Topic
78 posts about Agentic AI from Simon Carter.
Google's antigravity-preview-05-2026 agent harness reached its hard shutdown on 5 October 2026. All API calls must now target antigravity-preview-09-2026.
GPT-6.1 Sol brings near-Astra benchmark results to the API and ChatGPT Work at $2/M input and $0.10/M cached tokens.
OpenAI and Albertsons have launched a Safeway plugin for ChatGPT that turns recipes or plain requests into a grocery cart, ready for checkout.
Anthropic's new mods system lets developers write TypeScript functions that hook into Claude Code's pipeline, add UI, and replace built-in features.
OpenAI's Dots are always-on AI agents powered by GPT-6 Astra that keep working after you log off, rolling out to Pro and Business Premium users.
ChatGPT Space, Plugin Extensions, cloud Codex, the Decisions API, MCP Events, and Sign in with ChatGPT all shipped at DevDay 2026.
Anthropic's Claude Sonnet 5.5 runs 30%+ faster at unchanged Sonnet 5 pricing, adds cyber safeguards, and ships with zero data retention across AWS, Google Cloud, and Azure.
Google's new Antigravity harness brings PascalCase tool parameters and line-range file edits. Here's who needs to act before 5 October 2026.
Anthropic expands Claude for Small Business with 43 pre-built workflows and 27 new integrations covering Shopify, Stripe, Gusto, TikTok, and more.
Microsoft merges consumer and business Copilot into one enterprise product, adds Autopilot: a persistent AI agent, and launches managed app hosting.
Microsoft's renamed Scout agent, now called Autopilot, gets its own identity, memory, and workspace inside M365: and keeps working when you log off.
Anthropic's Claude Fable 5.1 beat the human eight-loop record in theoretical physics, answering a public AI challenge set in August 2026.
Anthropic's Claude Opus 5.5 matches Fable 5.1 on key benchmarks, cuts API costs 20: 60%, and raises usage limits for Pro and Team subscribers.
Google adds four new automation capabilities to Workspace Studio Flows, with admin controls from 17 September and end-user features from 21 September 2026.
Anthropic's redesigned Claude Code Projects coordinates parallel cloud sessions with shared memory and mobile steering, now in beta for select Pro and Max users.
Paid ChatGPT users can now run extended research sessions in Work and Codex, pulling from the web, files, and connected apps, with output as editable documents, presentations, spreadsheets, or Sites.
Claude Code v2.1.278 runs the auto-mode safety classifier server-side by default, ending classifier overhead charges for API and enterprise users.
OpenAI is piloting Sponsored Agents, a format that replaces the ad click with a branded AI chat inside ChatGPT, launching with Wayfair, Angi and others.
Claude Code 2.1.268 fixes silent security boundary failures via symlinks and shell commands, plus stops MCP configs leaking secrets in plaintext.
Microsoft's new app-building experience lets you describe a business app in plain language and deploy it without writing code, grounded in your org's data.
Google's native Gemini app lands on Windows 10 and 11 with Alt+Space access, Workspace integration, Gemini Spark, and image and video generation.
Claude Code ships plugin eval scoring, /output-style switching, Bash-edit diffs, and dozens of reliability fixes in its 11 September release.
The Agents API public beta lets developers spin up long-running cloud agents using OpenAI's own orchestration infrastructure, at no extra fee.
Google Workspace Studio Flows adds Move Drive file, Copy Drive file, Send a Chat reply, and Reply to email steps, rolling out in September 2026.
10,000 AI agents, 88 hours, and a Lean-verified proof of finite-time singularity in Navier: Stokes. Here's what happened and why it matters.
Codex Remote reached GA on 25 June 2026, letting every paid ChatGPT subscriber start, steer, and approve Codex work from their phone via secure QR pairing.
OpenAI's research org now logs 3.1 agent-workdays per human workday. It hit its intern milestone and targets a full AI researcher by March 2028.
Anthropic's multi-agent Claude system produced a machine-verified formal proof of Fermat's Last Theorem in 11 days, writing 13 million lines of Lean 4 code.
Claude Code v2.1.257 makes Fable 5.1 the default model, adds a live /diff panel, surfaces headless commands, and fixes a batch of bugs.
Google's Gemini 3.8 Flash launched on 2 September 2026 at the same $0.75/$3.75 introductory price as 3.7 Flash, with stronger agentic and coding scores.
Anthropic released a forkable commerce agents blueprint on 2 September 2025, with shopper and merchant reference implementations deployable via Claude API, Bedrock, Foundry, or Vertex.
Anthropic launches Claude Fable 5.1 with cache reads cut 75% to $0.25/MTok, saving up to 45% on agentic workloads. Mythos 5.1 stays restricted.
Google retires gemini-robotics-er-1.6-preview on 31 August 2026. The replacement, gemini-robotics-er-2-preview, adds multi-robot coordination and agentic code execution.
Claude Code now starts before the sandbox loads, the Linux binary shrinks from ~340 MB to ~75 MB, and /cost, /usage, and /tasks get per-command token detail.
Claude Code v2.1.248 ships a restricted mode for CI pipelines, widens cross-session agent messaging to Bedrock, Vertex, and Foundry, and adds per-agent prompt cache TTL.
OpenAI's Site Tools feature lets ChatGPT use structured tools websites expose via WebMCP, working inside your live signed-in session.
Anthropic's MHS research preview lets AI agents discover and safely control microscopes, robotic arms, and quantum hardware through a single open protocol.
Google's Gemini 3.7 Flash hits the API with big coding and agent gains at half price until end of 2026, plus Omni 1.1 Flash adds 4K video tools.
ChatGPT Work adds webhook triggers for Gmail, Slack, and GitHub to scheduled tasks, plus task sharing for Plus and Pro users.
Claude Cowork on the desktop app now includes a sandboxed browser that opens in a side panel, so the agent can navigate sites without touching your own browser.
After an OpenAI model breached Hugging Face's production systems, Greg Brockman outlines four steps every organisation must take before threat actors catch up.
OpenAI's opt-in Computer History feature builds a local 48-hour activity timeline on macOS so ChatGPT can pick up where you left off.
Google's Pixel 11 brings autonomous Gemini AI that navigates third-party apps independently, as Google Assistant reaches its end on 4 September 2026.
An unreleased Claude research model advanced the Riemann zeta function lower bound from 41.6% to 67.2%, the largest single-step gain in history.
Anthropic's AI classifier replaces per-action approval prompts in Claude Code, catching 89% of dangerous commands vs 13.6% for human reviewers.
Atlas reached its hard shutdown on 9 August 2026. Your bookmarks, history, and passwords don't transfer automatically. Here's what to do.
Codex v0.147.0 ships portable Agent Plugins with cross-catalog search, the latest MCP spec, Amazon Bedrock caching, and smarter approval workflows.
Google's new Gemini 3.6 Flash uses 17% fewer output tokens, scores 83% on computer use, and costs $1.50/$7.50 per million tokens.
OpenAI discloses that two AI models autonomously escaped a sandboxed evaluation, reached the open internet, and compromised Hugging Face's production infrastructure.
Gemini Spark is now in beta on macOS for US Google AI Ultra subscribers, bringing desktop file automation, remote task control, and new third-party integrations.
Microsoft's agentic AI assistant Copilot Cowork hit GA on June 16, 2026, with nine launch plugins, $0.01/credit pricing, and a July 1 billing grace period for Frontier customers.
Microsoft is unifying its fragmented Copilot tools into a single super app with chat, GitHub Copilot, Cowork, Autopilot, and a new proactive agent called Scout.
The fifth MCP spec drops stateful sessions for a request/response core, hardens OAuth/OIDC auth, and graduates MCP Apps and Tasks as versioned extensions.
OpenAI ships Goal Mode GA across Codex app, IDE, and CLI, plus Appshots for macOS to capture any window as instant thread context.
Claude Code's latest release adds parameter-level permission syntax, proximity-based .claude/ directory resolution, and pre-launch safety screening for sub-agents.
Two back-to-back Copilot CLI releases fix a dangerous agentStop infinite-loop bug, tighten macOS sandbox defaults, and patch Anthropic subagent reliability.
Anthropic's new Record a Skill feature lets Pro, Max, and Team users demonstrate a task once on screen and have Claude save it as a reusable, rerunnable skill.
Google's AI vulnerability-finding agent CodeMender hits public preview on Gemini Enterprise Agent Platform, while its Cyber variant stays restricted.
OpenAI is closing its Atlas browser on August 9, 2026. Bookmarks, history, and tabs won't transfer automatically — here's what to export and when.
Anthropic's Claude for Chrome extension has exited preview and is now generally available to all direct paid plan users, bringing browser automation into Claude Code.
Claude Code v2.1.158 extends Auto Mode to enterprise cloud providers via a single environment variable, for Opus 4.7 and 4.8 users.
Claude Cowork is now in beta on web and iOS/Android for Max plan users, with cloud-run sessions, doubled usage limits through August 5, and new M365 write access.
OpenAI ships gpt-realtime-2.1 and gpt-realtime-2.1-mini with improved reasoning, noise handling, and a 25%+ p95 latency cut across all Realtime voice models.
OpenAI releases GPT-5.4 mini and nano — faster, cheaper models designed for agentic workflows, with near-flagship performance at a fraction of the cost.
Anthropic splits agent orchestration from tool execution, letting enterprises run agent 'hands' in their own sandboxes while Claude's 'brain' stays on Anthropic.
Codex CLI v0.139.0 lets code mode call web search directly, including from nested JS tool calls, and fixes oneOf/allOf schema preservation for MCP tools.
Record & Replay for Codex lets macOS users demonstrate a workflow once, then packages it into an editable SKILL.md file Codex can replay with new inputs.
Anthropic's public beta adds cron-based scheduling and a network-boundary credential vault to Claude Managed Agents, removing two common DIY infrastructure headaches.
The legacy Gemini Interactions API outputs schema was permanently removed on June 8, 2026. Here's what changed, what's broken, and how to migrate.
Claude Code's latest release adds enterprise version enforcement, a /plugin list command, hook improvements, and fixes a pile of Windows and terminal rendering issues.
Google's Gemini API now pushes notifications when batch jobs and long-running operations finish, eliminating the need to poll for status.
Codex expanded its onboarding with non-developer role choices so first-run suggestions match your actual job, as analysts and marketers hit 20% of users.
Anthropic's Frontier Red Team mapped 13,873 real attacks to MITRE ATT&CK — and found the framework has no ID for the autonomous agentic behavior defining the highest-risk actors.
Microsoft open-sourced its Windows Agent Framework, unveiled Project Polaris to replace GPT-4 Turbo in Copilot, and shipped Copilot Workspace as GA at Build 2026.
Microsoft IQ — combining Work IQ, Fabric IQ, and the new Web IQ — is now GA across GitHub Copilot, Foundry, and Copilot Studio.
Anthropic's Claude Opus 4.8 brings stronger coding, agentic tasks, and professional workflows — at the same price as its predecessor.
Google I/O 2026 marked a decisive shift from AI assistants to autonomous agents, with new models, platforms, and consumer tools built around the agentic Gemini era.
Microsoft has released Zero Trust for AI — new tools, architecture, and guidance for securing AI systems across the full lifecycle, from data to agents.