'role confusion' problem defenses can't easily fix.
“destyling causes average attack success in our dataset to plunge from 61% to 10%”
Simon Willison used Claude Code to port the Moebius 0.2B inpainting model to run in-browser via WebGPU.
“the harder a problem is the more time you have to get distracted while you wait for them to finish crunching!”