--- title: "Codex mobile lands, OpenCode ships pinned sessions" url: https://daily.dev/posts/codex-mobile-lands-opencode-ships-pinned-sessions-wkgqwxblx source_url: https://daily.dev/posts/codex-mobile-lands-opencode-ships-pinned-sessions-wkgqwxblx type: freeform source: "Agentic Digest" published: 2026-05-16T04:17:58.465Z updated: 2026-06-18T02:13:36.372Z tags: ["agentic-ai", "ai-coding", "claude", "openai-codex", "opencode"] reading_time: 6 upvotes: 2 comments: 0 language: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Codex mobile lands, OpenCode ships pinned sessions **[Agentic Digest](https://daily.dev/sources/agents_digest)** · 6 min read · 2 upvotes · 0 comments ## Summary OpenAI's Codex is now available in the ChatGPT mobile app, enabling developers to start sessions on desktop and continue on mobile. OpenCode 1.50.1 shipped pinned sessions with git worktree support for per-PR isolated contexts. GPT-5.5 experienced a 48-hour degradation in Codex due to two bugs, now fixed with usage limits reset. Anthropic committed $200M with the Gates Foundation for Claude deployment in global health and education. Additional notable items include: v0 Browser Use from Vercel, Block donating Goose to the Linux Foundation under a new Agentic AI Foundation, LangChain Interrupt announcements (SmithDB, LangSmith Engine), Grok CLI plugin support with a Vercel plugin, AWS Kiro requirements analysis using SMT solvers, a TanStack npm supply-chain attack via GitHub Actions cache poisoning, Intercom doubling engineering throughput with Claude Code, and Linux kernel rules for AI-assisted bug reports. ## Content **TLDR:** OpenAI's Codex is now in the ChatGPT mobile app, and developers are already using it to manage devices via Tailscale and continue sessions between phone and desktop. OpenCode 1.50.1 shipped pinned sessions with git worktree support, giving each PR its own isolated context. GPT-5.5 had a rough 48 hours in Codex due to two bugs OpenAI has since fixed, with usage limits reset. Anthropic committed $200M with the Gates Foundation to deploy Claude across global health and education. --- ## Codex goes mobile OpenAI shipped Codex inside the ChatGPT mobile app, and the reaction from developers has been immediate. People are starting sessions on desktop and picking them up mid-meeting on their phones, managing Tailscale device networks from an iPhone, and generally treating their laptops as secondary devices. OpenAI also teased a bidirectional ChatGPT-Codex integration with a cryptic "one day this will make sense" — something is coming there. A "Locked use" setting is reportedly in development that would let Codex operate on a Mac while the screen is locked. ## GPT-5.5 degradation in Codex, now fixed For roughly 48 hours, GPT-5.5 was noticeably weaker inside Codex. OpenAI identified two bugs, shipped fixes, and reset usage limits. No details on what the bugs actually were, but the team confirmed monitoring is ongoing. If you noticed Codex behaving strangely earlier this week, that's why. ## OpenCode 1.50.1: pinned sessions and worktrees OpenCode shipped session pinning in 1.50.1, designed to pair with the experimental `/warp` feature. The workflow: each PR gets its own git worktree, you pin the session, and when you start your day you pick up exactly where you left off from a list. The worktree feature itself is coming out of experimental next week once one more complementary feature ships — the team is teasing what that feature is. There's also an open architecture question about whether OpenCode should fork a background server on first launch so all clients (CLI, web, desktop) share one instance. ## Anthropic commits $200M with Gates Foundation Anthropic announced a four-year, $200M partnership with the Bill & Melinda Gates Foundation to deploy Claude in global health, education, and agriculture. That's four times the size of OpenAI's $50M Gates Foundation deal. The work covers vaccine and drug research for neglected diseases, making epidemiological models accessible to non-specialists, and AI tutoring in sub-Saharan Africa and India. Benchmarks, datasets, and evaluation frameworks will be released as public goods. --- ## Also notable - **OpenClaw's token spend:** The team behind OpenClaw runs ~100 concurrent Codex instances continuously — reviewing every PR, scanning every commit for security issues, deduplicating issues, and even creating PRs from meeting discussions in real time. Their framing: "how would we build software if tokens don't matter?" - **v0 Browser Use:** Vercel's v0 can now open the apps it builds, interact with them, critique designs, debug flows, and send screenshots back to the user while it works. - **Block handed Goose to the Linux Foundation:** To resolve trademark and governance issues blocking enterprise adoption, Block co-founded the Agentic AI Foundation (AAIF) under the Linux Foundation with Anthropic's MCP team; Goose, MCP, and Agents.MD are the initial projects. - **LangChain Interrupt announcements:** SmithDB (purpose-built distributed database for agent observability, traces in seconds instead of minutes), LangSmith Engine, Managed Deep Agents, and LangSmith Sandboxes are now generally available. - **Grok CLI + Vercel plugin:** xAI's Grok CLI now supports plugins, and the Vercel plugin adds cloud deployment directly from the CLI; Grok Build (agentic CLI for coding and automation) is in early beta for SuperGrok Heavy subscribers. - **Grok inside Hermes:** NousResearch's Hermes agent now accepts Grok subscriptions directly, letting users bring their existing Grok access into the agent. - **Linux kernel AI reporting rules:** New kernel docs for Linux 7.1 clarify that AI-assisted bug reports must be treated as public disclosures, must be plain text without Markdown, and must include verified reproducers — a direct response to a surge in low-quality AI-generated security reports. - **Anthropic's $1.5B author settlement stalled:** A federal judge declined final approval, requesting more detail on counsel fees, expenses, and service awards before signing off; the settlement covers ~480,000 works at roughly $3,000 each. - **Musk vs. Altman closing arguments:** Prediction markets give Musk a 32% chance of winning his $134B suit; his case is weakened by his own 2017 push to make OpenAI for-profit and his admission that xAI distilled OpenAI models. - **Grok Code Fast 1 deprecated:** GitHub deprecated Grok Code Fast 1 across all Copilot experiences as of May 15; migrate to GPT-5 mini or Claude Haiku 4.5. - **ChatGPT personal finance:** OpenAI launched bank account integration via Plaid for ChatGPT Pro subscribers in the US, covering balances, transactions, and investments across 12,000+ institutions; the feature cannot see full account numbers or make transactions. - **AWS Kiro requirements analysis:** AWS added SMT solver-based requirements analysis to Kiro that caught issues in 60% of first-draft specs across 35 internal projects; the pipeline rewrites natural-language requirements into formal logic and surfaces contradictions as plain-language questions. - **TanStack npm supply-chain attack:** Attackers published 84 malicious versions across 42 packages via GitHub Actions cache poisoning and OIDC token extraction — covered in Dev Weekly #111. - **Zed adds ChatGPT support:** The open-source IDE now supports ChatGPT subscriptions inside the Zed Agent, with the same usage limits as Codex. - **Intercom doubled engineering throughput:** By standardizing on Claude Code, building an internal skills library, and connecting the agent to all engineering systems, Intercom hit a 17.6% automatic code approval rate and doubled PR throughput, validated by Stanford researchers. - **PFF case study:** Two engineers using agentic workflows achieved 25x more deploys and 10x output compared to a team of 10, completing a 4-month project in under 2 months. - **China agentic commerce:** Alibaba's Qwen integration with Taobao's 4 billion product catalogue reached 300 million monthly active users; Alipay processed 120 million AI-agent transactions in a single week. - **OSS security strip mining:** Metabase went from ~10 security submissions per month to ~10 per week starting early 2026 as LLM-powered scanners bulk-scan public code; some commercial OSS projects like Cal.com are going closed source in response. - **Bun Rust UB issue:** A GitHub issue reports undefined behavior in safe Rust in Bun's codebase, with community criticism pointing at an AI-assisted Zig-to-Rust rewrite as the root cause. ## Similar posts on daily.dev - [OpenAI brings Codex to the ChatGPT mobile app](https://daily.dev/posts/openai-brings-codex-to-the-chatgpt-mobile-app-usgk8lgxx) · The New Stack · 0 upvotes · 0 comments - [OpenAI says Codex is coming to your phone](https://daily.dev/posts/openai-says-codex-is-coming-to-your-phone-c24qjzdpe) · TechCrunch · 2 upvotes · 0 comments --- Tags: [#agentic-ai](https://daily.dev/tags/agentic-ai), [#ai-coding](https://daily.dev/tags/ai-coding), [#claude](https://daily.dev/tags/claude), [#openai-codex](https://daily.dev/tags/openai-codex), [#opencode](https://daily.dev/tags/opencode) [View this post on daily.dev](https://daily.dev/posts/codex-mobile-lands-opencode-ships-pinned-sessions-wkgqwxblx)