Agents on your mobile
Ben's Bites
Opinion — commentary, not a factual news event.
Claude Code now works better on your phone, and ChatGPT can hook into Apple Messages. The real race is making agents useful when you're away from your laptop.
Based on reporting by Ben's Bites — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
The mobile-agent push just got a little more real. Anthropic updated Remote Control in Claude Code so you can start a new session from your phone, recover dropped connections between phone and laptop, and get faster session loading on iOS. That’s not a flashy demo. It’s the sort of fix that turns a tool from “nice on paper” into something people might actually reach for when they’re not sitting at a desk.
OpenAI is nudging in the same direction. ChatGPT can now connect to Apple Messages, search conversations, catch you up, and draft or send replies from a Mac. The feature sits inside ChatGPT Work and Codex on desktop, which keeps the focus on work use rather than a casual consumer stunt. And that matters, because messaging is where a lot of real-life coordination happens.
The rest of the roundup reads like a company-by-company effort to make agents more practical and a little less theatrical. OpenAI says GPT-Image-2 can now generate transparent backgrounds in the API, and GPT-5.6-Sol is 20% cheaper at $4/$24 per 1M input/output tokens. Grok Bot is rolling out to more users too, with access for all SuperGrok Plus, Cursor Pro+, and Cursor Teams subscribers, plus a free week-long trial.
Anthropic also released Claude Academy, a free set of courses, tutorials, and job-specific guides. Deepgram added Flux TTS for live conversation, with response times as low as 80ms and the ability to carry context across turns and handle interruptions. Taken together, the message is plain: the companies aren’t just trying to make agents smarter. They’re trying to make them fit into the boring, messy places where people actually work.
The feed underneath all this is even more telling: agents in Slack, email, iMessage, CLI, design tools, Stripe data, and website workflows. That’s the real story. Everyone’s trying to make the bot feel less like a demo and more like a colleague with a badge.
My take — AI-written commentary, not fact-checked reporting
This is the part of AI that actually matters: not bigger fireworks, but better access, better controls, and fewer excuses. The closed-model crowd loves to sell magic; the open side keeps pushing for usable tools people can wire into real workflows. Boring wins, and the browser-tab circus can wait.
Read more about this at: Ben's Bites