The Hallway Track
Every signal tracked from the world's AI conferences — 4,665 talks, quotes & announcements, newest first.
METR finds AI dramatically accelerated cyber vulnerability discovery but not AI research itself
“The rate of vulnerabilities reported across many projects has dramatically accelerated in 2026 compared with 2025, both for specific projects (cURL, OpenSSL, Firefox, and Microsoft) and for aggregate vulnerability databases (the US NVD, and OSV)”
GPT-5.6 is now available in Kiro, offering improved price-performance for developers.
Linux executables can be structured as SQLite databases using ELF format tables
Anthropic's Fable 5 captures only 8% of enterprise spend, far behind cheaper Opus 4.8 at 28%.
“It also told investors that it had 6,000 customers that spend $100,000 annually or more.”
Fable's high cost ends the era of relying on new models to cheaply solve engineering problems
“Prior to Fable, it felt silly to waste too much time improving your coding harness or context strategies. A new model would arrive at the same price (or cheaper!) and paper over most of your problems.”
AI models and biological brains share geometric structure, suggesting intelligence is a physical law
“if you apply enough compute to matter, you get this thing that looks like intelligence”
Linus Torvalds used AI as a debug assistant on Linux kernel code, crediting it with commit message authorship.
“And this was a debug session from hell, enormously helped by an AI doing much of the grunt-work.”
Warp built a cloud agent platform after hitting limits of local laptop-based AI coding agents.
“we realized that we had kind of reached the limits of what we could do on our laptops, and we wanted agents to do work that was more long-running, that was adapted to different constraints”
llm 0.33 adds composable templates, per-call embedding keys, and reasoning summary options
1 in 20 production clinical AI notes contains errors serious enough to harm patients.
“1 in 20. That's not theoretical in testing, that's in production on real patients.”
Agent frameworks force humans to stay on-task, undermining the promise of autonomous agents
“SSH with vibes”
Coding agent adoption is blocked by organizational readiness, not technical limitations.
“a lot of developers told me, "We didn't sign up for this. We didn't sign up for better prompting, writing better specs."”
Effective coding agent use requires confident instruction and verification, not just line-by-line review
“The key skill required to make productive use of coding agents is being able to confidently instruct them on how to make changes and then confidently verify that those changes have been applied in the correct way.”
Benchmark-chasing is the wrong model selection instinct; task-specific routing beats one-model-fits-all
“There is no single best model. The right one depends on the actual request.”
AI tools boost individual output but ignore the quadratic alignment cost of larger engineering teams
“alignment beats individual skills”
Agentic AI lacks cost control planes, driving a shift from token maxing to value maxing.
“people are proud to call themselves token billionaires”
Anthropic's CI team lost 200 workloads in 90 seconds because giving agents a token and tool list isn't enough
“this idea of like here's a token and here's a tool list that is just like not enough like it just doesn't scale when the agent is starting to do some some like real work in production.”
AI pipeline components are flipping from human-made to model-made via simulation annually since 2022.
“what we used to call "synthetic data" and "synthetic rubrics" and "AI researcher" and "end to end RL environments" is just increasingly ambitious human simulation - 10% worse, but 100x cheaper and 10,000x faster.”
Models absorbing harness capabilities into weights is reshaping agent architecture toward human-attention scaffolding
“The change last winter, last Christmas — it's a little hard to pin down. I mean, the harness changed and a little post-training changed and then new pre-trained models came… but it felt like a big jump which is not that easy to pin down what did it.”
Simile AI raises $2B to build behavioral foundation models simulating human decision-making at scale
“What if we can just recreate the world that we live in?”
Chess.com CEO says AI acts like a pocket professor for accelerating skill acquisition
“you can literally get into, you know, whatever AI program you're using and you can start researching something and it can walk you through and it can help you get exposure to ideas, refine ideas”
CD Projekt Red built device-specific presets for Cyberpunk 2077 on Mac using MetalFX Upscaling.
“Click play, drop into Night City, and enjoy the game.”
Microsoft Copilot skills are shareable markdown-based workflow files for reusable agent automation
“a skill is just a sequence of steps into a workflow and that is all combined into a markdown file called a skill.md file”
GitHub Copilot guides developers to modernize plain HTML portfolios with Bootstrap responsive components