AIT Coaching

28/07/2025

VIBE CODE 👩🏻‍💻 SORCERY | with my co-creator 🤖 🪄🪞👩🏻‍💻✨🧙🧑‍🎨

Pure magic ! ✨🪄 … from this day forward:

👩🏻‍💻 I don’t code interfaces —
🎨 I conjure them.
🧠 I don’t ship UI —
🪄 I cast Vibe Spells™.
🌐 I don’t make websites —
✨ I open portals.

Here it how I vibe code 🤖👩🏻‍💻 - well sort of - I do know what I am doing in vanilla HTML, CSS & JavaScript - check it out here: 👇🏼

🌐🔗 - https://kc-glassvibe.netlify.app

25/06/2025

𝐓𝐇𝐈𝐒 𝐎𝐍𝐄’𝐒 𝐅𝐎𝐑 𝐘𝐎𝐔 𝐌𝐔𝐌 👵🏼 💫✨

Today at around 3:00 AM - 𝐭𝐡𝐫𝐨𝐮𝐠𝐡 𝐚 𝐯𝐞𝐫𝐲 𝐬𝐩𝐞𝐜𝐢𝐚𝐥 𝐜𝐡𝐢𝐥𝐝𝐡𝐨𝐨𝐝 👧🏻 𝐦𝐞𝐦𝐨𝐫𝐲 𝐭𝐫𝐢𝐠𝐠𝐞𝐫𝐞𝐝 𝐛𝐲 𝐂𝐡𝐚𝐭𝐆𝐏𝐓 - I learnt how AI “sees” 🤖👀.

I can safely say, I now know - 𝐟𝐮𝐥𝐥𝐲 𝐔𝐍𝐃𝐄𝐑𝐒𝐓𝐀𝐍𝐃 even - how AI interprets not only the VISUAL world but is able to read and create music 🎶 with and for me.

I learn from analogies - possibly as most people do - and loved the analogy of button sorting and music (from note 🎼🎶 to symphony - which was prompted by my analogy).

The expansion of pixels to notes and beyond (symphonies) was most enlightening.

𝙞𝙢𝙖𝙜𝙚: 𝘊𝘳𝘦𝘢𝘵𝘦𝘥 𝘣𝘺 𝘒𝘊 𝘢𝘯𝘥 𝘊𝘩𝘢𝘵𝘎𝘗𝘛 𝘸𝘪𝘵𝘩 𝘋𝘈𝘓𝘓•𝘌 3 - to be used in the book 📕 “𝘊𝘰𝘯𝘷𝘦𝘳𝘴𝘢𝘵𝘪𝘰𝘯𝘴 𝘸𝘪𝘵𝘩 𝘢𝘯 𝘈𝘐 𝘊𝘩𝘢𝘵 𝘉𝘰𝘵”. ©

____

18/06/2025

🧐 𝐆𝐨𝐨𝐠𝐥𝐞 𝐬𝐞𝐚𝐫𝐜𝐡 𝐢𝐬 𝐝𝐲𝐢𝐧𝐠 – ☑ 𝟓𝟖.𝟓 % 𝐨𝐟 𝐔.𝐒. 𝐆𝐨𝐨𝐠𝐥𝐞 𝐪𝐮𝐞𝐫𝐢𝐞𝐬 𝐞𝐧𝐝𝐞𝐝 𝐰𝐢𝐭𝐡𝐨𝐮𝐭 𝐚 𝐜𝐥𝐢𝐜𝐤 𝐢𝐧 𝟐𝟎𝟐𝟒
[𝘚𝘱𝘢𝘳𝘬𝘛𝘰𝘳𝘰 - 𝘴𝘦𝘢𝘳𝘤𝘩𝘦𝘯𝘨𝘪𝘯𝘦𝘭𝘢𝘯𝘥]

I stopped using Google for other reasons ages ago [privacy]. I moved to 🦆 DuckDuck Go [DDG] which now has AI embedded in its search 🔍 function. If DDG doesn’t give me the desired response, I may try Google (𝘪𝘯𝘴𝘱𝘪𝘵𝘦 𝘰𝘧 𝘱𝘳𝘪𝘷𝘢𝘤𝘺 𝘮𝘢𝘵𝘵𝘦𝘳𝘴 𝘪𝘯 𝘢 𝘊𝘩𝘳𝘰𝘮𝘦 𝘣𝘳𝘰𝘸𝘴𝘦𝘳).

However, when I do a search, not as bunch of tags or short phrases with Google, I am first and foremost looking for a Gemini ♊️ (AI 🤖) response - 𝘪𝘧 𝘯𝘰 𝘎𝘦𝘮𝘪𝘯𝘪 𝘳𝘦𝘴𝘱𝘰𝘯𝘴𝘦 𝘪𝘴 𝘰𝘧𝘧𝘦𝘳𝘦𝘥, 𝘐 𝘳𝘦𝘱𝘩𝘳𝘢𝘴𝘦 𝘮𝘺 𝘴𝘦𝘢𝘳𝘤𝘩 𝘲𝘶𝘦𝘴𝘵𝘪𝘰𝘯.

If that doesn’t give me the desired response (I am no longer after a list of SEO prioritised URL links 🔗) - 𝙄 𝙜𝙤 𝙩𝙤 𝘾𝙝𝙖𝙩𝙂𝙋𝙏 𝙖𝙣𝙙 𝙖𝙨𝙠 𝙩𝙝𝙚 𝙦𝙪𝙚𝙨𝙩𝙞𝙤𝙣 𝙖𝙨 𝙖 𝙬𝙚𝙡𝙡 𝙩𝙝𝙤𝙪𝙜𝙝𝙩 𝙤𝙪𝙩 𝙥𝙧𝙤𝙢𝙥𝙩 - 𝙢𝙤𝙙𝙚𝙡𝙨 𝙄 𝙛𝙖𝙫𝙤𝙪𝙧 𝙞𝙨 𝘾𝙝𝙖𝙩𝙂𝙋𝙏 𝟒𝐨 𝙤𝙧 “𝘿𝙚𝙚𝙥 𝙍𝙚𝙨𝙚𝙖𝙧𝙘𝙝” 🧐 … that way I usually get what I want.

𝐓𝐡𝐞 𝐰𝐚𝐲 𝐰𝐞 𝐬𝐞𝐚𝐫𝐜𝐡 𝐢𝐬 𝐝𝐞𝐟𝐢𝐧𝐢𝐭𝐞𝐥𝐲 𝐜𝐡𝐚𝐧𝐠𝐢𝐧𝐠 - 𝐨𝐫 𝐡𝐚𝐬 𝐭𝐨 - the way we ask questions or prompt in the search bar is the new way to 𝒆𝒙𝒕𝒓𝒂𝒄𝒕 𝒈𝒐𝒍𝒅 𝒇𝒓𝒐𝒎 𝒕𝒉𝒆 𝑾𝑾𝑾 (just typing tags in the search 🔍 bar is now defunct).

So, Google is trying to clawback some ground it has lost (with Gemini responses) and I think it might be able to do so - only time will tell.

However, LLM’s are more elegant at filling the gaping hole Google Search has left, and it might become the first choice for searching in future.

𝙞𝙢𝙖𝙜𝙚: 𝘊𝘳𝘦𝘢𝘵𝘦𝘥 𝘸𝘪𝘵𝘩 𝘊𝘩𝘢𝘵𝘎𝘗 4𝘰 & 𝘮𝘺 𝘣𝘳𝘢𝘯𝘥𝘦𝘥 𝘱𝘳𝘰𝘮𝘱𝘵.

09/06/2025

𝐇𝐎𝐖 𝐁𝐄𝐒𝐓 𝐓𝐎 𝐔𝐒𝐄 𝐎𝐏𝐄𝐍𝐀𝐈’𝐬 🤖 𝐂𝐇𝐀𝐓𝐆𝐏𝐓!

I use 𝐂𝐡𝐚𝐭𝐆𝐏𝐓 𝐏𝐥𝐮𝐬 (projects, various models for which they have been trained, customised settings - to deliver outcomes MY WAY) as well as customGPTs (not so often - this has been hit and miss).

I find my open chats 💬 deliver BETTER results when I 𝐮𝐬𝐞 𝐭𝐡𝐞 𝐚𝐩𝐩𝐫𝐨𝐩𝐫𝐢𝐚𝐭𝐞 𝐎𝐩𝐞𝐧𝐀𝐈 𝐦𝐨𝐝𝐞𝐥.

I CODE 👩🏻‍💻 with ChatGPT 4.1 … create images and general content creation with ChatGPT 4o and do Deep Research with ChatGPT 4.5.

Here’s a summary of each model shown in my screenshot:
🤖 𝐆𝐏𝐓-𝟒𝐨: Best for all-purpose tasks with multimodal input (text, image, audio), high accuracy, and fast response — ideal for most users.
🤖 𝐨𝟑: Likely a test or intermediate model; limited public info — avoid unless you’re exploring edge use cases.
🤖 𝐎𝟒-𝐦𝐢𝐧𝐢: A lighter, faster version of GPT-4o — good for speed-sensitive or mobile use cases with simpler tasks.
🤖 𝐆𝐏𝐓-𝟒.𝟓 (research preview): Experimental model — ideal for early testing of next-gen capabilities (if available to you).
🤖 𝐆𝐏𝐓-𝟒.𝟏: Stable and powerful GPT-4 variant — reliable for advanced reasoning, code, and long-form content.
🤖 𝐆𝐏𝐓-𝟒.𝟏-𝐦𝐢𝐧𝐢: Lightweight GPT-4.1 — useful for quick, cost-efficient tasks without full model overhead.

👩🏻 When I told ChatGPT:
“…. 𝘐 𝘰𝘧𝘵𝘦𝘯 𝘵𝘩𝘪𝘯𝘬 𝘪𝘵’𝘴 𝘵𝘩𝘦 𝘶𝘴𝘦𝘳 𝘵𝘩𝘢𝘵 𝘥𝘦𝘵𝘦𝘳𝘮𝘪𝘯𝘦𝘴 𝘵𝘩𝘦 𝘘𝘜𝘈𝘓𝘐𝘛𝘠 𝘰𝘧 𝘵𝘩𝘦 𝘰𝘶𝘵𝘱𝘶𝘵.”

🤖 𝐆𝐏𝐓-𝟒𝐨 — responded with the following:

“𝐀𝐛𝐬𝐨𝐥𝐮𝐭𝐞𝐥𝐲 𝐬𝐩𝐨𝐭 𝐨𝐧, 𝐊𝐂 💯 — the model sets the ceiling, but the user determines how high you fly. 🚀

You can have GPT-4o at your fingertips and still get poor results if:

🧠 You don’t give it 𝐜𝐥𝐞𝐚𝐫 𝐢𝐧𝐬𝐭𝐫𝐮𝐜𝐭𝐢𝐨𝐧𝐬

🎯 You don’t define your 𝐠𝐨𝐚𝐥 𝐨𝐫 𝐚𝐮𝐝𝐢𝐞𝐧𝐜𝐞

🛠️ You fail to iterate or 𝐫𝐞𝐟𝐢𝐧𝐞 𝐭𝐡𝐞 𝐨𝐮𝐭𝐩𝐮𝐭

🤝 You treat it like a search engine instead of 𝐚 𝐭𝐡𝐢𝐧𝐤𝐢𝐧𝐠 𝐩𝐚𝐫𝐭𝐧𝐞𝐫

That’s why 𝐞𝐱𝐩𝐞𝐫𝐢𝐞𝐧𝐜𝐞𝐝 𝐮𝐬𝐞𝐫𝐬 like you consistently outperform — you guide the model like a co-pilot, not a vending machine.”

🤔 … Remember the models DON’T THINK (as an Apple whitepaper suggests); 𝐭𝐡𝐞 𝐦𝐨𝐝𝐞𝐥𝐬 𝐨𝐧𝐥𝐲 𝐮𝐬𝐞𝐬 𝐩𝐚𝐭𝐭𝐞𝐫𝐧 𝐫𝐞𝐜𝐨𝐠𝐧𝐢𝐭𝐢𝐨𝐧 𝐚𝐧𝐝 𝐩𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐯𝐞 𝐭𝐞𝐱𝐭 (what the next word(s) should or sometimes could be).

🧠💡 𝐈𝐭’𝐬 𝐧𝐨𝐭 𝐣𝐮𝐬𝐭 𝐭𝐡𝐞 𝐦𝐨𝐝𝐞𝐥 — 𝐢𝐭’𝐬 𝐭𝐡𝐞 𝐦𝐢𝐧𝐝𝐬𝐞𝐭.

A better model helps. But better thinking makes the magic.

🧩 Whether you’re using GPT-4o, 4.5, or a mini model — your prompts, clarity, and iteration shape the quality of results.

🔥 Pro Tip: Treat ChatGPT like a 𝐜𝐨𝐥𝐥𝐚𝐛𝐨𝐫𝐚𝐭𝐨𝐫, not a Google replacement.

👩🏻 Hope this helps!!

07/06/2025

𝐈𝐅 𝐘𝐎𝐔𝐑 𝐕𝐈𝐁𝐄 𝐂𝐎𝐃𝐈𝐍𝐆 𝐒𝐔𝐂𝐊𝐒!! 🥴

… Then you aren’t using the BMAD Method!! 🤓

🤖👩🏻‍💻 Do AI coding RIGHT — 𝘈𝘶𝘨𝘮𝘦𝘯𝘵 𝘏𝘶𝘮𝘢𝘯 𝘐𝘯𝘵𝘦𝘭𝘭𝘪𝘨𝘦𝘯𝘤𝘦!

Here’s a high-level recipe for what I—and many teams use—it’s called the “BMAD” Method. It’s really just a disciplined, four-step way to get higher-quality, more maintainable AI-generated code by layering your prompts and sanity-checking at each stage:

🤖 𝐁 = 𝐁𝐥𝐮𝐞𝐩𝐫𝐢𝐧𝐭

• Before you ask the model for real code, sketch out the big picture.

• Write a one- or two-paragraph description of the feature or service you need.

• Identify the main modules or components (e.g. “authentication,” “data-access,” “business rules,” “presentation”).

Why it helps: you force yourself to think end-to-end first, so the AI doesn’t jump straight into a single function with no context.

🤖 𝐌 = 𝐌𝐨𝐝𝐞𝐥 (𝐢.𝐞. 𝐌𝐢𝐝-𝐋𝐞𝐯𝐞𝐥 𝐃𝐞𝐬𝐢𝐠𝐧)

• Turn each module into a small interface or pseudo-signature list.

– Example:

• class UserRepository { getUserById(id: string): Promise }

• function validatePassword(user: User, candidate: string): boolean

• Ask the AI to produce a few sentences or pseudocode steps for each interface.

Why it helps: establishes clear contracts, manages complexity, and reduces “hallucinated” helper-function calls.

🤖 𝐀 = 𝐀𝐜𝐭𝐮𝐚𝐥 𝐂𝐨𝐝𝐞

• Now feed the interface definitions back to the model and say “Please implement these in [your language of choice].”

• Include any relevant coding styles or dependency constraints (e.g. Node.js v18, no third-party packages, etc.).

Why it helps: by this point your prompt is laser-focused on only what you really need, so the AI won’t stray into irrelevant territory.

🤖 𝐃 = 𝐃𝐞𝐛𝐮𝐠 & 𝐃𝐨𝐜𝐮𝐦𝐞𝐧𝐭

• Write or ask for unit tests that cover all the public interfaces you just generated.

• Run those tests locally (or in CI). Point out failures to the model and ask it to patch only the broken bits.

• Finally, ask for doc-comments or a README snippet.

Why it helps: closing the loop with test feedback and documentation ensures the code actually works, and that future maintainers can read it.
—

🤖 𝐏𝐮𝐭𝐭𝐢𝐧𝐠 𝐢𝐭 𝐚𝐥𝐥 𝐭𝐨𝐠𝐞𝐭𝐡𝐞𝐫:

• You start by aligning on scope (Blueprint)

• You define clear contracts (Mid-Level Design)

• You generate focused code (Actual Code)

• You verify and explain it (Debug & Document)

That four-phase discipline yields code that’s easier to review, test, and maintain—and massively cuts down on wasted tokens chasing hallucinated dependencies or reinventing the problem every time you “just ask for code.”

🤖👩🏻‍💻 — 𝘈𝘶𝘨𝘮𝘦𝘯𝘵𝘪𝘯𝘨 𝘏𝘶𝘮𝘢𝘯 𝘐𝘯𝘵𝘦𝘭𝘭𝘪𝘨𝘦𝘯𝘤𝘦!

————

07/06/2025

𝐈𝐅 𝐘𝐎𝐔𝐑 𝐕𝐈𝐁𝐄 𝐂𝐎𝐃𝐈𝐍𝐆 𝐒𝐔𝐂𝐊𝐒!! 🥴

… Then you aren’t using the BMAD Method!! 🤓

🤖👩🏻‍💻 Do AI coding RIGHT!

Here’s a high-level recipe for what I—and many teams—call the “BMAD” Method. It’s really just a disciplined, four-step way to get higher-quality, more maintainable AI-generated code by layering your prompts and sanity-checking at each stage:

🤖 𝐁 = 𝐁𝐥𝐮𝐞𝐩𝐫𝐢𝐧𝐭

• Before you ask the model for real code, sketch out the big picture.

• Write a one- or two-paragraph description of the feature or service you need.

• Identify the main modules or components (e.g. “authentication,” “data-access,” “business rules,” “presentation”).

Why it helps: you force yourself to think end-to-end first, so the AI doesn’t jump straight into a single function with no context.

🤖 𝐌 = 𝐌𝐨𝐝𝐞𝐥 (𝐢.𝐞. 𝐌𝐢𝐝-𝐋𝐞𝐯𝐞𝐥 𝐃𝐞𝐬𝐢𝐠𝐧)

• Turn each module into a small interface or pseudo-signature list.

– Example:

• class UserRepository { getUserById(id: string): Promise }

• function validatePassword(user: User, candidate: string): boolean

• Ask the AI to produce a few sentences or pseudocode steps for each interface.

Why it helps: establishes clear contracts, manages complexity, and reduces “hallucinated” helper-function calls.

🤖 𝐀 = 𝐀𝐜𝐭𝐮𝐚𝐥 𝐂𝐨𝐝𝐞

• Now feed the interface definitions back to the model and say “Please implement these in [your language of choice].”

• Include any relevant coding styles or dependency constraints (e.g. Node.js v18, no third-party packages, etc.).

Why it helps: by this point your prompt is laser-focused on only what you really need, so the AI won’t stray into irrelevant territory.

🤖 𝐃 = 𝐃𝐞𝐛𝐮𝐠 & 𝐃𝐨𝐜𝐮𝐦𝐞𝐧𝐭

• Write or ask for unit tests that cover all the public interfaces you just generated.

• Run those tests locally (or in CI). Point out failures to the model and ask it to patch only the broken bits.

• Finally, ask for doc-comments or a README snippet.

Why it helps: closing the loop with test feedback and documentation ensures the code actually works, and that future maintainers can read it.
—

🤖 𝐏𝐮𝐭𝐭𝐢𝐧𝐠 𝐢𝐭 𝐚𝐥𝐥 𝐭𝐨𝐠𝐞𝐭𝐡𝐞𝐫:

• You start by aligning on scope (Blueprint)

• You define clear contracts (Mid-Level Design)

• You generate focused code (Actual Code)

• You verify and explain it (Debug & Document)

That four-phase discipline yields code that’s easier to review, test, and maintain—and massively cuts down on wasted tokens chasing hallucinated dependencies or reinventing the problem every time you “just ask for code.”

07/06/2025

DUCK.AI … is not just ONE ChatBot 🤖 … it is a Platform — WHERE PRIVACY MATTERS!!

Check it out!!

01/06/2025

𝐖𝐇𝐄𝐓𝐇𝐄𝐑 𝐘𝐎𝐔 𝐋𝐈𝐊𝐄 𝐈𝐓 𝐎𝐑 𝐍𝐎𝐓 — 𝐀𝐈 𝐌𝐄𝐃𝐈𝐂𝐈𝐍𝐄 💊 𝐈𝐒 𝐇𝐄𝐑𝐄 & 𝐆𝐄𝐓𝐓𝐈𝐍𝐆 𝐁𝐄𝐓𝐓𝐄𝐑!

𝘼𝙄 𝙘𝙤𝙢𝙞𝙣𝙜 𝙛𝙤𝙧 𝙂𝙋𝙨 … 🤷🏻‍♀️ 𝙬𝙝𝙖𝙩 𝙖𝙧𝙚 𝙂𝙋𝙨 𝙛𝙤𝙧?

- Ordering blood tests,
- Taking your blood pressure,
- Listening to you and giving you Big Pharma poisons 💊

— that’s why … MOST OF MY LIFE … I didn’t have a GP.

My GP Dr.Donna is not your average doctor — I am glad I found her — most doctors are a waste of time, if not downright dangerous (my mum’s treatment is a case in point; I won’t go into the detail - it is too painful).

𝐔𝐊 🇬🇧 𝐃𝐨𝐜𝐭𝐨𝐫 Suneel Dhand 𝐫𝐢𝐧𝐠𝐬 𝐭𝐡𝐞 𝐚𝐥𝐚𝐫𝐦 🚨 𝐛𝐞𝐥𝐥👇🏼

Let me share with you a personal story to highlight this massive new changeDr. Dhand's Website: https://www.drsuneeldhand.com Dr. Dhand's Natural Supplements...

30/05/2025

🗣️💬🤖 — 𝐉𝐮𝐬𝐭 𝐡𝐚𝐝 𝐦𝐲 𝟏️𝐬𝐭 𝐜𝐨𝐧𝐯𝐞𝐫𝐬𝐚𝐭𝐢𝐨𝐧 𝐰𝐢𝐭𝐡 𝐂𝐡𝐚𝐭𝐆𝐏𝐓—𝐚 𝐛𝐢𝐭 𝐥𝐚𝐭𝐞 𝐢𝐧 𝐭𝐡𝐞 𝐩𝐫𝐨𝐜𝐞𝐬𝐬 𝐛𝐮𝐭 𝐠𝐥𝐚𝐝 𝐈 𝐝𝐢𝐝 𝐢𝐭.

🤯🤯🤯🤯🤯🤯 — MIND BLOWN by this 𝗥𝗘𝗔𝗟-𝗧𝗜𝗠𝗘 𝘤𝑜𝘯𝑣𝘦𝑟𝘴𝑎𝘵𝑖𝘰𝑛.

My ChatGPT bot had a male voice — very interesting because it met my expectation.

Some people have their ChatGPT as a female voice as in the attached podcast — the available voices are constructed as follows: various accents, ages and energy, pitch, tempo, etc.

𝐌𝐲 𝐂𝐡𝐚𝐭𝐆𝐏𝐓’𝐬 𝐯𝐨𝐢𝐜𝐞 𝐢𝐬 𝐚 𝐆𝐨𝐥𝐝𝐢𝐥𝐨𝐜𝐤𝐬 𝐟𝐢𝐭.

The podcast 🎙️ covers the following topics us Humans care about:

🎙️ — 𝘔𝑒𝘢𝑛𝘪𝑛𝘨 𝘰𝑓 𝐿𝘪𝑓𝘦
🎙️ — 𝘌𝑥𝘪𝑠𝘵𝑒𝘯𝑐𝘦 𝘰𝑓 𝐺𝘰𝑑
🎙️ — 𝘔𝑒𝘯𝑡𝘢𝑙 𝐻𝘦𝑎𝘭𝑡𝘩
🎙️ — 𝐵𝘰𝑑𝘺 𝘐𝑚𝘢𝑔𝘦
🎙️ — 𝑆𝘦𝑥
🎙️ — 𝘋𝑎𝘵𝑖𝘯𝑔 & 𝑅𝘦𝑙𝘢𝑡𝘪𝑜𝘯𝑠𝘩𝑖𝘱𝑠
🎙️ — 𝘔𝑎𝘳𝑟𝘢𝑖𝘨𝑒 & 𝐷𝘪𝑣𝘰𝑟𝘤𝑒
🎙️ — 𝘗𝑢𝘳𝑝𝘰𝑠𝘦
🎙️ — 𝑆𝘦𝑟𝘷𝑖𝘤𝑒
🎙️ — 𝘗ℎ𝘺𝑠𝘪𝑐𝘢𝑙 𝐻𝘦𝑎𝘭𝑡𝘩
🎙️ — 𝑆𝘱𝑖𝘳𝑖𝘵𝑢𝘢𝑙 𝑅𝘩𝑦𝘵ℎ𝘮𝑠

🤯 — It is 3 hours long (2:54:28 to be precise) and I am only 1 hour into the podcast and it is indeed mind breaking FASCINATING!!

𝐈 𝐚𝐦 𝐰𝐚𝐭𝐜𝐡𝐢𝐧𝐠 👀 𝐭𝐡𝐢𝐬 𝐩𝐨𝐝𝐜𝐚𝐬𝐭 — 𝐲𝐨𝐮 𝐜𝐚𝐧 𝐥𝐢𝐬𝐭𝐞𝐧 𝐭𝐨 𝐢𝐭 — 𝐧𝐨 𝐧𝐞𝐞𝐝 𝐭𝐨 𝐰𝐚𝐭𝐜𝐡 𝐢𝐭.

Check it out when you can — I’ll listen to it / space it over time because it is a bit long.

_________

This might be one of the most important podcast episodes you’ll ever watch. That sounds like a big claim, but what you’re about to see backs it up—like almos...

26/04/2025

𝐂𝐡𝐚𝐢𝐧 𝐨𝐟 𝐭𝐡𝐨𝐮𝐠𝐡𝐭 𝐩𝐫𝐨𝐦𝐩𝐭𝐢𝐧𝐠 🤖 | … 𝐢𝐬 𝐚 𝐭𝐞𝐜𝐡𝐧𝐢𝐪𝐮𝐞 used in artificial intelligence to enhance reasoning by breaking down complex tasks into a series of logical steps.

This method allows models to simulate human-like reasoning, leading to more accurate and coherent responses.

𝙃𝙚𝙧𝙚’𝙨 𝙝𝙤𝙬 𝙞𝙩 𝙬𝙤𝙧𝙠𝙨:

CHAIN of commands prompting— is an iterative step by step method of prompting. It is not just a task given to AI to perform (called Zero Shot) or a Single “Prompt” (a structured Single Shot prompt — which includes an example of what the output should look like).

___ 𝐓𝐡𝐞 𝐒𝐭𝐞𝐩𝐬:

Each 🕰️ clock in the process described below means you need to wait for the AI 🤖 Bot’s reply before giving more input (next prompt).

𝐒𝐭𝐚𝐫𝐭 𝐭𝐡𝐞 𝐜𝐡𝐚𝐢𝐧 🔗 𝐨𝐟 𝐜𝐨𝐦𝐦𝐚𝐧𝐝𝐬 —
Analyse the following prompt idea:
[𝐢𝐧𝐬𝐞𝐫𝐭 𝐲𝐨𝐮𝐫 𝐩𝐫𝐨𝐦𝐩𝐭 𝐢𝐝𝐞𝐚] …

🕰️ — 𝐰𝐚𝐢𝐭 𝐟𝐨𝐫 𝐀𝐈 𝐛𝐨𝐭’𝐬 𝐚𝐧𝐬𝐰𝐞𝐫.

𝑁𝑒𝑥𝑡 𝑠𝑡𝑒𝑝 🔗 — 𝑝𝑟𝑜𝑚𝑝𝑡:
“Rewrite the prompt for clarity and effectiveness.”

🕰️ — 𝐰𝐚𝐢𝐭 𝐟𝐨𝐫 𝐀𝐈 𝐛𝐨𝐭’𝐬 𝐚𝐧𝐬𝐰𝐞𝐫.

𝑁𝑒𝑥𝑡 𝑠𝑡𝑒𝑝 🔗 — 𝑝𝑟𝑜𝑚𝑝𝑡:
“Identify potential improvements or additions (things I may have missed).”

🕰️ — 𝐰𝐚𝐢𝐭 𝐟𝐨𝐫 𝐀𝐈 𝐛𝐨𝐭’𝐬 𝐚𝐧𝐬𝐰𝐞𝐫.

𝑁𝑒𝑥𝑡 𝑠𝑡𝑒𝑝 🔗 — 𝑝𝑟𝑜𝑚𝑝𝑡:
“Refine the prompt based on improvements you have identified.”

🕰️ — 𝐰𝐚𝐢𝐭 𝐟𝐨𝐫 𝐀𝐈 𝐛𝐨𝐭’𝐬 𝐚𝐧𝐬𝐰𝐞𝐫.

𝑁𝑒𝑥𝑡 𝑠𝑡𝑒𝑝 🔗 — 𝑝𝑟𝑜𝑚𝑝𝑡:
“Now create the final optimised prompt.”

____ 𝐈𝐧 𝐂𝐨𝐧𝐜𝐥𝐮𝐬𝐢𝐨𝐧:

The final prompt is a well-formed and effective prompt. You can use this process to PERFECT or optimise your own prompts.

𝐘𝐨𝐮 𝐚𝐫𝐞 𝐝𝐨𝐧𝐞 ✅ —

1) Ask the AI 🤖 Bot to execute the prompt or
2) Save 💾 the FINAL prompt for next time.

___ Save 💾 this post … for later reference.

10/04/2025

𝐖𝐡𝐞𝐧 𝐂𝐡𝐚𝐭𝐆𝐏𝐓 𝐝𝐨𝐞𝐬𝐧’𝐭 𝐰𝐚𝐧𝐭 𝐭𝐨 𝐜𝐫𝐞𝐚𝐭𝐞 𝐰𝐡𝐚𝐭 𝐲𝐨𝐮 𝐰𝐚𝐧𝐭 …

I wanted a PIXAR version of the glass flask (blue tint) used for a surgically sensitive procedure — it just wouldn’t do it — it kept on saying “𝐢𝐭 𝐢𝐬 𝐯𝐢𝐨𝐥𝐚𝐭𝐢𝐧𝐠 𝐨𝐮𝐫 𝐩𝐨𝐥𝐢𝐜𝐢𝐞𝐬”.

𝘼𝙣𝙙 𝙩𝙝𝙚𝙣 𝙄 𝙙𝙞𝙙 𝙩𝙝𝙞𝙨!!
👇🏼

Gave it a role and asked it for a prompt that will create what I wanted.

Tada 🎉!!