The Hallway Track
Every signal tracked from the world's AI conferences — 4,665 talks, quotes & announcements, newest first.
Pinterest built an agentic LLM tool using MCP and ReAct to auto-diagnose Apache Spark job failures
“Our vision for a diagnostics agent was to ask it simply, 'Why did a job fail?' and get back a deep research document which provides evidence on the root cause of the failure.”
Trisca built a 4-agent AI pipeline automating oncology prior authorizations with zero human review
“confidence is a key metric that we were working towards”
Leaked Altman email reveals OpenAI planned open-source release to suppress rivals and block their funding.
“We think this helps discourage others from releasing similarly-powerful models, and makes it harder for new efforts to get funded.”
Altos Labs is applying foundation models to single-cell biology to model cellular rejuvenation and aging
“restore cell health and resilience through cellular rejuvenation to inverse disease and disabilities that can happen throughout the life”
Human-focused API architecture fails at agent scale, requiring ground-up rethinking of auth systems.
“the human-focused architecture doesn't scale well for agents”
Thundra built a runtime intelligence layer enabling coding agents to continuously fix production performance issues.
“Google just published their Dora metrics for 2026 and we can see that the biggest impact of AI adoption on engineering is individual effectiveness... But the second one is software delivery instability.”
Lyft applies ML model evaluation rigor to AI agents before production deployment.
“if we are running um offline evaluations for our machine learning model before that goes to productions, I think we should do the same for AI agents as well”
Corporate AI mania is driving irrational decisions disconnected from actual tool experience
“I have seen an executive confess that they had never even used ChatGPT or any AI tool in their life, immediately after producing a technical strategy for an organisation with $2B+ in revenue which was entirely centered around AI.”
Claude Code v2.1.181+ ships Bun's Rust rewrite in production across millions of devices
“Claude Code v2.1.181 (released June 17th) and later use the Rust port of Bun. Startup got 10% faster on Linux but otherwise, barely anyone noticed. Boring is good.”
AI coding agents plateau on scientific tasks because they run out of novel hypotheses, not implementation ability.
“they ran out of ideas uh or you know what people call them research taste”
Graph fundamentals offer underutilized opportunities for smarter, cheaper AI applications
“the more I learn about the fundamentals of graph data structures and algorithms, the more interesting opportunities seem to present themselves”
AI has a serious UX problem as features outpace user-friendly design
“Your scientists were so preoccupied with whether or not they could, they didn't stop to think if they should.”
AI self-improvement loops require domain-specific evaluation functions, not just agentic design.
“does the code compile or not”
Enterprises are abandoning rented AI inference after runaway costs force infrastructure ownership.
“It's almost as if you're loading credits inside a casino. You put some and then you pull it and then you are so addicted to it then you end up doing more and more of it and by some time you realize that you've blown past the threshold that you had mentally kept in mind.”
Simon Willison used Claude Fable to build a browser-based SQLite query plan explainer.
“Maybe one day I'll learn to read a query plan.”
Code is becoming the dominant medium for technical communication, replacing traditional content formats.
“increasingly code is becoming the way that we communicate”
AI agents need verifiable receipts and budget management, not just more tools, for scientific collaboration
“whenever AI agent workflow automation matures, organizations will give to the agents not only tools, they will give them budgets”
AI agents lack persistent state checkpoints, making debugging and replay impossible today.
“all of that is lost and it is only stamped as a read-only trace by the end, which is sitting in another tool far away from where the actual code is.”
AI agents performing high-stakes actions ship without feature flag safety infrastructure web teams mastered in 2012
“we are shipping them the way web team used to ship in 2008”
Anthropic reverses Fable 5 removal, making it permanent in Max and Team Premium plans due to competition.
“Beginning July 20, Claude Fable 5 will be included in all Max and Team Premium plans, at 50% of limits.”
Quixote Python web framework still receives commits 21 years after initial release
Kimi K3 shifts AI competitive narrative from compute moats to efficiency stacks, pressuring US labs.
“Kimi K3 is really, really good”
Practitioners debate whether AI agent loops live up to their hype in production software factories
Apple's new Music Understanding framework analyzes songs on-device with three lines of code.
“Your app can now understand music and it takes zero machine learning experience to set it up. All you need is three lines of code.”