Top 10 Underrated Claude Tips

Most people have no clue that Anthropic recently shipped this.

Live Artifacts allow Claude to build and maintain persistent, interactive HTML dashboards that stay connected to your apps.

Unlike older static artifacts, they sit in your sidebar, remember their history, and automatically pull fresh data every time you open them.

Perfect for building personal apps within Claude that require live data.

To start using Live Artifacts, open Claude Cowork, and say: “I want to build a live artifact for [x].”

Use this for creating a personal OS, fitness/health app & more (simple apps inside Claude).

#9

MCPs are the easiest way to unlock Claude and instantly boost your productivity.

Think of MCP servers as “bridges” that let you bring your favorite apps directly into Claude.

Prompt your Slack channels, Notion databases, create Figma designs & more - all through natural language prompting.

To get started with connecting your apps in Claude:

Customize → Connectors → Start connecting your most-used apps

Our essential connectors: Notion, Excalidraw, Google Calendar, Slack, Google Drive.

We haven’t seen any other creator talk about this.

You should create “custom” slash commands in Claude.

Example: Anytime you type “/grammar,” Claude instantly conducts a grammar check.

This is essentially a way to build skills in Claude without committing to building a full skill.

Good for simple workflows you often do.

To get started, you can simply tell Claude: “Anytime I say /[x], do [x].”

Especially useful in long threads.

This is a recently shipped Claude CLI feature for the more advanced users.

Dynamic workflows allow Claude to work through a problem step by step inside a single conversation with a verification layer.

It writes a plan as a script, spins up tens to hundreds of parallel agents to execute different parts of that plan simultaneously, and has those agents check each other's work (verification).

Getting started

Type /workflow inside your Claude Code CLI.

Claude Design is hands-down the most powerful AI design tool on the market.

Anthropic recently doubled usage limits, allowing you to create more designs.

We highly recommend using Design for your websites, apps, and anything that needs a pretty UI/UX.

You can access Claude Design here:

This is a tip Miles has talked about many times on the AI Edge YouTube channel.

In Claude Cowork, you can attach your AI agents to locally stored desktop folders.

One of the best use cases for this is having Claude update markdown files on your behalf.

Useful for creating: session logs, personal memory databases, and you can even use this to create a custom notes storage app with Claude.

In Cowork, you’ll want to attach a brand new desktop folder and send prompts like:

“Create a new .MD that logs this entire session.”

“Create a .MD where I send you personal notes.”

“Update my .MD file with all my preferences.”

An excellent way to locally store your context and AI data.

Everyone should be using this.

Learn Mode in Claude lets your AI agents teach you any concept.

Think of this like NotebookLM (Claude version).

This becomes 10x more powerful when you pair it with custom Claude Skills and MCPs that can source from your most-used apps.

Two weeks ago, Anthropic made it possible to swap between reasoning efforts in real time across the Claude app.

We recommend using “High” as your default, and only resorting to Max + Thinking for your most important tasks.

In our experience, “Low" and “Medium” efforts can cause more hallucinations, so only use them for basic tasks.

This is an obvious tip that most users are ignoring.

Dictation is significantly better inside of Claude than typing.

This is how you can voice-dump your entire thoughts and interact with Claude in real time.

To make Dictation more powerful, attach the “WhisprFlow” connection, which optimizes your voice prompts in real-time

Ok, now this is powerful.

Ultracode is a Claude Code setting that combines xhigh reasoning effort with automatic workflow orchestration.

With it on, Claude plans and executes entire tasks instead of waiting for your approval.

As expected, this does burn through tokens, though.

To enable it, run "/effort ultracode" in Claude Code CLI.

Use them to start boosting your AI productivity now.

It was a huge week in AI, so you won’t want to miss this!

The US government issued an export control directive on June 12, ordering Anthropic to block Fable 5 and Mythos 5 for all foreign nationals, including foreign employees working in the United States.

Anthropic said it could not apply the restriction selectively and disabled Fable 5 for all users.

Watch Miles’ emergency update here

SpaceX began trading on Nasdaq on June 12 under the ticker $SPCX after pricing its IPO at $135 per share.

The stock opened at $150, an 11% first trade gain, while the company raised $75B at an implied valuation of about $1.77T.

The stock is currently pumping - trading at $200+/share (at time of writing).

Yesterday morning, SpaceX officially acquired Cursor in an all-cash deal valued at $60B.

Cursor Mobile was also just announced.

OpenRouter recently announced the Fusion API, which achieves Fable-level intelligence at half the price.

How it works:

https://x.com/OpenRouter/status/2065856853989270011?s=20

zAI just launched GLM-5.2 - an open-weight model built to achieve frontier-level coding intelligence.

Chat with it here:

Anthropic says it is working to restore Fable 5.

The next government decision will show whether the shutdown becomes a one-off dispute, a permanent restriction on Mythos-class systems, or a template for controlling future model releases.

We expect Mythos to be back online within the next few weeks.

Fable 5 is banned, but the ENTIRE system prompt is leaked.

You can plug the exact Fable 5 system prompt .md into Opus, GPT-5.5, or any other LLMs.

Note: You cannot get a 1-1 Fable, but you can tell other models to act like it

Steal the system prompt here:

https://github.com/elder-plinius/CL4R1T4S/blob/main/ANTHROPIC/CLAUDE-FABLE-5.md

A new Cursor model was just teased.

It’s rumored to be:

We expect this new model to be released within the coming weeks.