The Powerful AI – In my 23 years of watching AI evolve, and especially over the last four years of writing code and building local, air-gapped systems, I have never seen a technology that scales this poorly. Think about everything else we have built in the last few decades.
We scaled global video streaming, put smartphones in billions of pockets, and migrated whole industries to the cloud. We did all of that without threatening the stability of global power grids or damaging my drinking water or causing permanent chip shortages.
That early era was where my obsession with absolute system determinism truly took root. Sitting before dense stacks of code, building expert systems and hardcoded rule sets by hand, there was no room for probabilistic guessing or architectural drift; code either executed with mathematical certainty, or it did not execute at all. Now, looking back at how AI has evolved over the last 23 years, those foundational lessons in strict logic remain the ultimate anchor.
Over the last two decades, as I moved through massive fintech transformations, scaled high-throughput financial services engines with AI, Blockchain and several architectures with deep data, commercial and tech intensity, and took on executive technology leadership roles, that foundational craving for structural precision never wavered.
Why determinism is a life-or-death engineering issue—: In 1991, a Patriot missile defense system of US failed to intercept an incoming Scud missile because of a microscopic floating-point rounding error in its software. The system tracked time in tenths of a second but stored the values as a 24-bit binary fraction. Because $1/10$ has an infinite repeating binary expansion, the math was slightly truncated. After the system ran continuously for 100 hours, this tiny discrepancy accumulated to exactly 0.34 seconds. At supersonic speeds, that split-second lag caused the radar to look in the wrong place, missing the target by 687 meters—a tragic reminder of why absolute mathematical determinism is everything in mission-critical engineering.
The Lure of the Easy Way Out
I look at the code running on my screen today, and then I think back to how we built systems in 2003. Back then, if you wanted a system to work, you had to understand it. You had to design the data boundaries, write the logic, and take responsibility for every single byte.
- Illusion of Innovation: The tech industry is mistaking the integration of unvetted, third-party cloud API wrappers into core systems for genuine software architecture.
- Flawed Strategic Outsourcing: Relying entirely on multi-billion-parameter models hosted in external clouds replaces critical engineering with a risky strategy of “renting someone else’s brain.
It is a dependency trap that robs your company of its autonomy, drains your budget, and leaves your systems fragile. True technology leadership requires an unyielding commitment to architectural clarity and systemic discipline. If we don’t start building our own foundations again, our architectures will collapse under the first sign of real enterprise pressure.
Are We Solving Intelligence or Just Scaling Costs?
Everyone is talking about the race to AGI and superintelligence, but after spending years building large-scale technology platforms, I keep asking myself one simple question: Do the engineering and the economics actually work? What I’m seeing is an industry investing exponentially more compute, power, and capital for increasingly smaller gains. Somewhere along the way, we started believing that if a model isn’t smart enough, the only answer is to make it bigger. I am not convinced that’s the path forward.
This isn’t a blind criticism of Generative AI—I fully believe it will transform how we work. But we have to face the reality of diminishing returns. Every new model generation demands a staggering increase in GPUs, energy, and infrastructure, yet the actual intelligence improvements are becoming more incremental.
- Scale has limits. Bigger models don’t guarantee proportionally better intelligence.
- Efficiency is the real challenge. AI must become more economically sustainable, not just computationally larger.
- Innovation beats brute force. The next leap forward will come from smarter architectures, not simply more parameters.
As engineers, we’ve always known that brute force is a lazy solution. Better architecture always beats bigger architecture. I believe the next major breakthrough in AI won’t come from building the largest model on earth. It’ll come from building the smartest, most efficient systems.
There is almost an engineering blindness happening right now because people want to believe they are replicating a human mind. But let’s look under the hood. These are statistical engines designed to guess the next word. They lack common sense, they can’t reliably recall basic facts, and they operate nothing like a biological brain.
Three Takeaways Every AI Leader Should Be Thinking About
As I look at where AI is heading, I don’t see this as a debate about whether Generative AI will succeed—it absolutely will. The real question is how we actually get there. If we continue to measure progress only by model size and raw compute, we risk missing the architectural innovations that will truly define the next generation of technology.
| Key Takeaway | What It Means |
|---|---|
| Scale Has Limits | Bigger models don’t automatically produce proportionally better intelligence. Beyond a certain point, the returns become increasingly incremental while the costs continue to skyrocket. |
| Efficiency Is the New Competitive Advantage | The future belongs to AI systems that deliver more capability with less compute, lower energy consumption, and sustainable economics—not simply more GPUs. |
| Innovation Beats Brute Force | The next breakthrough won’t come from adding another trillion parameters. It will come from better architectures, improved reasoning, adaptive learning, and fundamentally smarter system design. |
As engineers and technology leaders, we’ve learned this lesson many times before. Lasting breakthroughs rarely come from making something bigger—they come from making it better. I believe AI is approaching that same inflection point, where thoughtful engineering will matter far more than brute-force scaling.
Beyond the AI Hype: Building Systems That Actually Work
After years of building and scaling technology platforms, I believe we are entering a defining moment in the AI journey. The industry has created incredible possibilities, but we need to move past experimentation, polished slide decks, and endless theoretical conversations. The next phase belongs to the builders—the people who can translate AI’s potential into reliable, scalable, production-grade systems that deliver real, measurable value.
- Execution Over Presentation: The future will not be won by the prettiest slides, the biggest press releases, or the loudest marketing claims. It will be won by teams that can actually design, build, deploy, and operate systems that solve real-world problems.
- Engineering Depth Matters: As AI becomes deeply embedded into critical enterprise platforms, strong software fundamentals—architecture, data pipelines, security, and reliability—will become more valuable than ever.
- Efficiency Will Separate Leaders From Followers: Sustainable technology requires systems that are highly performant, deterministic, cost-aware, and built for long-term operational success—not solutions that simply consume infinite resources.
- The Builders Will Shape What Comes Next: As the AI market matures, the organizations and engineers who can deliver practical, trustworthy, and scalable solutions will create the lasting impact.
That’s exactly why I’m stepping beyond the traditional corporate path and building an independent group of Forward-Deployed Engineering practitioners. We must not pay our school fees twice—once in the form of capital, and a second time in the form of our data, thought processes, designs, and IP. We must reject slideware, buzzwords, and theoretical discussions.
Instead, we must work directly with technology to transform our dreams and complex, legacy engineering environments into modern, intelligent, AI-at-the-core systems designed to perform today and scale tomorrow. The AI journey will eventually separate excitement from execution. When that happens, the people who understand how to build efficient, reliable, and meaningful technology will be the ones creating the future. Let’s build systems that don’t just look impressive on a screen—let’s build ones that actually work.
I am refusing to be raw material for the AI hype machine
I think a lot of us are feeling burned out right now. We are being pushed to treat ourselves, our work, and our user data as mere fuel to feed the corporate AI hype machine, and I think it’s time we stood up and talked about what’s actually happening behind the scenes. We need to call out the real problems we are seeing every day:
- Confidently incorrect systems: We keep acting like these models “know” things. But a hallucination isn’t just a minor data gap; it’s a model being completely, falsely certain about something it made up. It’s generating highly convincing fiction, and we are treating it like truth.
- Anxiety disguised as innovation: Right now, there is a frantic, almost panicked rush to slap an “AI” sticker on everything. We are throwing unvetted, raw API wrappers into production because management is terrified of looking left behind. This isn’t “agile innovation”—it’s panic-driven development, and we are the ones who have to stay up late fixing it when it breaks.
- Throwing money at massive clouds when we don’t need to: We’ve been conditioned to route every single task to massive, closed, expensive cloud models. It’s such a waste. We are overpaying for giant, generic minds to do simple, specific jobs. We could be running smaller, incredibly smart, local models (like Qwen, Llama, or Phi) right on our own machines—keeping our data safe, our costs low, and our speed fast.
We need to stop letting corporate anxiety dictate our engineering ethics. It’s time to build smarter, keep our data local, and put our foot down.
Why bad engineering decisions break my heart (and our systems)
Let’s be honest: true tech leadership isn’t about being the loudest person in the room or chasing the newest shiny object. It’s about caring enough to build things that actually last. Looking back at how we used to write code in 2003 compared to where we are heading now, one thing remains painfully true: when we take shortcuts to please a timeline, the system always collapses later under actual real-world pressure. Here is what I’ve learned from years of watching teams build themselves into corners:
- We pay for our architectural choices forever. Every shortcut we take today is a debt we force our future selves (or the next team) to pay. Building strong foundations and setting clear boundaries isn’t just “dev stuff”—it’s what keeps a company alive and able to adapt ten years down the road.
- Going fast means nothing if you’re running in the wrong direction. I know there is always pressure to ship right now. But speed without discipline is just chaos. I’ve never seen a rushed, messy fix outlast a system built with care, clear standards, and actual pride in the craft.
If we want to build things we can actually be proud of, we have to stop accepting “cowboy coding” as the norm. We have to draw clear boundaries around our data, write logical code we can actually prove works, and stop settling for “good enough for now.” When it comes to building platforms that thousands of real people rely on, we should aim for craft and care, not just crossing things off a checklist.
Conclusion – Staring at those two screens on my desk, I realized they represent the heartbeat of a complete leader: one reflecting the strategic vision of the boardroom, the other pulsing with the raw, executable logic of the “Builder.”
Strategy without execution is a hallucination, but execution without strategy is just drift. The tech industry frequently stumbles because leaders separate the abstract vision from the physical reality of the deployment environment, treating engineering as an outsourced detail rather than the core machine.
We must step forward and refuse to let our craft, our hard work, and our data be treated as mere fuel to feed the corporate AI hype machine. It is time to reclaim our perimeter, build with local, optimized intelligence, and take pride in our engineering. No shortcuts. No rented brains. For us, it is, and always will be, perfection or nothing.
—
Points to Note
it’s time to figure out when to use which tech—a tricky decision that can really only be tackled with a combination of experience and the type of problem in hand. So if you think you’ve got the right answer, take a bow and collect your credits! And don’t worry if you don’t get it right.
Books + Other readings Referred
- Research through open internet, news portals, white papers and imparted knowledge via live conferences & lectures.
- Lab and hands-on experience of @AILabPage (Self-taught learners group) members.
Feedback & Further Question
Do you have any burning questions about Big Data, “AI & ML“, Blockchain, FinTech,Theoretical PhysicsPhotography or Fujifilm(SLRs or Lenses)? Please feel free to ask your question either by leaving a comment or by sending me an email. I will do my best to quench your curiosity.
====================== About the Author ======================
Read about Author at : About Me
Thank you all, for spending your time reading this post. Please share your feedback / comments / critics / agreements or disagreement. Remark for more details about posts, subjects and relevance please read the disclaimer.
FacebookPage ContactMe Twitter =============================================================
