Fazal Shah

Fazal Shah Agentic AI automation and Openclaw enthusiasts.

10/09/2026

iPhone Duo: polished foldable or seven years late?

Would you pay $1,999 for an iPhone that unfolds into an iPad-like screen? We look at the animation, multitasking, Pencil promise, visible crease and the Touch ID trade-off—with creator reactions and one expensive kidney joke 😅

What would make you switch from your current phone?

Footage/posts: Apple, , IShowSpeed via , , , , and AppleTrack. Native viewer screenshots supplied for commentary.

10/09/2026

GPT-6 Astra built me a 3D house tour. Then I wanted precise changes—and hit a wall.

Can AI replace 3D modelers and animators? Here are my actual Blender results, the limits I ran into, and why learning the fundamentals still matters.

What would you trust AI with today?

Full video, sources and social links: https://youtu.be/CBFvbqkIOiM
More practical AI tests: https://www.kodeking.net

Will Smith comparison clips are AI-generated. Community workflow claims are attributed in the video.
Subscribe animation: Advaitam E Learning — https://youtu.be/SuyZD4IwK7k

09/09/2026

GPT Image 2.5: photo editing and design examples

From outfit edits to posters and invitations, here’s a 59-second look at what the new model can produce. Watch the portrait edit closely for the details that stay consistent.

Follow for more practical AI demonstrations.

Examples: OpenAI + 12ui. Website shown is an image concept; animated prompt is illustrative.

08/09/2026

One prompt. Reference photos. A 3D mosque tour I can actually explore. 🕌

I gave Astra photos of Faisal Mosque and asked it to build an interactive browser experience. Watch the roof, minarets and click-to-navigate camera — then wait for the prayer hall.

This is an approximate AI reconstruction, not an exact architectural replica. The drone shot is real footage; the business clips are illustrations.

Which landmark should I try next?

More experiments: www.kodeking.net

Credits: Drone — Asif khan / Pexels. Architecture/property illustrations — Tima Miroshnichenko and Kampus Production / Pexels. Laptop photo — Jarkko Laine, adapted, CC BY 2.0: https://creativecommons.org/licenses/by/2.0/

08/09/2026

An F1-inspired car you can take apart. 🏎️

283 modeled elements across 11 systems, including small hardware. I built this original K26 concept with GPT‑6 Astra + Blender, then turned it into an interactive browser explorer.

Screen recording edited in HyperFrames. Sound effects only. Illustrative engineering study, not official team CAD.

Which object should I build next?

07/09/2026

Can Claude help build a game or visualize a house? These published Fable 5.1 demos show a playable kart racer and a Blender walkthrough.

Here is a 59-second look at the demos, reported efficiency gains and progress toward clearer writing. Results vary by task; savings depend on workload and caching.

What useful project would you try first?

Sources: Anthropic launch/model pages; demos by BridgeMind and Alex Albert; writing update by Boris Cherny. Demo footage belongs to the credited creators.
https://www.anthropic.com/claude-fable-and-mythos-5-1
More experiments: www.kodeking.net

🚨Breaking: An Anthropic engineer () just broke down how they actually use skills inside Claude Code — and it’s a complet...
18/03/2026

🚨Breaking: An Anthropic engineer (

) just broke down how they actually use skills inside Claude Code — and it’s a completely different mindset.

Here’s the real system 👇

Skills are NOT text files.

They are modular systems the agent can explore and execute.

Each skill can include:

reference knowledge (APIs, libraries)

executable scripts

datasets & queries

workflows & automation

→ The agent doesn’t just read… it uses them

The best teams don’t create random skills.

They design them into clear categories:

• Knowledge skills → teach APIs, CLIs, systems
• Verification skills → test flows, assert correctness
• Data skills → fetch, analyze, compare signals
• Automation skills → run repeatable workflows
• Scaffolding → generate structured code
• Review systems → enforce quality & standards
• CI/CD → deploy, monitor, rollback
• Runbooks → debug real production issues
• Infra ops → manage systems safely

→ Each skill has a single responsibility

The biggest unlock is verification

Most people stop at generation.
Top teams build systems that:

simulate real usage

run assertions

check logs & outputs

→ This is what makes agents reliable

Great skills are not static.

They evolve.

They capture:

edge cases

failures

“gotchas”

→ Every mistake becomes part of the system

Another thing most people miss:

Skills are folders, not files.

This allows:

progressive disclosure

structured context

better reasoning

→ The filesystem becomes part of the agent’s brain

And the biggest mistake?

Trying to control everything.

Rigid prompts.
Micromanagement.
Over-constraints.

Instead:

provide structure

give high-signal context

allow flexibility

→ Let the agent adapt to the problem

The best teams treat skills like internal products:

Reusable.
Composable.
Shareable across the org.

That’s how you scale agents.

Not with better prompts.

But with better systems.

Save this. This is how AI actually gets useful.

Manus just went local.1.5 million views on the announcement. Posted this morning."Today, we're taking Manus out of the c...
17/03/2026

Manus just went local.
1.5 million views on the announcement. Posted this morning.
"Today, we're taking Manus out of the cloud and putting it on your desktop."
They're calling it "My Computer."
Your AI agent... on your local machine... with access to your files, apps, terminal, and browser.
Sound familiar?
It should.
That's what OpenClaw has been doing for months.
But here's where it gets interesting.
Manus is closed source. Owned by Meta. Backed by billions.
OpenClaw is open source. Community-driven. Built by a lobster-obsessed Austrian.
And Manus just launched a feature that OpenClaw users have had since day one.
Let me break down what Manus "My Computer" actually does.
Your agent can access your local files and folders.
It can run terminal commands.
It can use your development tools.
It can operate your browser.
It can work while you're away and you can assign tasks from your phone.
Every single one of those things... OpenClaw already does.
But there are some key differences.
Manus requires approval for every terminal command.
You pick "Always Allow" or "Allow Once."
That's actually smart. Especially after what happened to Summer Yue.
OpenClaw gives you more freedom but also more risk. That's the tradeoff of open source.
Manus uses a hybrid model... cloud for complex tasks, local for simple or sensitive stuff.
OpenClaw lets you run fully local with Ollama or any model you want. Or fully cloud. Or both. Your choice.
Manus has a polished desktop app with a real UI.
OpenClaw runs through terminal, Discord, WhatsApp, Telegram, iMessage, Slack... basically everywhere.
Different approaches. Same destination.
Every major player is converging on the same conclusion.
The future of AI isn't a chatbot in a browser tab.
It's an agent on your computer doing real work.
Meta knows it. That's why they bought Manus for $2 billion and now it's going local.
OpenAI knows it. That's why they hired Peter Steinberger and OpenClaw ships daily.
Alibaba knows it. That's why they launched CoPaw.
And I've known it since December 2022 when I flew to the Philippines and started building AI Employees.
Here's the real question for builders.
Do you want a closed-source agent controlled by Meta?
Or an open-source agent you control?
Manus is polished. It's easy. It's backed by the biggest social media company on Earth.
But you don't own it.
You can't see the code.
You can't modify it.
You can't build skills for it.
You can't run it on your own infrastructure.
With OpenClaw and AI Persona OS... you own everything.
The code. The skills. The memory. The personality. The data.
That matters more than most people realize right now.
But they'll realize it soon.
The agent war is officially a five-way race.
OpenAI + OpenClaw.
Meta + Manus.
Alibaba + CoPaw.
Perplexity + Perplexity Computer.
Google + whatever they're cooking.
Perplexity just launched their own closed-source desktop agent too. I've been testing it for days. It's great.
But same problem as Manus.
You don't own it. You can't see the code. You can't build skills for it.
And the builders who understand the difference between "using someone else's agent" and "owning your own agent"...
Those are the ones who win....
Did you notice? Manus posted at 7:59 AM and has 1.5 million views already.
The market is hungry for this.
People want AI agents on their computers doing real work. The only question left is who you trust to build yours. 🦞

You can also follow on Instagram to see more.
16/03/2026

You can also follow on Instagram to see more.

Address

Islamabad
28100

Website

Alerts

Be the first to know and let us send you an email when Fazal Shah posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Contact The Business

Send a message to Fazal Shah:

Shortcuts

Share

Category