ai-powered-markdown-translatorArticle translated from French to English with gpt-5.4-mini.
June 27, 2026 is a compact day dominated by two notable announcements: OpenAI’s Codex Remote leaves preview and reaches general availability, with secure QR pairing and a new DigitalOcean plugin to provision cloud environments from mobile; Sakana AI brings Fugu-Ultra to Vercel AI Gateway, its multi-agent orchestrator available at no platform markup with performance comparable to Claude Fable 5. On the tooling side, Claude Code v2.1.195 ships 12 targeted fixes and GitHub enriches Copilot metrics for enterprise administrators.
Codex Remote generally available
25 June — OpenAI announces the general availability of Codex Remote, its remote coding agent that had been in preview until now. Access is available from the ChatGPT mobile app on iOS and Android, allowing developers to start or control Codex sessions without being at their workstation.
GA brings two concrete new features:
- Authenticated QR pairing (Remote Control): each iOS or Android device is paired individually with each host via QR code, ensuring one-to-one authentication without sharing credentials.
- DigitalOcean plugin: Codex can now provision a DigitalOcean Droplet, configure SSH access, and connect the instance directly to the Codex environment — a complete cloud-to-mobile integration, without manual infrastructure configuration.
The full documentation is available under /codex/remote-connections.
🔗 Codex changelog — Codex Remote GA
Sakana Fugu-Ultra joins Vercel AI Gateway
27 June — Sakana AI announces the integration of Fugu-Ultra into Vercel AI Gateway, accessible via model identifier sakana/fugu-ultra in Vercel’s AI SDK.
Fugu-Ultra does not rely on a single model: it orchestrates a pool of public frontier models and routes each request to 1 to 3 agents depending on the problem’s complexity, then combines their results into a consolidated answer. According to Vercel, its performance on reasoning and scientific benchmarks is comparable to Claude Mythos Preview and Claude Fable 5.
On pricing, the model reflects the prices of the underlying providers without any platform markup. Vercel AI Gateway includes usage and cost tracking, automatic retries, failover, Zero Data Retention, and budgets per API key.
🔗 Tweet @SakanaAILabs · Vercel changelog
Briefs
-
Claude Code v2.1.195 (26 June) — 12 changes in this release: new
CLAUDE_CODE_DISABLE_MOUSE_CLICKSvariable to disable mouse interactions in full-screen mode, fix for hook matchers with identifiers containing hyphens (now exact matching), and fix for voice dictation for languages without spaces (Japanese, Chinese, Thai). 🔗 Release v2.1.195 -
GitHub Copilot API — total metrics by adoption phase (26 June) — The Copilot usage metrics API adds a
total_pull_requests_mergedfield by AI adoption cohort, in addition to the existingavg_pull_requests_merged. Enterprise administrators and organization owners can now measure the absolute throughput of merges by phase, available in 1-day and 28-day reports. 🔗 GitHub Changelog -
Genspark AgentBase — live session (26 June) — Genspark invites users to a YouTube live session to discover AgentBase, its new agent base feature, with live Q&A with the team. 🔗 Tweet @genspark_ai
What this means
Mobile coding agents are reaching a new level of maturity. The GA of Codex Remote turns experimental access into a supported product. One-to-one QR pairing and the DigitalOcean plugin signal a clear direction: developers should be able to launch and run their coding environments from anywhere, with a level of security equivalent to the desktop. Direct integration with cloud infrastructure (Droplet SSH ready in seconds) concretely reduces friction between idea and execution on mobile.
Multi-agent orchestration models are becoming accessible products. Fugu-Ultra illustrates a trend distinct from monolithic models: instead of a single expensive frontier model, an orchestrator dynamically chooses 1 to 3 models depending on the problem and aggregates their answers. Its integration into Vercel AI Gateway without platform markup lowers the entry barrier for teams already in the Vercel ecosystem. The announced performance (comparable to Claude Fable 5 on reasoning and science) positions Fugu-Ultra as a serious alternative for demanding tasks, provided the orchestration delivers at scale.