Good morning! Here's what's happening in AI today:
ChatGPT Sites lets you build and publish full projects
Claude Fable 5 access expands to more subscription plans
Google Gemini adds custom avatars for any scene or style
How to set up Hermes Agent the right way from day one
And 4 new AI tools worth trying today
AI BUILDING
OpenAI introduced ChatGPT Sites, letting anyone build and publish a working website directly from a conversation, no separate hosting setup or code editor required to get something live.
People are already using it to build simple games, launch a small business page, or publish research findings, all without writing a single line of code or configuring a server anywhere.
It also works well for lighter personal projects, like sharing a photography portfolio, setting up an automated out of office handoff, or spinning up a small hub for a team to coordinate around.
Every site built with ChatGPT Sites can be published instantly and shared with a live link, and OpenAI is showcasing real examples people have already built to demonstrate just how far the tool can stretch beyond a simple demo.
Building even a simple website has traditionally required at least some technical setup, from hosting to basic code to a developer willing to take on a small job. ChatGPT Sites removes that barrier entirely by turning a plain conversation directly into a published, shareable site. For anyone with an idea but no development background, this closes a gap that used to require hiring someone else just to get something online.
AI MODEL
Anthropic shared that Claude Fable 5 is rolling out more broadly across subscription plans starting this week, expanding beyond the original limited release as more server capacity comes online.
Claude Fable 5 will now be included in all Max and Team Premium plans at half of standard usage limits starting this week, giving a much larger group of subscribers direct access without needing to upgrade further.
Pro and Team Standard users keep their access to Fable through usage credits, with a one time credit added to their account so they can try the model without committing to a new plan first.
Anthropic said demand for Fable has been difficult to predict since launch, which is why access has rolled out in stages rather than all at once, a gradual approach meant to balance availability against reliability for everyone already using it.
Rolling out a high demand model in stages is a real tradeoff between availability and reliability, and Anthropic chose to expand access as capacity allowed rather than promise something it could not deliver at scale. This update means more subscribers can now try Fable 5 without hitting a wall immediately after signing up. It also reflects just how much genuine demand the model has generated since it first became available.
AI PRODUCTIVITY
Higgsfield introduced Free Mode, letting anyone brainstorm, write prompts, and plan out projects at no cost directly inside Higgsfield Supercomputer, without needing to spend credits just to think through an idea.
Free Mode removes the cost barrier from the early planning stage, so you can explore ideas, draft prompts, and map out a project before deciding whether it is worth spending credits to actually generate anything.
This is aimed at the part of the creative process that usually gets skipped over, the messy brainstorming phase, giving it a dedicated space instead of forcing you straight into paid generation.
Because it lives inside Higgsfield Supercomputer, anything you plan in Free Mode can move directly into a paid generation step without switching tools or losing the context you already built.
Most creative AI tools charge you the moment you start experimenting, which discourages the kind of loose, exploratory thinking that actually leads to good ideas. Free Mode gives that exploration its own space, free from the pressure of spending credits on something that might not work. For anyone who wants to plan before they commit, this changes how the earliest part of a project actually feels.
HOW TO AI
I've spent many years teaching students the difference between a tool that works once and a system that compounds. The pattern is always the same. Someone gets a good result on day one and assumes the hard part is done.
Most people who install an AI agent make exactly this mistake. They ask it something fun, get a decent answer, and assume everything's running correctly. A month later, nothing has "learned" anything.
That's because they tested the chat. They never tested the agent.
What makes Hermes different
Every chatbot runs the same loop: you type, it predicts a response, done. Hermes runs a process instead, assembling context, calling tools, checking results, before it ever responds. Two agents can look identical in a screenshot. The only way to tell them apart is to test what they actually do.
Step 1: Pick where it lives
Hermes Desktop is the right call for almost everyone starting out. It's a native app for both Mac and Windows, no server, no cloud account. Sessions save automatically, and the agent is available whenever your machine is on.
Step 2: Prove it actually remembers
This is the step almost everyone skips. Start a short conversation, fully close the app, then reopen it and resume that same conversation. If it's still there, the foundation holds. If it starts blank, nothing downstream will ever work.
Step 3: Test the tools, not the talk
A good conversational answer proves the model works. It says nothing about whether the agent works. Run three tests instead: ask it to run a command like checking disk usage, ask it to search the web for something current, and ask it to save a file, then confirm that file actually exists.
If all three come back with real output, the agent is functional. If one fails, fix it before doing anything else.
Where it lives, whether it remembers, and whether its tools actually work aren't the exciting part of setting this up. They're the only part that matters. Everything else this agent can eventually do only compounds if this part holds first.
P.S. You can access all my prompts, workflows and AI trainings if you upgrade here
Microsoft updated VS Code 1.129 with new agent workflows, letting you run terminal commands directly from chat and delegate complex tasks across sessions.
Anthropic kept Claude Code weekly limits 50% higher through August 19 for all Pro, Max, Team, and seat based Enterprise users.
Meta made Muse Spark 1.1 available on OpenRouter for US based developers to build with directly.
🖼️ Nano Banana 2: Turn a selfie into custom trading cards, murals, or cartoons.
🪄 Gemini Custom Avatars: Set up your avatar once, generate new scenes anytime.
🚀 Qwen3.8: Try a live preview of Alibaba's 2.4 trillion parameter model.
🛍️ NoimosAI Advisor: Analyzes your store's performance and drafts the next move.
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