BetaWeekly digests are a beta — we're trying something new. Feedback welcome.
Coding Agents weekly digests
- Week of October 12, 2026Memory arrives for coding agents as Codex retires its old default model
Devin gained cross-session memory with a nightly review of past work, while Codex retired GPT-5.5 for ChatGPT sign-ins and pushed steering and a searchable /config into its CLI. Claude Code and Cursor also spent the week tightening defaults, permissions, and integrations.
- Week of October 5, 2026Agents get faster models, safer sandboxes, and a phone in the loop
OpenAI brought GPT-6.1 Sol Ultrafast to Codex as its new default, Anthropic hardened Claude Code's sandbox and shipped a cheaper Haiku default, and Cursor let developers steer their local agents from an iPhone.
- Week of September 28, 2026Strong models land across the coding agents
OpenAI's GPT-6.1 Sol became the default engine for Codex, Claude Sonnet 5.5 arrived with a 1M context window, and Devin pushed deeper into Jira and cross-device workflows — while a heavy week of fixes hardened every agent's terminal and sandbox.
- Week of September 21, 2026GPT-6 lands in Codex as the agents grow review and governance features
OpenAI brought GPT-6 Sol and Luna to Codex and ChatGPT Work at lower token prices than GPT-5.6, while Cursor launched deploy-monitoring and security-review bots and Devin reworked how automatic reviews get triggered.
- Week of September 14, 2026Coding agents learn to talk, and to clean up after themselves
Claude Code shipped a dense week of reliability work — self-healing corrupted sessions, proxy fixes, and AGENTS.md support — while Codex CLI and Devin both added live voice conversations, and Grok Build took Memory out of preview.
- Week of September 7, 2026Cloud agents, plugin packages, and a model migration deadline
Cursor launched Projects, turning its editor into a place where a coordinator agent plans work and delegates it to parallel cloud subagents, while Anthropic and OpenAI both pushed plugin and model-lifecycle changes that will force developers to update their configurations.
- Week of August 31, 2026Agents break out of the cloud as Codex retires its MCP server
Cursor shipped self-hosted machines that keep code, build outputs, and secrets inside your own network, while OpenAI removed Codex's standalone MCP server and Anthropic hardened Claude Code's auto-approval rules against credential-fetching escapes.
- Week of August 24, 2026Agents get triggers, guardrails, and faster starts
Claude Code shipped a hardening-heavy week with a restricted mode and a prompt-cache fix, Codex scheduled tasks can now fire on Gmail, Slack, and GitHub events, and Cursor's cloud agents no longer need a repo to get started.
- Week of August 3, 2026Security hardening dominates as Claude Code patches permission bypasses
Claude Code shipped three consecutive security-focused releases fixing Bash permission bypasses and sandbox escapes, while Devin graduated automations to GA with queueing support and Anthropic expanded Claude Code's deployment options with self-hosted runners.
- Week of July 27, 2026Cursor on iPad, Codex model migrations, and the identity layer for agentic tools
Cursor launched a full iPad app with PR review and an India pricing plan, OpenAI announced GPT-5.4 retirement from Codex while rolling out Sign in with ChatGPT, and several coding agents shipped improvements to MCP tooling, session management, and enterprise features.
- Week of July 20, 2026Model routing, voice, and sandbox hardening dominate the week
Claude Code ships nested subagents and a strict sandbox allowlist, Cursor Router intelligently selects models for Auto mode, and Codex adds ChatGPT Voice and multi-folder projects — a week of architecture-level changes across the coding agent landscape.
- Week of July 13, 2026Agents stop acting on their own
Claude Code stops auto-running verification and review skills, fixes a permission bypass in PowerShell, and redesigns its fork command. Cursor in Slack shares plans before starting work, and Devin simplifies its session sizing model.
- Week of July 6, 2026Side conversations take center stage
Codex lands in the ChatGPT desktop app with inline editing and PR review, while Claude Code fixes a critical terminal freeze and auto-mode rollout, and Cursor and xAI's Grok Build add durable side chats and conversation search.
- Week of June 29, 2026Permission models, default models, and the week of the agent platform
Claude Code shipped a new default model with a 1M-token context window and changed its permission architecture, while Grok Build delivered a dozen updates across voice, media, and air-gapped deployments, and Cursor expanded team marketplaces to MCPs.
- Week of June 22, 2026Codex Remote goes GA as Cursor and Claude Code polish mobile and terminal workflows
Codex Remote reached general availability with QR pairing and DigitalOcean plugin support, while Cursor launched an iOS mobile app for cloud agents and Claude Code v2.1.191 cut streaming CPU usage by 37% alongside permanent agent-stop fixes.
- Week of June 15, 2026The configuration layer arrives: customizable agents and unified management
Cursor unified plugins, skills, and MCPs into a single Customize page, while OpenAI's Codex added per-host personality settings and a first-party Record & Replay skill creator, and Claude Code shipped new blocking rules for destructive git commands.
- Week of June 8, 2026The biggest Fable in the story so far
Anthropic launched its most capable model yet, Claude Fable 5, while Cursor's Bugbot got dramatically faster and cheaper, and OpenAI's Codex iOS client gained workspace-level file management. Several tools also shipped new features for managing sessions and MCP tooling.
- Week of June 1, 2026Devin and Cursor broaden enterprise controls as platforms ship multi-agent polish
This week Devin added personal automations and per-PR spend limits, Cursor launched multi-team organizations with centralized governance, and agents across OpenAI Codex, Claude Code, and Grok Build shipped desktop handoffs, fallback models, and leader-session stability.
- Week of May 25, 2026Run mode security and agent orchestration shape the week
Codex CLI hardened command safety against repository-supplied Git hooks and browser-based websocket exploits, while Cursor introduced an Auto-review run mode and Claude Code debuted dynamic workflows for orchestrating tens to hundreds of background agents.
- Week of May 18, 2026iOS, Jira, and desktop shortcuts extend coding agents beyond the terminal
Codex landed on iOS with Spotlight and Shortcuts support, Cursor shipped a Jira integration and shared canvases, and multiple agents tightened security around permission prompts and sandbox escapes.
- Week of May 11, 2026Agent infrastructure, smarter sheets, and Opus 4.7 lands everywhere
Cursor and Claude Code shipped major agent workflow upgrades — Cursor with configurable development environments and full-screen tabs, Claude Code with an agent view and goal-completion mode — while Anthropic's Opus 4.7 reached fast mode across the ecosystem and OpenAI's Codex gained a richer TUI and iOS notifications.