Graham_space

Just scribing and typing.

[Graham_space @Graham__space]

🚀 Just released TOON Context MCP Server!

Transparent token optimization for coding agents - converts JSON/CSV to 30-60% more efficient format automatically. No workflow changes needed, just 50% less tokens! 🧵 @kilocode @cline @GithubProjects

#AI #MCP #Tokens #Coding

[Akshay 🚀 @akshay_pachaar]

A simple trick cuts your LLM costs by 50%!

Just stop using JSON and use this instead:

TOON (Token-Oriented Object Notation) slashes your LLM token usage in half while keeping data perfectly readable.

Here's why it works:

TOON's sweet spot: uniform arrays with consistent fields per row. It merges YAML's indentation and CSV's tabular structure, optimized for minimal tokens.

Look at the example below.

JSON:

{
"𝘂𝘀𝗲𝗿𝘀": [
{ "𝗶𝗱": 𝟭, "𝗻𝗮𝗺𝗲": "𝗔𝗹𝗶𝗰𝗲", "𝗿𝗼𝗹𝗲": "𝗮𝗱𝗺𝗶𝗻" },
{ "𝗶𝗱": 𝟮, "𝗻𝗮𝗺𝗲": "𝗕𝗼𝗯", "𝗿𝗼𝗹𝗲": "𝘂𝘀𝗲𝗿" }
]
}

Toon:

𝘂𝘀𝗲𝗿𝘀[𝟮]{𝗶𝗱,𝗻𝗮𝗺𝗲,𝗿𝗼𝗹𝗲}:
𝟭,𝗔𝗹𝗶𝗰𝗲,𝗮𝗱𝗺𝗶𝗻
𝟮,𝗕𝗼𝗯,𝘂𝘀𝗲����

It's obvious how few tokens are being used to represent the same information!

To summarise, here are the key features:

💸 30–60% fewer tokens than JSON
🔄 Borrows the best from YAML & CSV
🤿 Built-in validation with explicit lengths & fields
🍱 Minimal syntax (no redundant braces, brackets, etc.)

IMPORTANT!!

That said, for deeply nested or non-uniform data, JSON might be more efficient.

In the next tweet, I've shared some benchmark results demonstrating the effectiveness of this technique in reducing the number of tokens and improving retrieval accuracy with popular LLM providers.

Where do you think this could be effective in your existing workflows?

Find the relevant links in the next tweet!

[Graham_space @Graham__space]

@llamaonthebrink I have an idea how to fix the liquidity limitations if you wanna chat?

[Graham_space @Graham__space]

@ctodotnew @convex Who cares

[Graham_space @Graham__space]

Master the basics! 💡 New post on Python Data Types!

Everything you need to know about:

Integers, Strings, Booleans.

Mutable vs. Immutable objects.

Getting the variable type with type().

Essential reading for every #Python developer. Check it out!

Link below👇:

#programming #tutorial #beginner

[Graham_space @Graham__space]

✨ NEW ARTICLE ALERT! ✨

Tired of only using NotebookLM for your AI note-taking? 🧠 The landscape is exploding with incredible alternatives!

My new deep-dive reveals the Top AI Tools for Note-Taking in 2025, going beyond NotebookLM to explore tools like:

🍎 Elephas (Privacy & Mac-friendly)
🔗 Obsidian (Ultimate customization with AI plugins)
🚀 Notion (All-in-one workspace with integrated AI)
🔒 Self-Hosting options for maximum privacy!

Find the perfect NotebookLM alternative to supercharge your productivity and knowledge management! Which one will you choose? Link below:👇

#AINoteTaking #NotebookLM #AItools #Productivity #KnowledgeManagement #Obsidian #NotionAI #TechNews #Elephas

[Graham_space @Graham__space]

https://t.co/Z56OQpr6tu
Stars, forks, and shares are pure rocket fuel ❤️ Pull requests for new icons/themes very welcome! #LinktreeAlternative #OpenSource #IndieDev #WebDev #NoCode

[Graham_space @Graham__space]

🚀 Just open-sourced **Linktree Builder** – a completely free, no-BS alternative to Linktree that runs 100% in your browser! No accounts, no servers, no tracking, no monthly fees. link below: 👇

✅ Real-time preview while you edit
✅ 50+ built-in platform icons & colors (Twitter/X, Instagram, TikTok, GitHub, Spotify, Patreon, Ko-fi, and tons more)
✅ Drag-to-reorder, custom bio, emoji/avatar ✅ Auto-save + export as a single standalone HTML file
✅ Deploy instantly to GitHub Pages, Cloudflare Pages, Netlify, Vercel… anywhere One file. Zero dependencies. Works offline.

Download the HTML, open it, build your page in minutes, export → done.

[Graham_space @Graham__space]

I'll do ya one better, you can literally run test scripts yourself in my toon MCP server: https://t.co/BEeA6KRNSp

[Theo - t3.gg @theo]

Crazy how much more informed you can be by just reading the README file on the repo.

Armin is wrong here, the training data doesn't matter. TOON makes most models more accurate for lookup.

[Graham_space @Graham__space]

@maxtumin @saltyAom Check out my MCP for the test scripts

[Graham_space @Graham__space]

@natsukashixo @peach2k2 Exactly, mcps like mine that use toon or convert it to another human-readable format, it's the best way to optimize data input for now. Deepseek is working on just using images for context tho.

[Graham_space @Graham__space]

@natsukashixo @peach2k2 The issue is, csv and normal json can outperform it in these baby examples. It's probably just better to use them.

[Graham_space @Graham__space]

@peach2k2 Prime example aka me:

https://t.co/isPClva4JQ

[Graham_space @Graham__space]

🚀 Just released TOON Context MCP Server!

Transparent token optimization for coding agents - converts JSON/CSV to 30-60% more efficient format automatically. No workflow changes needed, just 50% less tokens! 🧵 @kilocode @cline @GithubProjects

#AI #MCP #Tokens #Coding

[Graham_space @Graham__space]

🚀 Just released TOON Context MCP Server!

Transparent token optimization for coding agents - converts JSON/CSV to 30-60% more efficient format automatically. No workflow changes needed, just 50% less tokens! 🧵 @FactoryAI @cline @claudeai @AnthropicAI

#AI #MCP #Tokens #Coding

[Graham_space @Graham__space]

@DataChaz dont forget my TOON mcp server: https://t.co/isPClva4JQ

[Graham_space @Graham__space]

🚀 Just released TOON Context MCP Server!

Transparent token optimization for coding agents - converts JSON/CSV to 30-60% more efficient format automatically. No workflow changes needed, just 50% less tokens! 🧵 @kilocode @cline @GithubProjects

#AI #MCP #Tokens #Coding

[Graham_space @Graham__space]

@colinhacks He forgot to ask Claude for help😂

[Graham_space @Graham__space]

Bold prediction for 2026:

JavaScript will still be "dying" Python will still be "too slow" Rust will still be "the future" And PHP will still be running 70% of the internet

Tech Twitter predictions are entertainment, not strategy.

Build with what works. Ship fast. Make money.

The tech stack wars are won by those who ignore them.

[Graham_space @Graham__space]

I'm doing a case study interview for @ctodotnew on Monday, have anything interesting I should mention from your experience?

[Graham_space @Graham__space]

@AnthropicAI Well, if you're a government employee in Maryland....might wanna start sending out resumes.

[Graham_space @Graham__space]

@iyoushetwt "don't make mistakes"

[Graham_space @Graham__space]

@AnthropicAI This is why it's kinda silly when people glaze Chinese models so much, the reason why they can only spend 6 million on training (they actually just fine tune) is because I'm sure the intellectual theft rate is extremely high like other sectors.