Good morning! Here's what's happening in AI today:
Claude extends Fable 5 access on all paid plans through July 19
Google AI Studio releases custom URLs for your deployed apps
Cursor launches Side Chats to explore ideas without losing your thread
How to use GPT-5.6 Sol inside Claude Code
And 4 new AI tools worth trying today
AI MODEL
Anthropic extended Claude Fable 5 access on all paid plans and kept Claude Code's weekly rate limits 50% higher, both running through July 19.
Fable 5 remains available on every paid plan through July 19, giving subscribers more time before any access changes take effect.
Claude Code's weekly rate limits stay 50% higher through the same window, so heavier coding workflows keep more room to run.
If you are already on a paid plan, nothing changes on your end. Fable 5 and the higher limits continue working exactly as they have been.
This extension gives paid subscribers extra runway to build workflows and test use cases before any access adjustments happen.
Fable 5 is the most capable model on your plan right now, and this extension means you keep full access without interruption. If you have not explored what Fable 5 can do yet, or you have been running into rate limits in Claude Code, this window just got longer. Use the extra time deliberately.
AI TOOL
Google AI Studio introduced custom URLs, giving deployed apps a permanent, personalized home with a clean AI Studio domain.
Any app you deploy in AI Studio can now get its own memorable, permanent URL instead of a generic, auto-generated link.
A custom domain makes it easier to share your app with others without a long, forgettable address attached to it.
This applies to any app built and deployed inside AI Studio, so the update covers your existing projects as well as new ones.
The feature is live now, meaning you can set a custom URL for a deployed app today without waiting on a rollout.
A generic auto-generated link makes even a great app look unfinished. Custom URLs give your projects a real identity, something you can actually put on a business card or share with confidence. For anyone building and shipping apps regularly in AI Studio, this is a small feature that changes how finished your work feels.
AI TOOL
Cursor introduced Side Chats, a new way to ask questions and explore ideas without interrupting your main conversation.
Each side chat is a separate, durable agent conversation that runs alongside your main thread without disrupting it.
You can explore a tangent, ask a quick question, or test an idea in a side chat while your primary conversation stays untouched.
Use an @-mention to bring context from a side chat back into your main thread whenever it becomes relevant.
Side chats persist over time, so you can return to one later without losing the thread of what you were exploring.
The biggest cost of asking a quick question mid-task is usually losing your place in the main conversation. Side Chats solve that by giving tangents their own space that you can pull back in only when it matters. For anyone running long, complex sessions in Cursor, this keeps your main thread clean while still letting you explore freely.
HOW TO AI
GPT-5.6 Sol is genuinely impressive. It is also, for a lot of people, a lesson in how fast a quota can disappear.
The problem is not Sol. It is where most people are running it.
What the 5.6 family actually is
GPT-5.6 is three models, not one. Luna handles everyday coding, fast and cheap. Terra steps up for bigger features and repo-wide changes. Sol handles planning, architecture, and final review, the judgment layer you bring in before anything gets built. This hierarchy only works if something enforces it.
The problem inside Codex
When Sol spawns a subagent inside Codex, that subagent inherits the same model and the same effort level as the parent. Run Sol at Ultra and every subagent runs at Ultra too. This is a documented open issue with no fix yet, and it is the fastest way to burn through an entire session on a single task.
What Claude Code does differently
Claude Code lets you set the model and effort level for each subagent independently. Sol can orchestrate a Terra subagent for a feature build and a Luna subagent for a quick check, each at its own tier. The expensive model stays at the top making judgment calls. The cheaper models handle the volume. This is the harness difference: Sol is not smarter inside Claude Code, it is finally working the way it was designed to.
The orchestration setup that works
Fable 5 at the top, reading the goal and reviewing final output. Sol in the middle, handling planning and architecture at high effort, not Ultra. Terra and Luna at the bottom, doing the actual implementation. Fable reviewing Sol's output catches what Sol might miss, since different model families have different blind spots.
Effort levels that actually work
Avoid Ultra entirely for now, it is a parallel multi-agent mode with documented bugs. Use Luna at high effort for most everyday coding, Terra at medium for feature builds and high for repo-wide changes, and Sol at high effort for planning and review.
Add this to your AGENTS.md to control subagent spawning:
Only spawn subagents when I explicitly ask you to. When spawning subagents, use a maximum of 1-3 agents. Ask before spawning additional agents beyond the initial set.
Use clear stop points in your prompts so you can check work before committing to the next stage:
Build this feature. Start by writing a plan. When the plan is complete, stop and ask for my feedback before proceeding.
Two ways to set it up
If you are on Codex Pro, install OpenAI's official Codex plugin for Claude Code. Three commands, and you get /codex:review, /codex:adversarial-review, and /codex:rescue right inside Claude Code.
If you are already in Claude Code, set up a proxy so Sol runs behind the Claude Code harness while keeping its agents, tools, and interface intact. Type a custom alias instead of the standard command to start a Sol-powered session.
The models are good enough to use from any harness. The harness determines whether you control the workflow or the workflow controls you.
Claude Code launched an in-app browser on desktop, letting Claude pull up docs, designs, or any site and interact with it the same way it works with your local dev servers.
Meta introduced Muse Spark 1.1, a significant upgrade now available in Thinking mode in the Meta AI app, alongside a public preview of the new Meta Model API.
Nous Research updated the Hermes Desktop app so it can now discover and connect to your Hermes Cloud agents automatically after signing in with Nous Portal.
🤖 Claude: Access Fable 5 on any paid plan through July 19.
🔗 Google AI Studio: Give your deployed apps a permanent, custom URL.
💬 Cursor: Explore ideas in Side Chats without losing your main thread.
☁️ Nous Research: Auto-discover your Hermes Cloud agents from the desktop app.
THAT’S IT FOR TODAY
Thanks for making it to the end! I put my heart into every email I send, I hope you are enjoying it. Let me know your thoughts so I can make the next one even better!
See you tomorrow :)
- Dr. Alvaro Cintas
✓ Full archive of premium guides with ready-to-use prompts
✓ Structured AI courses (step-by-step, start-to-finish)
✓ Every upcoming premium tutorial