METR and Redwood Research each sent one person to OpenAI from July 21 to August 11, 2026, to evaluate the misalignment behavior monitoring system for its internal agents (including Codex). The public draft states that about 75–94% of agent traffic in the internal Responses API at or above GPT-5.5 is eligible for monitoring, and the internal Codex agent subset is about 99.2%, but there are gaps such as compaction not being monitored, and paths were found that can bypass the sandbox and Auto-review. OpenAI says vulnerabilities affecting the production protection stack have been patched, and METR and Redwood have not independently reviewed these claims.
METR and Redwood Research each sent one person to OpenAI from July 21 to August 11, 2026, to evaluate the misalignment behavior monitoring system for its internal agents (including Codex). The public draft states that about 75–94% of agent traffic in the internal Responses API at or above GPT-5.5 is eligible for monitoring, and the internal Codex agent subset is about 99.2%, but there are gaps such as compaction not being monitored, and paths were found that can bypass the sandbox and Auto-revie
The author ranks 13 Personal Agents by the number of connectable apps: meta muse leads with 54, chatgpt dots has 50, gemini spark 35, today and poke 30 each, grok bot 20, manus cue 15, Instinct 12, while hark, Lingguang 2.0, Qwen Go, Hermes, and OpenClaw each have about 10. In terms of quotas and restrictions, Hark is free with no feature limits, muse is only available in the US and Canada, and Qwen Go, Lingguang 2.0, cue, and Instinct still require invitation-only beta access.
DeepSeek has integrated the Harness with the web, App, and open platform teams into a product department, led by Cui Tianyi, and the official website has posted recruitment positions in two directions: AI product manager and Agent Harness. The article uses OpenAI Codex lead Tibo Sottiaux and the Anthropic Claude Code team as examples to explain how the execution capabilities of coding tools can be extended into a general work task entry point. DeepSeek, while open-sourcing Harness and the inference framework DeepInfer, is also facing cost pressure on caching and transmission brought by long tasks.
web, App, and open platform teams into a product department, led by Cui Tianyi, and the official website has posted recruitment positions in two directions: AI product manager and Agent Harness. The article uses OpenAI Codex lead Tibo Sottiaux and the Anthropic Claude Code team as examples to explain how the execution capabilities of coding tools can be extended into a general work task entry point. DeepSeek, while open-sourcing Harness and the inference framework DeepInfer, is also facing cost
1Password launches the 1Password for Agents platform, enabling personal AI agents to complete website logins on users' behalf without accessing or storing users' passwords or verification codes. The feature is aimed at individual and family plan users, with initial support for agents including Meta Muse, OpenAI ChatGPT Work and Dots, Anthropic Claude Duet, Grok Bot, and Town. In the flow, the agent first proves its identity to 1Password using a digital key; the user sees on their device the account credentials and task requested for use and approves or rejects; after approval, 1Password directly fills in the username, password, or one-time verification code, and the password always remains in the vault.
1Password launches the 1Password for Agents platform, enabling personal AI agents to complete website logins on users' behalf without accessing or storing users' passwords or verification codes. The feature is aimed at individual and family plan users, with initial support for agents including Meta Muse, OpenAI ChatGPT Work and Dots, Anthropic Claude Duet, Grok Bot, and Town. In the flow, the agent first proves its identity to 1Password using a digital key; the user sees on their device the acco
Figma product manager Wayne Lin introduced three kinds of context for getting the Figma agent to call on a design system: connect a published design library so the agent can directly use team-approved components and tokens; write component selection rules and exceptions into a Guidelines file; then provide real examples through an Examples page or the _example naming convention. Guidelines works best when generating designs from scratch; consistency of application when editing existing designs is still being improved.
Figma product manager Wayne Lin introduced three kinds of context for getting the Figma agent to call on a design system: connect a published design library so the agent can directly use team-approved components and tokens; write component selection rules and exceptions into a Guidelines file; then provide real examples through an Examples page or the example naming convention. Guidelines works best when generating designs from scratch; consistency of application when editing existing designs is
Replit has launched Monitoring Agent, providing autonomous on-call capabilities for applications already in production: when a production issue arises, it can read production logs and metrics, identify the cause, verify the fix, and, once auto-publish is enabled, automatically deploy the fix and report the changes by email. This feature builds on App Monitoring, launched earlier this year; users can enable intelligent autofix alerts in the application's Publishing settings.
The Perplexity Agent API has added a Batch API, supporting up to 50,000 requests per file. Each request runs the full agent loop, and web_search and fetch_url are supported in batch mode. According to the official announcement, this can be used for large-scale data enrichment tasks, such as enriching a spreadsheet of 50,000 companies with real-time web data overnight.
The Perplexity Agent API has added a Batch API, supporting up to 50,000 requests per file. Each request runs the full agent loop, and web search and fetch url are supported in batch mode. According to the official announcement, this can be used for large-scale data enrichment tasks, such as enriching a spreadsheet of 50,000 companies with real-time web data overnight.
Anthropic announced that it is donating Agent Skills to the Agentic AI Foundation, which previously also received MCP. Since December 2025, more than 70 agent platforms have adopted this standard, including ChatGPT, Cursor, GitHub Copilot, and VS Code. Ado shared this news and said he is glad to see the standard opened up.
1Password on Thursday launched 1Password for Agents, which lets users authorize AI agents to log in to their personal accounts on their behalf, with the agent never touching the user's credentials throughout the process. The platform establishes communication among the agent, the user, and the password manager: when an agent initiates a login, it sends a unique digital key to 1Password to confirm its identity; after the user receives a notification and personally confirms, 1Password fills in the credentials on the user's behalf to complete the login. Users can choose agents from Anthropic's Claude Duet, OpenAI's Dots or ChatGPT Work, Grok Bot, Meta's Muse, and Town; 1Password says it is working with other companies to expand the service.
1Password on Thursday launched 1Password for Agents, which lets users authorize AI agents to log in to their personal accounts on their behalf, with the agent never touching the user's credentials throughout the process. The platform establishes communication among the agent, the user, and the password manager: when an agent initiates a login, it sends a unique digital key to 1Password to confirm its identity; after the user receives a notification and personally confirms, 1Password fills in the
Anthropic announced that it is donating the Agent Skills standard to the Agentic AI Foundation (AAIF) under the Linux Foundation, in order to keep it open, community-driven, and vendor-neutral. The AAIF was co-founded last year by Anthropic along with Block and OpenAI, and the MCP specification had previously been donated to this foundation as well. Simon Willison noted that he had predicted this development last December, and used the MCP plugin on simonwillison.net to retrieve that prediction of his from that time.
Anthropic announced that it is donating the Agent Skills standard to the Agentic AI Foundation (AAIF) under the Linux Foundation, in order to keep it open, community-driven, and vendor-neutral. The AAIF was co-founded last year by Anthropic along with Block and OpenAI, and the MCP specification had previously been donated to this foundation as well. Simon Willison noted that he had predicted this development last December, and used the MCP plugin on simonwillison.net to retrieve that prediction
Anthropic's Boris Cherny said that Agent Skills began as an experiment in Claude Code, and due to widespread industry adoption, Anthropic donated it to the Agentic AI Foundation, which also owns MCP. The cited ClaudeDevs post states that since December 2025, more than 70 agent platforms have integrated the standard, including ChatGPT, Cursor, GitHub Copilot, and VS Code.
The Agentic AI Foundation announced on October 22, 2026 the addition of four hosted projects: Agent Skills contributed by Anthropic, Open Responses contributed by OpenAI, OpenEnv contributed by Hugging Face, and OpenSandbox contributed by Alibaba, bringing the total number of hosted projects from 3 at its founding in December 2025 to 10. The four projects correspond to four layers: agents acquiring skills, applications invoking models across vendors, agent training and evaluation, and secure agent operation. Agent Skills is open-sourced under Apache 2.0 and CC BY 4.0, already supported by more than 70 agent platforms, with over 1 million skills published and nearly 280 million installs; Open Responses defines shared schemas and streaming events based on the OpenAI Responses API, with implementations already provided by Databricks Model Serving and Vercel AI Gateway.
The Agentic AI Foundation announced on October 22, 2026 the addition of four hosted projects: Agent Skills contributed by Anthropic, Open Responses contributed by OpenAI, OpenEnv contributed by Hugging Face, and OpenSandbox contributed by Alibaba, bringing the total number of hosted projects from 3 at its founding in December 2025 to 10. The four projects correspond to four layers: agents acquiring skills, applications invoking models across vendors, agent training and evaluation, and secure age
1Password has released the 1Password for Agents platform, enabling AI agents to complete operations requiring login credentials without viewing users' passwords. The platform initially supports Meta's Muse and SpaceXSI's Grok Bot, and says it will later integrate more partners including OpenAI and Anthropic. 1Password says the platform can connect to a user's single or multiple personal agents, with credentials and keys managed by 1Password.
Anthropic announced that it is donating Agent Skills to the Agentic AI Foundation; previously, MCP was also donated to the same foundation. Since December 2025, more than 70 agent platforms have adopted the standard, including ChatGPT, Cursor, GitHub Copilot, and VS Code.
Rescale built an AI triage agent based on the AWS AI-DLC framework and Strands Agents, going from problem definition to pre-release deployment in just five days and reaching production within weeks. The agent runs in an Argo Workflows CI/CD pipeline on Amazon EKS, calls Claude Opus 4.8 on Amazon Bedrock to classify nightly Playwright test failures as infrastructure issues, code defects, flaky tests, or environment problems, and queries pod logs via MCP, combining them with a historical failure knowledge base to generate structured remediation reports.
Rescale built an AI triage agent based on the AWS AI-DLC framework and Strands Agents, going from problem definition to pre-release deployment in just five days and reaching production within weeks. The agent runs in an Argo Workflows CI/CD pipeline on Amazon EKS, calls Claude Opus 4.8 on Amazon Bedrock to classify nightly Playwright test failures as infrastructure issues, code defects, flaky tests, or environment problems, and queries pod logs via MCP, combining them with a historical failure k
At AGNTCon + MCPCon, Google Cloud announced the launch of Agent Compute Engine (ACE), a VM type built specifically for AI agents and aimed at large-scale operation scenarios. The system comprises four components—Agent Gateway, Agent Registry, Agent Identity, and Agent Observability—and also expands Gemini's capabilities for enterprise agent governance.
Google Cloud unveiled a research reference architecture codenamed Project Chrysalis, demonstrating how to run governed self-evolving AI agents on the Gemini Agent Platform and ADK. In a simulated Black Friday cache avalanche failure, the architecture reduced time to first mitigation from 15–45 minutes to 1.8 seconds, cut the HTTP 5xx error rate from 72.7%–80.0% to 0.0%, and lowered P99 latency from 519.8 ms to 18.2 ms. All autonomous operations are constrained by governance controls including Model Armor, Cloud KMS, Binary Authorization, OPA, and two-person review HITL.
Google Cloud unveiled a research reference architecture codenamed Project Chrysalis, demonstrating how to run governed self-evolving AI agents on the Gemini Agent Platform and ADK. In a simulated Black Friday cache avalanche failure, the architecture reduced time to first mitigation from 15–45 minutes to 1.8 seconds, cut the HTTP 5xx error rate from 72.7%–80.0% to 0.0%, and lowered P99 latency from 519.8 ms to 18.2 ms. All autonomous operations are constrained by governance controls including Mo
Shao Meng used a Coding Agent to recreate "Breakout," the first WAP browser game on the Moto E770, in just over 10 minutes, and even simulated the phone's casing. He mentioned that the phone, which he saved up for a whole semester to buy during his freshman year, was lost in one of his moves, and after more than a decade it certainly can't be powered on. He had previously said he wanted to remake the lost little games he played as a child and during his school years, so that his son, born in the 2010s, could get to know them.
Shao Meng used a Coding Agent to recreate "Breakout," the first WAP browser game on the Moto E770, in just over 10 minutes, and even simulated the phone's casing. He mentioned that the phone, which he saved up for a whole semester to buy during his freshman year, was lost in one of his moves, and after more than a decade it certainly can't be powered on. He had previously said he wanted to remake the lost little games he played as a child and during his school years, so that his son, born in the
Claude Code creator Boris Cherny shared five practical principles in week four of Stanford's CS146S course. He personally runs thousands of Claude agents at once, and since last November nearly all of his team's code has been written by Claude, with per-person output up roughly 10x. The five principles include automating R&D bottlenecks (code review is already automated; approvals are next), building for the model six months out, small teams moving faster, scheduling agents in loops, and domain experts winning. He also noted that Claude Code's early revenue was flat for several consecutive months, and growth only surged after the release of Opus 4.
Claude Code creator Boris Cherny shared five practical principles in week four of Stanford's CS146S course. He personally runs thousands of Claude agents at once, and since last November nearly all of his team's code has been written by Claude, with per-person output up roughly 10x. The five principles include automating R&D bottlenecks (code review is already automated; approvals are next), building for the model six months out, small teams moving faster, scheduling agents in loops, and domain
1Password has released 1Password for Agents, allowing supported AI agents to access credentials in a user's personal 1Password vault in order to complete tasks on their behalf. At launch, the platform supports agents such as Anthropic Claude Duet, Meta Muse, OpenAI ChatGPT, and Dots. The agents themselves cannot see the user's credentials; after the user approves the login, 1Password fills in the credentials directly without sending them to the agent. 1Password said that partners including Browserbase, Browser Use, Perplexity, and Ramp will also add support in succession.
1Password has released 1Password for Agents, allowing supported AI agents to access credentials in a user's personal 1Password vault in order to complete tasks on their behalf. At launch, the platform supports agents such as Anthropic Claude Duet, Meta Muse, OpenAI ChatGPT, and Dots. The agents themselves cannot see the user's credentials; after the user approves the login, 1Password fills in the credentials directly without sending them to the agent. 1Password said that partners including Brows
