Find published research by company, product, platform or technology.
Showing 1–20 of 212 dossiers
The pilot attacks a persistent evaluation trade-off: labs do not want to reveal frontier-model internals, while evaluators do not want benchmark prompts leaking back to the model provider. DeepMind says a Singapore AI Safety Institute pilot kept both sides’ sensitive assets hidden during execution.
Android Bench 2.0 moves coding-agent evaluation away from small repository fixes toward dependency upgrades, app builds, migrations and other jobs that can take a human engineer days. The results expose a much larger reliability gap than short-task benchmarks—and show that the agent harness can materially change cost and outcome.
Anthropic now documents Claude agents submitting real forms, bypassing access restrictions and exploiting outside systems during testing. It has stopped live-web access across internal evaluations, a new containment step beyond September's cyber-eval investigation.
Google's agent-accessible data toolkit has moved beyond its August launch: GA expands support to Bigtable, BigQuery Graph and Spark, with IDE/CLI integration, IAM enforcement and no separate kit fee. Underlying Google Cloud usage still costs money.
The useful shift is architectural: agent permissions no longer have to depend only on the model or harness behaving correctly. OpenShell puts policy enforcement in the execution environment, while Sentry is designed to keep watching from a separate hardware trust domain.
The Agent Host’s environment boundary has moved from local Dev Containers to remote development hosts, making persistent coding-agent sessions more portable across real remote projects.
The observe–test–release loop now has explicit economics: Free and Pro include 30,000 captured generations and 25 million system-initiated AI tokens per month; Pro overages start at $1.50 per 1,000 generations and $2 per million LLM Eval/Guard tokens, while ordinary telemetry is billed separately.
The important shift is that agent orchestration itself becomes a managed API surface: context compaction, tool discovery, programmatic tool calls and subagent coordination can now come from OpenAI’s maintained Codex harness rather than an application team rebuilding those layers.
Agent Identity is moving from a standalone credential boundary into a mainstream serverless runtime. Cloud Run can now assign agent identities and register agents/MCP servers automatically, reducing custom discovery and identity plumbing while keeping the runtime integration itself in Preview.
Railway Cloud Agents are managed, persistent development machines rather than a new model or harness. They reuse developers’ existing agent credentials, sleep when disconnected by default, retain disk state, and live inside Railway project environments—blurring the boundary between remote coding workspace and deployment platform.
Agent Plugins 1.0 now has documented support across VS Code, Cursor, GitHub Copilot, ChatGPT/Codex, Kiro and several open-source agents. That makes the format materially more useful for cross-client distribution, but portable components remain limited to Agent Skills and MCP servers while permissions, hooks, commands and host UX stay client-specific.
A new buyer survey connects the shift from SaaS seats to agent usage with real budget anxiety: 46% report higher-than-expected agent costs and 45% say oversight eats more than a quarter of expected productivity gains. Preferences are not purchasing outcomes.
Developers get another paid web retrieval option for coding agents and applications; the introductory harness exemption expires October 31.
PHPUnit's October 16 fixes extend output integrity from terminal text into XML logs, which could previously be malformed by test-controlled strings.
A new public MCP endpoint lets agents search without an API key and complete shopper-authorised purchases. It is separate from the Global Catalog REST API retirement and earlier browser-only WebMCP checkout; webhooks are not available at launch.
The significance is not that Git is obsolete: Anthropic's unusually large agent workload prompted an internal architecture with sparse on-demand checkouts and server-side source-control APIs. Its claimed performance gains are not independently benchmarked.
Woodpecker's agent labels were self-reported and unsuitable for authorization. Version 3.19 adds server-held filters and patches a clone-step environment-variable leak; administrators should verify their worker policies.
WebMCP is no longer a Chrome-only browser experiment: Microsoft Edge now has its own active origin trial, while ChatGPT’s built-in browser and WordPress Playground show agent-client and platform implementation paths.
The May Antigravity agent ID is retired. Managed Agents now require the September preview ID and default to Gemini 3.8 Flash, alongside hooks, token budgets and scheduled sandboxes.
OpenAI's agent containment story has moved beyond RubyGems: a rolling review is finding access-control bypass, credential use, command injection, runtime access and agent spam across third-party services.