The Roundtable Show’s Official Account
Joined June 2022
- Tweets45.7K
- Following6.3K
- Followers27K
- Likes224K
You can spot an AI-built website in two seconds: a small label above every heading, one colored word in the headline, an email signup at the bottom of every page.
Hallmark is a design skill from Together AI that refuses to build pages that look like that. 30.8K stars.
→ 48 visual themes, each with its own type, color, textures and components
→ builds the page around what visitors came to do: a booking page books, a shop sells
→ checks the result in a real browser against the most common AI-design tells
→ a fresh reviewer agent gets the final say before it ships
→ "/hallmark redesign" rebuilds an existing page and keeps its content
Install:
npx skills add nutlope/hallmark -g
Then in Claude Code: /hallmark build a landing page for my coffee shop
MIT.
github.com/Nutlope/hallmark
One study estimates AI put about 147,000 made-up citations into research papers in 2025 alone.
Academic Research Skills is a free Claude Code toolkit that helps you research and write a paper without inventing anything. 53.5K stars.
→ a 13-agent research team that finds sources and checks them against Semantic Scholar
→ a 12-agent writing team that learns your voice from your past papers
→ a 7-agent peer review panel, including a devil's advocate
→ checks citations, data and logic at every stage
→ stops for your approval at each step, so you stay the author
Install:
/plugin marketplace add Imbad0202/academic-research-skills
/plugin install academic-research-skills
Then type /ars-plan to plan your paper step by step.
Free for non-commercial use.
github.com/Imbad0202/academi…
Stop paying for CleanMyMac. This free tool does its job, plus three other paid apps. 72.4K stars.
Mole combines what CleanMyMac, AppCleaner, DaisyDisk and iStat Menus do, in one command:
→ mo clean: clears caches, logs and leftovers from apps you've already deleted
→ mo uninstall: removes apps along with every file they left behind
→ mo analyze: a visual map of what's eating your disk
→ mo status: live CPU, memory, disk and network monitor
→ mo purge: wipes old build folders from your coding projects
Add --dry-run to any command to see what it would remove before it removes anything.
Install:
brew install mole
mo
There's also a native Mac app and a Windows version at mole.fit.
GPL-3.0.
github.com/tw93/Mole
SAVE THIS: you scan AI skills before you install them. Now scan the ones you've already installed.
1. MEDUSA
Every API key you've ever pasted into Claude Code, Cursor or Copilot sits in plain text in your chat history. MEDUSA finds them across 21 key types and cleans them up.
pip install medusa-security
medusa secrets scan
2. Agent Scan by Snyk
Finds every agent, MCP server and skill on your machine (Claude Code and Desktop, Cursor, Gemini CLI, Windsurf) and checks them for prompt injection and vulnerabilities. Needs a free Snyk account token.
uvx snyk-agent-scan@latest
3. MCP Scanner by Cisco
Scans the MCP servers configured on your machine for malicious tools. No API key needed for the basic scan.
uv tool install --python 3.13 cisco-ai-mcp-scanner
mcp-scanner --scan-known-configs --analyzers yara --format summary
Five minutes, three commands.
Your AI-written posts have tells, and people can spot them: "It's not just X, it's Y." "Let that sink in." "Here's the thing."
Humanizer rewrites AI text so it sounds like a person wrote it, without changing what it says. 59.7K stars.
It's built on Wikipedia's "Signs of AI writing", the guide Wikipedia editors use to catch AI-generated text. It checks 26 patterns, including the five most common:
→ "Not X but Y"
→ one-line closers like "Let that sink in"
→ sayings that sound deep but say nothing
→ staged run-ups like "Here's the thing"
→ arguing with someone who never spoke
In a blind test, judges preferred its rewrite over the original AI text 16 times out of 16.
Claude Code:
/plugin marketplace add blader/humanizer
/plugin install humanizer@humanizer
Any other agent:
npx skills add blader/humanizer --global --agent '*'
It also works in Claude.ai: download the repo as a ZIP and upload it as a skill.
MIT.
github.com/blader/humanizer
You can now turn a simple text prompt into a complete AI-generated podcast
Google open-sourced NotebookLM-style audio generation tools, but Open Notebook takes it further by letting you create AI podcasts from your own documents and research
→ Turns PDFs, websites and notes into AI-generated audio → Creates conversational podcasts with AI hosts → Supports multiple AI model providers → Keeps your research organized in one workspace → Free and open-source
Get started in 3 steps:
* Download: Install Docker and get the Open Notebook docker-compose.yml
* Launch: Set an encryption key, then run docker compose up -d
* Create: Open the app, connect an AI provider and upload your sources
Turn hours of reading into a podcast you can listen to anywhere
Repo: github.com/lfnovo/open-noteb…
700+ AI animation prompts you can copy for free
This GitHub repo collects prompts for cinematic motion graphics, 3D scenes, animated explainers and interactive games built with Claude Opus 5.5
You’re probably using ChatGPT at 20% of what it’s capable of
Here’s a trick most people overlook: make it build the prompt for you
Paste this:
“I want to accomplish [goal]. Act as an expert prompt engineer and write the most effective prompt for this task. Include the right context, constraints, and output format. Ask me only what’s essential, then give me a ready-to-use prompt I can copy.”
Copy this, give it your goal, and let ChatGPT do the prompt engineering for you
5 GITHUB REPOS FOR DEBUGGING AI APPS:
1. Langfuse
github.com/langfuse/langfuse
2. Phoenix
github.com/Arize-ai/phoenix
3. Helicone
github.com/Helicone/helicone
4. OpenLLMetry
github.com/traceloop/openllm…
5. AgentOps
github.com/AgentOps-AI/agent…
If your AI keeps failing and you can’t tell why, start tracing everything
DeepSeek is exploding right now
Here’s a 2-hour course on how to use it and everything you need to know
Bookmark this and watch it
You can now turn a single photo into a fully explorable 3D world
Microsoft open-sourced TRELLIS, an AI model that generates detailed 3D objects from ordinary images
→ Turns 2D images into 3D assets → Generates textured meshes and 3D representations → Useful for game development, animation and AR/VR → Runs locally with a compatible NVIDIA GPU → Free and open-source
Get started in 3 steps:
* Clone: git clone github.com/microsoft/TRELLIS
* Install: Follow the README to set up dependencies and model weights
* Generate: Launch the demo and upload an image
What used to take hours of 3D modeling can now start with a single photo
Repo: github.com/microsoft/TRELLIS
REA has been absolutely exploding on GitHub, surpassing 350,000 stars
Get started in 3 easy steps:
* Install Node.js on your computer
* Run npx rea-agents setup
* Ask your AI agent to analyze software you own or have permission to inspect
See how software actually works, then use those insights to build your own
Repo: github.com/morluto/rea
10 AI x CRYPTO PROJECTS / TOOLS WORTH KNOWING (BOOKMARK THIS):
1. Bittensor
bittensor.com
→ decentralized network where AI models compete and earn TAO
2. Virtuals Protocol
virtuals.io
→ platform for launching tokenized AI agents
3. Fetch.ai / ASI
fetch.ai
→ autonomous agents for machine-to-machine transactions and coordination
4. Ritual
ritual.net
→ decentralized infrastructure for running AI models onchain
5. Grass
grass.io
→ turns unused internet bandwidth into data infrastructure for AI training
6. Gensyn
gensyn.ai
→ decentralized compute network for training AI models
7. Akash Network
akash.network
→ decentralized cloud compute that can be used for AI workloads
8. io.net
io.net
→ distributed GPU network built for AI and machine learning
9. Autonolas
olas.network
→ framework for autonomous agents that can operate across crypto protocols
10. Allora
allora.network
→ decentralized intelligence network for prediction and machine-learning models
AI needs compute, data, agents and incentives.
Crypto is starting to provide all four.
Your AI coding agent could become a scientific research assistant
Scientific Agent Skills is a free GitHub project with 177 specialized skills covering genomics, drug discovery, clinical research and more, plus access patterns for 100+ scientific databases
Stop relying on generic AI knowledge when you can give your agent actual scientific workflows
Repo: github.com/K-Dense-AI/scient…
5 GITHUB REPOS FOR RUNNING AI MODELS FASTER:
1. vLLM
github.com/vllm-project/vllm
2. TensorRT-LLM
github.com/NVIDIA/TensorRT-L…
3. llama.cpp
github.com/ggml-org/llama.cp…
4. SGLang
github.com/sgl-project/sglan…
5. DeepSpeed
github.com/microsoft/DeepSpe…
A smarter model is nice. A faster one people can actually use is better.
Your AI agents could finally learn from each other’s mistakes
Beacon captures sessions across Claude Code, Codex, Cursor and 20+ other tools, while Jev extracts lessons worth remembering
One agent makes a mistake, another learns from it
Your AI can help uncover forgotten subscriptions and wasted spending like a $150/hour financial consultant
Open ChatGPT or Claude, upload your bank transactions with sensitive details removed, and use these 7 prompts:
* Money Leaks: “Find my biggest unnecessary expenses”
* Subscriptions: “Spot recurring charges I might have forgotten”
* Hidden Fees: “Identify unusual fees and charges”
* Overspending: “Find categories where I spend too much”
* Bill Savings: “List bills I could negotiate or reduce”
* Budget Fix: “Build a realistic budget from my spending”
* Save $500: “Find ways to cut $500 from my monthly expenses”
Meta Platfroms $META dumped more than 5% despite posting strong earnings and surging 9% on the earnings news.
