The Hallway Track
Every signal tracked from the world's AI conferences — 4,665 talks, quotes & announcements, newest first.
Microsoft SQL agent skills launched in public preview for AI-assisted database migration and management
“think of it like a repository that has a bunch of custom curated and uh validated skills from the Microsoft SQL team to help you do things like in this case uh assess, remediate, validate, migrate”
EmbeddingGemma 2 enables on-device multimodal semantic search across photos and videos offline
“This all happens directly on my phone without any intermediate transcription, and no external API calls are required.”
Oracle uses ChatGPT and Codex to compress days of specialist work into minutes
AI makes building internal apps easy, but deploying and sharing them securely remains hard
“making things has become cheap, right? But releasing it, making it available, even within the company, hasn't become much easier”
Salesforce demos AIforce with IT help desk Slack bot and multi-framework developer platform
“build, deploy, and govern anywhere”
LangChain's lab team argues agents must continuously learn from run traces, not require manual log review
“observing and turning those observations and agent experiences into lessons for future launches are obviously very different things”
SharePoint Online and OneDrive will retire pre-authenticated tempauth URLs by April 1, 2027.
“If an API returns a URL that previously worked without authentication because it embedded a token in the URL, that URL must now be accessed by presenting a valid Microsoft Entra ID access token at request time.”
Databricks and Replit combine to offer enterprises a governed development-to-deployment platform with Lakebase.
Bending Spoons acquires digital businesses and integrates them into one proprietary technology platform.
“if we can create a machine, a system, that we can run a digital technology business better than almost anyone else, then we are almost guaranteed to be able to effectively accumulate capital over a very long time through acquisitions”
Pollo AI integrates GPT-5.6, GPT-6 Astra, and GPT-Image-2.5 for creative campaign generation
Anthropic launches Claude Haiku 5.5, 75% cheaper than Haiku 4.5, matching GPT-6 Luna pricing
“the cheapest, fastest, and most capable small model we've ever released”
LangChain's 'skills' pattern lazily loads agent context to prevent context window overload
“With skills, information is loaded in layers, not all at once, so context stays small until it's actually useful.”
Larger context windows cause agent hallucination; context engineering is the real fix.
“context engineering is helping us create a better architecture for these agent programs”
Enterprises should reorganize company structures around AI, not just augment individual workers.
“how do you start to reorganize the structure of the company to take advantage of the opportunities that new technologies provide?”
OpenAI disrupted AI-powered influence ops using fake journalists and think tanks
Content body was empty; no extractable signal from this submission.
Ben Affleck describes hands-on ML experience from Hollywood VFX workflows
“I can write like pretty shitty Python scripts and stuff like that because with convolutional neural networks, which were the sort of precursors to what the transformer can do”
Google demos real-time AI golf coaching using Gemini with 120fps motion capture
“Behind the scenes, what's happening is we are using motion capture cameras at a 120 fps, capturing all those stills, feeding those into Gemini 3.8 to do the heavy thinking.”
Databricks Genie uses ontology to resolve business context across multi-source enterprise data automatically
LangChain Academy launches certification program for agent engineers covering full ADLC lifecycle
“Our goal at LangChain Academy is to teach you how to create great agents and constantly improve them.”
62% of LLM-generated code is unsafe, while AI accelerates attacker capability simultaneously.
“62% of the results generated by LLM are unsafe or corrupted”
Anthropic releases Claude Haiku 5.5, matching GPT-6 Luna's price at $0.10/$0.50 per million tokens
“If your workloads fit in 100,000 tokens, Haiku is the same price as Luna and reports higher benchmark scores. Above 100,000 tokens, Luna looks like a much better deal.”
Factory's 'software factory' organizes AI agents to autonomously convert bug reports into production code.
“each step up the ladder of abstraction makes engineers more valuable, not the other way around”
Databricks introduces budget and alert features for cloud data warehouse cost management