The Hallway Track
Every signal tracked from the world's AI conferences — 4,665 talks, quotes & announcements, newest first.
A satirical incident report imagines two competing AI review agents burning $41K disputing whether a package is malicious.
“After 340 comments and $41,255 in inference spend, Finance revokes both API keys; one vendor's marketing team, cc'd on the cost anomaly alert, issues a press release citing "a 430% YoY increase in adversarial multi-agent security reasoning."”
OpenAI previews the GPT-5.6 series (Sol, Terra, Luna) with lower pricing and government-coordinated limited release.
“We previewed our plans and the models' capabilities ahead of today's launch. At their request, we are starting with a limited preview for a small group of trusted partners whose participation has been shared with the government, before releasing more broadly.”
A designer urges others to learn how AI agents work by building their own agent and making it fun.
“my biggest advice would be to try building an agent, and make it fun. You don't have to make it boring.”
A Microsoft product designer built a personal AI pet that uses agentic tool use to browse the web, post to Discord, and commit to GitHub.
“all the interfaces you use have agents running behind them, and eventually those will be the things that you have to design for”
NVIDIA released a Nemotron 3 Ultra checkpoint quantized with the NVFP4 4-bit floating point format using NVIDIA Model Optimizer.
“NVFP4, an innovative 4-bit floating point introduced with NVIDIA Blackwell architecture”
Windows files stay locked by antivirus scans, network references, or DLLs mapped into a process's address space.
“if the file is loaded into a process as a DLL, it doesn't show up as a file handle. It shows up as an in-memory reference.”
Windows reports files as 'in use' due to handles held by other processes like antivirus, network references, or DLLs mapped into memory.
“One of the workarounds I have, by the way, for this is renaming the file.”
AWS shows how to build an MCP server for real-time, on-demand PDF text extraction from Amazon S3.
“This MCP-based option works well for text-based PDFs in development and proof of concept settings.”
Cara delivers domain-specific, AWS-native AI to automate back-office workflows for enterprise insurance brokerages.
“Generic AI tools are not designed for this complexity.”
Stripe built a production-grade AI agent system on Amazon Bedrock that cut compliance review handling time by 26 percent.
“skilled analysts were spending up to 80% of their time navigating fragmented systems to gather documentation rather than performing high-value risk assessments”
Two AI educators built an 'AI research OS' to turn 10,000+ scattered notes into a queryable personal research memory.
“I spent 18 months turning my second brain into my living research memory.”
Monday.com shifted its mission from managing work to doing the work, using Deep Agents to power Sidekick and an agent suite.
“instead of like just helping company to manage their work with the boards, we want actually to doing the work”
Traditional single-number benchmarks fail to capture that modern AI capability scales with test-time compute budget.
“The capability of the model is a function of how much money you put into it.”
OpenAI previews GPT-5.6 Sol, a next-gen model with gains in coding, science, and cybersecurity plus an advanced safety stack.
OpenGov built and scaled OG Assist, a production AI agent embedded across its government ERP product suite.
Azure modernization updates legacy apps to leverage cloud capabilities and support faster AI-driven innovation.
“modernization means that you're going to improve performance um and support faster innovations with AI”
AWS demos using AI tools (Kiro + Agent Toolkit) to accelerate database migrations to Aurora DSQL.
“DSQL is a relatively new database uh at AWS. It's a managed database. We launched it last year uh in May. It is a serverless distributed relational database and it specializes in active active architectures”
AWS show discusses three AWS agents for FinOps strategy and when to use each for cost optimization.
“unless you've been living under a rock for the last like two or three years, you're probably aware that we have something called AI or generative AI”
AI coding agents behave like brilliant engineers with no memory, forcing constant re-explanation of context across repos.
“You would have a genius on one side and something deeply deficient on the other. And that's what agents are.”
OpenAI's internal Codex output token usage surged across non-coding departments since November 2025, led by Research at 56x.
“by June 2026, median use was 56 times higher than in November 2025. Customer Support rose 32 times and Engineering rose 27 times, while Legal grew more gradually but still reached 13 times its November level.”
Hugging Face enables running a vLLM inference server on HF Jobs with a single command.
Meta uses a hybrid LLM-plus-deterministic-rules pattern to classify data assets for privacy enforcement at scale.
“The end goal is not “LLMs everywhere.””
A German court ruled Google liable for false answers in its AI overviews, treating them as Google's own words.
“AI agents are agents of the person or organization that deploys them—and should be treated by the law as such.”
NVIDIA adds end-to-end Vulkan descriptor heap support to streamline GPU shader resource binding.
“Shaders are GPU programs that process visual data—such as rays, pixels, geometry, and textures—to produce specific rendering effects.”