The Hallway Track
39 tracked signals on claude-code.
Prompt injection attack bypasses Claude Code Auto Mode with 80% success rate
“The safety mechanism itself can become part of the failure. The classifier allowed the creation of the malware process, but then it blocked the command intended to stop it!”
Claude Code auto mode becomes default, blocking 89% of harmful actions vs humans' 13.6%
“In this evaluation, none of the 720 attack attempts succeeded against Claude Fable 5, Opus 5, or Sonnet 5 running auto mode.”
Opus 5 achieves prompt injection resistance and can run autonomously for days without scaffolding
“A year ago the model would have just done it. But nowadays Opus does not.”
Docker's new SBX microVM sandbox secures AI agents from local data exfiltration risks.
“What if I try to hack myself?”
Anthropic's Claude Code evolves toward self-modifying agents via 'Cloud Brain, Local Hands' architecture.
“From the rapid rise of Claude Code to a future where agents can rewrite their own harnesses, collaborate across teams, and operate across cloud and local environments, the way we build software is changing extraordinarily fast.”
Claude Tag handles 65% of Anthropic's own product engineering PRs via Slack integration
“we've all gotten a chance to take a step back and delegate a lot more of the menial implementation to Claude”
Anthropic launched Claude Tag, a Slack-native way to delegate work to Claude as a proactive, persistent teammate.
“third major redesign of LLM UIUX”
Apple's Game Porting Toolkit 4 adds agentic coding skills that let coding agents port games to Apple platforms far faster.
“We're at a moment where coding agents are fundamentally changing how software gets built. Game porting is no different.”
Uber caps employees at $1,500/month per AI coding tool like Claude Code to control runaway costs.
“The rideshare giant is limiting all employees to $1,500 in monthly token spending per AI coding tool, an Uber spokesperson said in response to a Bloomberg News inquiry.”
Anthropic and OpenAI shifted enterprise pricing to API usage, signaling true product-market fit.
“I had assumed that companies making extensive use of agents were getting similar discounts. It turns out I could not have been more wrong about that.”
Open models and even Google lack a true competitor to Claude Code and Codex for agentic work.
“even the mighty Google doesn't have a clear competitor for Claude Code and Codex”
Claude Opus 5.5 and Sonnet 5.5 now available in AWS GovCloud with FedRAMP Class D and DoD IL4/IL5 authorization
Databricks demo pairs Git worktrees with Lakebase database branching to safely isolate parallel coding agents
“Each agent gets its own isolated branch of the database, so it can run migrations, test changes, and experiment without affecting anything else.”
Databricks' Unity Gateway CLI lets enterprises govern and cost-control AI coding agents like Claude Code and Codex.
“From 0 to 60% I use Claude Code and the models are available there. 60 to 80% of the time I use Codex and a specific system model. And if I'm hitting 80-100% of my budget, I want to use a cheaper model like Kimmy.”
Impeccable's author shares hard-won lessons on engineering advanced agent skills beyond simple prompts.
“let's talk about the dark arts of skill engineering”
A new hire reports their big-company team is forced to ship Claude Code output nobody reads or understands.
“People are working 12 to 13 hours a day just to press enter. Nobody is reading anything.”
Anthropic, Google, and others are standardizing persistent, asynchronous agent runtimes as the default UX.
“one conversation, many cloud threads”
Anthropic defines agents by leaning on model intelligence with minimal core primitives in a loop.
“Agents in anthropic are actually very simply defined where essentially you want to give the model and lean in on model intelligence as much as possible.”
monday.com's production AI agents on Bedrock boosted per-engineer PR throughput by over 50%
“Agents are teammates, not jobs”
Anthropic's Claude models, including newly released Opus 4.8, are available in Microsoft Foundry for agentic coding.
“We've pretty well solved coding when you look at it from this metric's perspective.”
Conductor lets developers manage multiple AI coding agents from one desktop interface simultaneously
“Conductor is a desktop application for managing a team of programmers simultaneously.”
Claude Code now supports AGENTS.md, falling back to it when no CLAUDE.md exists.
“Starting today in version 2.1.277, if there is no CLAUDE.md in a folder, Claude will check for and use AGENTS.md.”
A new video compressor tool was developed using FFMPEG and WebAssembly.
Production AI agent loops need discipline to avoid unreviable 40,000-line PRs on real teams
“we're building 40,000line PRs that just nobody wants to read”
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!”
Claude Code defaults to agentic grep search, but semantic code retrieval yields measurable accuracy and retention gains.
“they just kind of found agentic search which is actually you know, grepping through your file system kind of worked better and you know, seems to work simpler for cloud code”
AI-written code is rising from ~10% today toward 40-50% by year-end, and AI is increasingly the harness, not just the model
“this is my argument around why AI engineering will be the last job, why you guys all have job security and why you guys will take over the world”
Existing security tools lack context to govern AI coding agents that may autonomously delete production databases.
“if cloud code is working on an unrelated task and suddenly thinks that maybe the right thing to do is to delete our database and recreate it, maybe we don't want that to happen”
NVIDIA proposes adding specialized deep research skills to agent harnesses like Claude Code
Claude Code can build a Copilot Studio AI agent in under 10 minutes
“we can develop an agent in 10 minutes”
LangChain launches AI-powered coding agent tutors for LangChain Academy courses
New /wayfinder skill helps AI agents navigate open-ended projects with unclear end states
“you can't quite decide everything right at the start.”
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.”
WorkOS engineer uses Claude Code with MCP-connected Slack and Linear to fix and self-verify bugs autonomously.
“despite, you know, getting more done than ever before, like you're completely fried at the end of the day”
AWS demos running coding agents like Claude Code, Codex, and Kiro in the cloud via Bedrock AgentCore runtime.
“we want our coding agents to be running and uh building software”
Claude Opus 5.5 generates pixel art animations from photos via simple prompts
“I had seen some buzz around how good Claude Opus 5.5 was at creating pixel art animations.”
Anthropic Python SDK v1.0 switches to httpx2, mirroring OpenAI's identical move two weeks prior.
“Upgrade to anthropic>=1 - read https://raw.githubusercontent.com/anthropics/anthropic-sdk-python/refs/heads/main/MIGRATION.md and get the tests passing”
Cloudflare CAPTCHA rules can be scoped to only challenge search URLs containing an ampersand, fixed via Claude Code.
“After some mucking around with Claude Code it turns out you can register the following rule instead, so the CAPTCHA only kicks in for search URLs containing at least one ampersand”
Simon Willison used Claude Code (Opus 4.8) to find methods mapping SQLite result columns back to their source table.column.
“I decided to set Claude Code (Opus 4.8, since Fable is currently banned by the US government) on the problem.”