There’s a fear running through every industry right now: AI is coming for my job.

Writers, designers, coders—they’re all worried about ChatGPT and Claude. Especially when a new model drops and demonstrates capabilities that seemed impossible six months ago. Each release ratchets up the anxiety: How long until this thing can do what I do?

But after a year of building heavily with these tools, I think this framing misses the point entirely.

AI isn’t replacing jobs. It’s replacing how jobs get done.

And the skill that’s emerging as universal? AI Engineering.

“The Hottest New Programming Language Is English”

Andrej Karpathy—former Tesla AI Director, OpenAI founding member—put it perfectly:

“The hottest new programming language is English.”

This isn’t a joke. It’s the new reality.

The ability to clearly articulate intent, break down problems, and iterate on AI outputs is becoming the core competency across every role. Whether you’re deploying infrastructure or designing a logo, the workflow is the same: define intent, prompt the system, validate the output, iterate.

That’s engineering. And everyone is doing it now.

What “AI Engineering” Actually Looks Like

Let me make this concrete with examples from my own work.

Cloud & DevOps → Prompt Engineering

In The Speed of AI-Assisted Development, I fixed a Python Lambda function, debugged a GitHub Actions pipeline, and deployed a new frontend—all in under an hour. I didn’t write the MIME parsing logic or look up the S3 CLI flags. I described what I needed. The agent wrote the code.

When I consolidated my domains using CloudFront Functions, I didn’t memorize the request.headers.host.value syntax. I defined the architecture—“redirect these hosts to this target with a 301”—and engineered the prompt to get working code.

The engineering wasn’t typing. The engineering was:

• Choosing the right architecture (CloudFront Function vs. S3 redirect buckets)
• Validating the AI’s logic against edge cases
• Orchestrating the deployment

Content Creation → Prompt Engineering

When I write for this blog, I’m often acting as an editor-in-chief for an AI staff writer. I provide structure, context, and voice. The AI drafts. I refine.

When I need visuals, I don’t open Photoshop. As I explored in AI Image Generation, I engineer the prompt—tweaking negative prompts, adjusting seeds, iterating until the output matches my vision.

Same discipline. Different domain.

Every Role Is Becoming AI Engineering

Think about what’s actually happening across industries:

Traditional Role	New Reality
Software Engineer	Orchestrating coding agents, validating AI-generated code
DevOps Engineer	Prompting infrastructure agents, reviewing IaC outputs
Writer	Directing LLMs, editing AI drafts for voice and accuracy
Designer	Engineering image generation prompts, curating outputs
Manager	Using AI for strategy synthesis, communication drafting

The job titles may stay the same. But the core skill is converging: the ability to effectively collaborate with AI systems.

That’s AI Engineering.

The Question That Matters Now

As I wrote earlier today in Ask Better Questions or Get Left Behind, Jensen Huang’s advice cuts through the noise:

“How does AI help me do my job better?”

Not hypothetically. Not someday. Today. In your role.

The people who keep asking that question—and keep refining their ability to work with AI—are the ones building leverage. Everyone else is just using a smarter Google.

The Human Stays in the Loop

Does this mean humans become irrelevant? No. As I discussed in Humans in the Loop in the AI Era, this shift makes human oversight more strategic, not less.

When AI handles execution, humans provide:

1. Judgment — Is this output correct? Appropriate? Good?
2. Context — What’s the real problem we’re solving?
3. Direction — Where should we go next?

AI isn’t replacing your job. It’s promoting you to Lead Engineer of your own AI workforce.

Final Thought

The hottest programming language is English.
The emerging universal skill is AI Engineering.

And the question isn’t whether AI will change your job.
It’s whether you’ll learn to engineer with it—or get left behind.

Jensen Huang on Using AI Early in Your Career

“The people who stand out aren’t the ones asking AI random questions, they’re the ones asking how AI helps them do their job better.”
— Jensen Huang, CEO of NVIDIA

That line hits harder in today’s market than it would have a few years ago.

Right now, unemployment for ages 20–24 has jumped to ~9.2%, a level usually seen during recessions. New grads are competing in a market where credentials matter less and leverage matters more.

AI isn’t the advantage by itself.
Knowing what to ask is.

The Wrong Way to Use AI (Most People Are Here)

Most people treat AI like:

• A smarter Google
• A code generator
• A homework helper
• A content writer

So they ask:

• “Explain Kubernetes”
• “Write this function”
• “Summarize this article”
• “Make me a resume”

That might save time.
It does not create leverage.

The Right Question (The One Jensen Is Pointing At)

Instead of asking:

“What can AI do?”

Ask this, persistently:

“How does AI help me do my job better?”

Not hypothetically.
Not someday.
Today. In my role.

That single question reframes AI from a tool into a multiplier.

The Question, Applied to You

Here’s how that question looks when you actually operationalize it.

If You’re a Student or New Grad

Wrong question:
“Can AI explain this topic?”

Better question:
“How can AI help me learn this faster than everyone else?”

Best question:
“How can AI help me build proof that I can already do this job?”

Concrete uses:

• Turn lectures → interview-ready explanations
• Convert assignments → portfolio artifacts
• Simulate interviews weekly
• Build tiny demos instead of just studying

Outcome:
You stop competing on GPA and start competing on evidence.

If You’re an Engineer

Wrong question:
“Can AI write this code?”

Better question:
“How can AI help me ship higher-quality systems faster?”

Best question:
“How can AI turn me from a ticket-closer into a systems thinker?”

Concrete uses:

• Pre-think designs before reviews
• Generate test strategies & edge cases
• Turn vague tasks into clear plans
• Document decisions and tradeoffs

Outcome:
You look senior earlier, because you think in systems, not snippets.

If You’re a Manager / SDM

Wrong question:
“Can AI help me write emails?”

Better question:
“How can AI help me lead people better?”

Best question:
“How can AI multiply my impact across people, projects, and stakeholders?”

Concrete uses:

• Draft clear feedback (especially hard feedback)
• Prepare for tough conversations
• Translate strategy → execution plans
• Produce crisp weekly updates that build trust

Outcome:
Less chaos. More clarity. Higher trust.

If You’re a Builder or Founder

Wrong question:
“What startup ideas does AI have?”

Better question:
“How can AI help me validate ideas faster?”

Best question:
“How can AI compress learning loops so I fail or win quickly?”

Concrete uses:

• Write landing pages in minutes
• Simulate customer objections
• Generate outreach messages
• Scope MVPs aggressively small

Outcome:
You test markets, not fantasies.

The Meta-Question (This Is the Cheat Code)

No matter your role, the question always reduces to this:

“If I keep asking AI better questions than everyone else, what compounds for me in 12 months?”

• Better judgment
• Faster execution
• Stronger communication
• Sharper intuition
• More optionality

AI rewards clarity of intent, not curiosity alone.

Why This Matters Right Now

In a weak job market:

• Average effort doesn’t stand out
• Credentials decay faster
• Titles matter less than output

AI creates a brutal divide:

1. People who use it to think
2. People who use it to ask

Jensen Huang’s advice isn’t about prompts.
It’s about ownership of your trajectory.

Final Takeaway

Don’t ask AI to replace your work.

Ask AI to:

• sharpen your thinking
• compress your learning
• multiply your impact
• leave artifacts behind

The people who win won’t be the ones who used AI the most.

They’ll be the ones who asked the right question early,
and kept asking it every day.

At their respective peaks, AI hype is likely higher than cloud hype, primarily due to the following factors:

1. Broader Public Awareness

• Cloud Hype (2010s): While the cloud revolutionized IT and business operations, its hype was more confined to tech-savvy audiences, enterprises, and investors. Everyday consumers indirectly benefited (e.g., through apps like Netflix or Dropbox), but they didn’t deeply engage with the concept.
• AI Hype (2020s): AI directly engages the public. Tools like ChatGPT, MidJourney, and DALL·E have reached millions of non-technical users, creating a cultural impact that extends beyond the tech world.

2. Immediate User Impact

• Cloud Computing: Delivered operational and cost advantages primarily for businesses (e.g., scalable infrastructure, SaaS). For end-users, its effects were indirect.
• AI: Offers tangible, personal utility—content creation, coding help, art generation, personalized recommendations—leading to stronger hype among individuals.

3. Media & Speculation

• Cloud: The conversation was heavily focused on B2B transformations and tech investments.
• AI: AI stirs debates about existential risks, societal impacts, and philosophical questions, grabbing broader headlines. The sheer excitement (and fear) surrounding AI’s potential future (e.g., AGI) amplifies the hype.

4. Financial Backing and Market Size

• Cloud (2010s): Cloud computing was massive and continues to grow, but much of the excitement was tied to its enterprise applications.
• AI (2020s): AI investments now span every sector (healthcare, finance, creative arts, education). The venture capital poured into AI startups in a single year (e.g., 2023-24) has outpaced funding for cloud in its earlier growth phase.

---

In Conclusion

While cloud hype laid the foundation for AI to flourish, AI has reached a broader cultural, technological, and economic peak of hype. However, AI hype may also burn faster if real-world applications don’t meet expectations or if challenges (e.g., ethical concerns, regulation) slow growth.

---

Disclaimer: This response was generated by AI based on the prompt: “Cloud hype vs AI hype, which is higher at peak?”

I’ve used many AI tools and apps so far. I’ve also built AI apps at work and as side projects. Some of these apps have great UX and I tend to go back to use it. Some don’t and I do not want to go back. It may seem like an obvious observation, but having a good UX is very important for AI tools.

Also, AI tools and software builders have the opportunity to differentiate themselves by experimenting with new flows and UX.

I’m seeing massive potential in Human-in-the-Loop (HITL) workflows as AI agents become more sophisticated. With specialized agents handling everything from code analysis to content generation, human oversight is evolving into a strategic advantage rather than just a safety check.

The magic happens in the feedback loop - humans refining AI outputs, which in turn makes the AI better at understanding human intent. I’ve found this particularly powerful in development workflows, where HITL systems help catch edge cases and maintain quality while significantly speeding up delivery.

As we move towards multi-agent systems working in concert, I believe HITL will become the key differentiator. It’s not just about having AI capabilities anymore - it’s about effectively partnering with AI to amplify what humans do best: providing context, judgment, and creative direction.

Written on December 7, 2024

#HITL #AIAgents #FutureOfWork

AI Agents are now becoming a thing, where an LLM based AI agent with instruction performs a tasks. There are several examples of AI agents which are already very impressive.

Soon there will be Trillions of AI agents on the web doing stuff. It’s gonna get chaotic but interesting.

I love both blockchain and AI, and I have delved deep into both of these fascinating technologies. Currently, I find myself captivated by the world of AI.

AI has a clear advantage when it comes to day-to-day applications. It is more readily applicable, whereas blockchain often requires users to learn new concepts and technologies. Convincing traditional art collectors to buy NFTs or persuading social apps to incorporate cryptocurrency for payments and tips requires a real resonance.

Existing businesses were initially skeptical about the potential of blockchain to enhance their operations. The magic of AI, on the other hand, has been witnessed by many, while the magic of blockchain has only been seen by a very few.

Digital artists have found a good platform to sell art with NFTs, and there are now several use cases of blockchain enhancing finance or making certain kinds of transactions easier than before.

However, blockchain has not made the case for a mainstream app. On the other hand, AI is right there at the forefront of the technological revolution. ChatGPT had 100 million+ users just two months after launch.

Why the AI bubble is better than the Blockchain bubble:

Broad Applicability: AI has applications in virtually every industry. Healthcare, finance, transportation, education, entertainment, and many more sectors are actively integrating AI technology into their operations. Blockchain, while revolutionary in its potential, primarily pertains to industries with significant data security and transactional needs like finance, supply chain, and legal contracts.

Research and Development: There is extensive ongoing research in the field of AI, with major breakthroughs happening regularly. While blockchain also receives research attention, AI has been around for a longer period, and its research domain is comparatively broader and more mature.

Integration with Existing Systems: AI technologies can often be integrated with existing systems, improving their capabilities. Blockchain often requires the creation of new systems or significant alterations to existing ones.

Regulatory Environment: While both AI and blockchain face significant regulatory hurdles, those faced by blockchain, particularly in the financial sector, can be especially steep due to concerns over fraud, money laundering, and financial stability.

Did you notice? The four points above were written by ChatGPT.

I’m still hopeful that blockchain will enhance certain areas of life, like being able to hold game collectibles in NFTs and being able to sell my hard-earned upgrades in a game via NFT transfers.

In summary, I love both AI and blockchain, but AI has proven to be more applicable in my experience.

DALL.E 2.

My first rodeo with AI image generation was with OpenAI DALL.E 2. I was blown away by DALL.E’s output. My prompt: “a Kathmandu street in the mt Everest region. There’s a yak and a yeti in the frame. Mountains are visible in the background.” gave me some realistic looking scenes from Nepal.

Stable Diffusion

Currently, my go-to for AI Image generation is using a MacOS app called DiffusionBee that creates images using Stable Diffusion. Stable Diffusion is a deep learning, text-to-image model released in 2022.

DiffusionBee: https://diffusionbee.com/

Midjourney

Midjourney is probably the best text-to-image generation AI progrom out there as of this writing. I have only used it a few times. The only way to generate image is to join their Discord Server and currently only paid subscribers can generate images. But the images are absolutely incredible. Cinematic scenes, fashion ideas and much more have been created with Midjourney. Here’s one of the twitter threads showing the power of Midjourney Version 5.1

---

I have a few to-do list items in terms of what I want to do with text-to-image generation. I’ll be sharing more updates on this blog and on Twitter. So follow me there @ravsau