👋 Hello hello,
Your different coding sessions in Claude Code can now message each other. That is, they can send each other targeted messages mid-task, like two teammates who've figured out how to talk.
OpenAI pushed a meaningful model upgrade for pretty much everyone, including free users. And a former employee who left OpenAI three months ago just went live with his own AI venture, and it’s worth a look.
🚨🚨 We’re hosting live sessions with Practicaly AI for Entrepreneurs members, where we will be building AI workflows from scratch, in real-time. If you haven’t attended any of these sessions, you are missing out! 🚨🚨
More details on this later. Okay, let’s get into it.
🔥🔥🔥 Three Curated AI Updates
If you've ever run two Claude Code sessions at once and had to manually sync what one session learned to the other, this update is for you.
Claude Code now supports cross-session messaging. Your sessions can reach out to each other and pass information without you having to copy-paste between terminals. The key thing to understand: it sends a text message (a summary), not your conversation history or files. Think of it less like syncing and more like one colleague calling another to say "hey, I just found something you need to know."
It works both ways. You can tell one session to ask another a question and get the answer back. Claude can also initiate this on its own — for example, if a change it just made would affect something another session is working on, it can warn that session automatically. The feature requires Claude Code v2.1.224 or later and runs on macOS and Linux. Available now.
For Plus and Pro users, GPT-5.6 Sol now powers both the Instant and deeper reasoning experiences in ChatGPT. Before this update, switching between quick and thoughtful responses could feel like you were talking to a different model. Now it's the same model, just thinking harder when you need it to. There's also a new slider that lets you dial how much thought goes into each response. A new internal evaluation found that factual errors in responses dropped significantly compared to the previous GPT-5.5 Instant.
For Free users, GPT-5.6 Luna is becoming the default model with unlimited text chats. There's also a new Think button for harder questions that need a bit more reasoning time. This is a real expansion of access, and free users getting unlimited text chats with a capable model is a bigger deal than the announcement made it sound.
Meta shipped a terminal coding agent that handles entire software engineering tasks end-to-end, from planning a change to writing the code to validating the result. It runs on Muse Spark 1.2, Meta's newest model. The interesting part is that the model and the agent were trained together, so they're designed to work as a unit rather than a general model plugged into a generic scaffold.
Multiple agents work in parallel on a task, with implementation running while background reviewers watch. In one internal test, the model optimized GPU kernels across more than 1,000 tool calls over 24 hours. This is currently in beta.
🔥🔥 Two Pro AI Tools
1.⚡ Energy
Energy is an AI workspace built around the idea that most knowledge work happens across a browser, an inbox, and a set of files — and you should be able to hand all of it off in plain language. You set up specialized assistants (inbox manager, sales coordinator, research scout) that connect to Gmail, Google Drive, Slack, and other tools you already use. Tell one what outcome you want, and it plans and executes the steps, then brings you the result to review.
It was built by a former OpenAI researcher who left three months ago to start this, and they claim users are doing days of work in hours. Worth checking out if you're drowning in coordination work.
2. 🌐 Context.dev
Context.dev is a web scraping API built specifically for AI agents and LLM pipelines. Here's the problem it solves: AI models only know what they were trained on, which goes stale fast. Context.dev lets you pull live web content — any URL converted to clean markdown, full-site crawls, structured data extraction, brand intelligence — and pipe it into your agent or workflow in real time.
It's especially useful if you're building AI tools that need to reason about current information, company data, or competitor pages. Backed by Y Combinator, free tier available, and teams report integrating in under 10 minutes.
🔥 One Pro Tip from Claude
If you're building with Claude Managed Agents (Anthropic's hosted infrastructure for running long-form AI agents), here are four updates that just shipped:
You can now set a session budget so your agent pauses when it hits a spending cap instead of quietly running up an unexpected bill. Sessions that hit the limit stop cleanly, and you raise the budget when you're ready to continue.
You can also control where inference runs. Set it to
usto keep all processing in-region (useful for compliance), orglobalto run wherever there's capacity at the standard rate. US-only is billed at 1.1x.If you already have skills stored in a
.claude/skills/folder in your repository for Claude Code, Managed Agent sessions now pick those up automatically at the start of each session. No extra setup needed.And the one that we think is most powerful: you can now give an agent an advisor. Add a stronger model as an advisor to your agent's roster, and the working agent can call it mid-session for a second opinion before making a complex decision. One extra line of configuration, and your agent has a smarter model in its corner when things get difficult.
🎤 Join Us Live This Friday
Are you looking to automate your business with AI? We’re hosting live build sessions with Practicaly AI for Entrepreneurs members every Friday.
Not a member? Join now to get 20% off one month
Join us live, build in real time, and get your questions answered on the spot.
Can’t make it live? Don’t worry, we’ll send out the recording after the session to everyone who RSVP’ed.
Event details:
📅When: Friday at 10AM PT/1PM ET
📍Where: Virtual, Google Meet
After you RSVP, you’ll be sent a link with the event details. Hope to see you there!
Did you find today's post useful?
Until next time,
Team @PracticalyAI
