Artificial Intelligence is entering a new phase of rapid advancement, with every new model release pushing the boundaries of what machines can assist humans with. From coding and debugging to reasoning and long-context analysis, the capabilities of modern AI systems are expanding quickly, prompting organizations to rethink how technology teams work, innovate, and build solutions. At the same time, these advancements are also sparking deeper conversations about workforce transformation, enterprise adoption, governance, and long-term technology strategy.
Amman Walia, Group CIO at Kanodia Group, believes the latest developments in AI models are not just incremental upgrades but indicators of a broader shift toward more capable and agent-like systems. In this interaction with DT, he shares his views on the launch of Anthropic’s latest models, the evolving role of engineers, the concerns organizations are grappling with, and the mindset young professionals must adopt to thrive in the AI era.
The Rise of Long-Context AI Systems
Commenting on the launch of the latest AI models by Anthropic, Amman Walia, Group CIO at Kanodia Group, explains that the February release of Claude Opus 4.6 and Claude Sonnet 4.6 may appear like a routine upgrade, but the details reveal something more significant.
“The February drop of Claude Opus 4.6 and Claude Sonnet 4.6 by Anthropic appears to be the usual pattern of model upgrades. However, the fine print indicates something more significant: both models come with a 1-million-token context window (beta), and this one feature alone raises the bar on what “useful” can be in AI applications. A larger context window is more than just the ability to process longer documents. It’s about maintaining intent, constraints, and context over the course of hours of work without having to constantly summarize, compress, or lose important assumptions. It’s the difference between an assistant and an agent,” he said.
“Anthropic has also made a significant point about preference: in Claude Code testing, users preferred Sonnet 4.6 over Sonnet 4.5 by about 70%. This is important because Sonnet is the sweet spot for many teams. It’s fast, scalable, and usually the go-to choice for production. If Sonnet continues to improve towards “frontier-like” levels of reliability, it narrows the gap between the high-end and standard options and changes the cost-benefit equation in a way that the market usually favors,” he observes.
Drawing from his own experiments, Walia remarks that Opus 4.6 stands out in areas requiring deeper reasoning. “My own experiments confirm the trend, with one key caveat. Opus 4.6 is, in my experience, the more accurate model for code generation and, in particular, debugging when bugs occur. It is superior at keeping a consistent mental map of the system during tracing of failure modes, suggesting patches that are consistent with the larger design, and adapting its strategy when the initial approach fails. This is most apparent in multi-step problems: solving dependency conflicts, tracing subtle logic regressions, and implementing changes across multiple files while maintaining consistency in interfaces. Opus 4.6 seems less susceptible to “confident drift,” where the output of the model slowly drifts away from the original specification,” he explains.
“Sonnet 4.6, on the other hand, performs well enough to often rival Opus 4.5 in the typical engineering workflow. For refactorings, unit test code generation, boilerplate code scaffolding, routine bug fixes, and code explanations, Sonnet 4.6 is good enough that I don’t feel compelled to call upon Opus for every problem set. That’s a big deal: the model that you can afford to deploy everywhere is the one that ultimately shapes the developer culture, team processes, and ultimately platform adoption,” he notes.
He stated, “Market feedback also indicates confidence. New versions appeared rapidly in the major distribution channels such as GitHub and AWS, which indicates that partners are confident enough in these models to roll them out broadly for enterprise experimentation. Rapid deployment is also a form of tacit endorsement: distribution partners rarely promote models that create headaches for support teams,” he adds.
“Nonetheless, the most important new capability for me is still the 1M-token context window. This is the capability that makes possible truly agentic systems: agents that can plan, remember, and act on complex tasks, codebases, and projects over long periods of time. Opus 4.6 and Sonnet 4.6 are important improvements, but the long context window is the key that unlocks the entire frontier of what AI can do,” he said.
Engineering Roles Are Moving Up the Stack
Discussing how engineering roles are evolving in the AI era, Amman explains that the pace of AI progress is already transforming technical work. “We are living in a period where the progress of AI is compounding at an increasingly rapid rate. There is ongoing discussion about how much current large language models can achieve in the next few years (some leaders have even proposed timelines as aggressive as 2028 for “superintelligence”). Whatever the specific date, the trend is clear: the engineering and technical work is already undergoing a transformation, and this will only accelerate,” he says.
“The first area where this is most apparent is in software development, where the work is increasingly shifting from “writing code” to “directing outcomes.” Engineers are increasingly enabled by copilots, code generation tools, automated tests, and agent-like workflows that can quickly generate functional code and mitigate common bugs,” he explains.
According to Amman, “This is already taking over a significant part of L1 work, scaffolding, standard CRUD operations, simple refactors, documentation generation, and routine unit tests. Even a significant part of L2 work (debugging and fixing) is getting compressed, since AI can point to likely causes of problems, identify fishy commits, generate patches, and suggest test cases to verify them. But the job doesn’t go away it just shifts up the stack. As code production becomes cheaper, the differentiator moves from code to judgment, architecture, and correctness. Engineers who remain relevant will be those (1)who can articulate the correct problem, rather than simply solving for a proposed solution; (2) architect systems that function correctly in the presence of real-world constraints; and (3) correctly validate their outputs.”
The most important skills for the future will involve:
Problem framing and product thinking: translating vague requirements into precise specifications, establishing success criteria, and trade-offs (speed vs. cost vs. reliability vs. security).
Systems thinking and architecture: scalability, robustness, observability, data architecture, integration complexity, and performance.
Deep debugging and reasoning: bugs will be more complex (distributed systems, dependency graphs, production-only bugs), and excellent root-cause analysis will remain a highly valued skill.
Verification and evaluation: with AI-generated code, the key skill is proving things are right improved testing approaches, CI best practices, contract testing, and for AI capabilities: evaluation frameworks, guardrails, monitoring, and rollback plans.
Security, privacy, and responsible engineering: preventing data exfiltration, managing prompt injection and supply chain attacks, and developing compliant and trustworthy systems.
AI-native execution: effective use of AI tools (prompting, agent workflows, internal accelerators), and understanding their limitations, hallucinations, cost, latency, and reliability.
“In short, AI will enable and speed up lower-level execution, particularly L1 and L2. Developers will remain valuable by leveraging AI as a multiplier and focusing on the skills that are most difficult to automate: critical thinking, creativity, system thinking, rigorous validation, and sound delivery of production-ready solutions,” he concludes.
Managing AI Excitement and Anxiety
Reflecting on the broader market response to rapid AI innovation, Amman says, “AI is advancing at such a rapid pace that I am witnessing a unique combination of excitement, nervousness, and confusion—sometimes within the same organization.”
“From a leadership perspective, there are two distinct groups. The first is driven by FOMO. These are owners and leaders who are eager to get on the AI train quickly, before they are left in the dust. While the sentiment is well-intentioned, the danger is real. Many leaders are eager to get on the train without understanding what is truly required, how the process of implementation will work, what data and processes need to be in place, and how to measure true ROI. The second group is at the other end of the spectrum. These leaders speak only in the language of ROI and expect immediate returns before allowing experimentation to occur. While this sounds prudent, in reality, it can stifle growth. AI adoption is not a one-time event; it is a capability-building process. When ROI is the only filter applied too soon, organizations fail to develop the learning muscle necessary to scale AI safely and effectively,” he explains.
“From boardrooms to offices and shop floors, the issue becomes very human. People fear job replacement “Will AI replace me?” but an even more daunting fear is competitive replacement: “Even if AI doesn’t replace me, will someone who uses AI more effectively and better replace me?” This fear is not unfounded. As AI is increasingly woven into the fabric of our daily work, the productivity divide can quickly grow between those who adopt early and those who lack access, training, and confidence,” he observes.
“Then, of course, there is the risk and governance side of the equation. Risk boards are rightly worried about data breaches, privacy, intellectual property, and regulatory compliance. But simultaneously, the same group of stakeholders is clamoring for the use of generative AI in research, summarizing complex regulations, creating compliance papers, and speeding up audit preparation. The message is clear: “We want the productivity gain, but we want control,” he added.
He emphasizes, “My conclusion is straightforward and has been true for previous technological revolutions, such as the machine revolution and the IT revolution. The key is that to remain relevant, one must continually upskill. Not everyone needs to be a data scientist, but everyone needs to be AI-literate, workflow-savvy, and have good judgment to know when to trust the output, when to check, and when to escalate. The organizations that will succeed will be those that find the right balance between speed and responsibility to experiment early to learn, to govern well to protect, to measure ROI in a thoughtful way as maturity increases, and to invest in people so that adoption is empowering, not threatening.”
Advice for Young Engineers Entering the AI Era
Offering guidance to young engineers beginning their careers during this technological transition, Amman says they are entering the industry at a uniquely transformative moment.
“Young engineers are entering a unique point in history: not only a new technology, but a new norm for how societies construct, decide, and scale. When technology adoption ramps up this quickly, it can be both exhilarating and intimidating—like standing on a moving train. My counsel is simple: lean into the change with open arms, but hold on even tighter to the human skills that make you valuable. First, accept that change is not a phase. AI-enhanced workflows, automation, and the rapid evolution of new platforms are the new normal. Fighting these changes only holds you back from growing. Curiosity is the best career move. Investigate what’s new and emerging, get your hands dirty with it, and view every new skill as an opportunity to stretch your reach,” he explains.
“But here’s the more important piece: don’t farm out your thinking. In a world where tools can write code, it’s not who can crank out the most work that will matter—it’s who can crank out the best judgment. Critical thinking is your guide: asking the right questions, challenging shaky assumptions, checking answers, and thinking through trade-offs. Creativity is your engine: dreaming up new user experiences, new products, and new ways to simplify tough problems,” he explains.
“As a new graduate, you have a distinct advantage: you don’t have to overcome deeply ingrained habits. You’re not tied to “the way it’s always been done.” And that’s a blessing. Leverage it to learn fast, adapt with confidence, and build with new patterns, not old ones,” he says.
He further encourages engineers to think beyond isolated features and focus on broader systems. “The coming decade will favor engineers who can transcend isolated work and think about systems. Learn to talk about systems and processes, not just features. How does a product behave under stress? What happens when it breaks? How do you build for reliability, safety, privacy, and accountability from day one? These are the questions that define the quality of technology—and the trust that society puts in it,” he explains.
Finally, Amman highlights the broader responsibility engineers carry. He stated, “Lastly, speak up more in society. Engineering is no longer a black box. The systems you build affect work, education, health, security, and public discourse. Your role in the world does not begin at deployment and end at the release notes. It includes speaking up in conversations about impact, advocating for responsible decisions, and helping teams and communities think clearly about the consequences. This is not just a moment about remaining relevant and remaining employed. This is a moment about remaining meaningful. Your power will come from deep thinking, careful design, and the willingness to build responsibly. The future will not be made by the loudest technology—it will be made by the people who use technology to move the human civilization forward.”
𝐒𝐭𝐚𝐲 𝐢𝐧𝐟𝐨𝐫𝐦𝐞𝐝 𝐰𝐢𝐭𝐡 𝐨𝐮𝐫 𝐥𝐚𝐭𝐞𝐬𝐭 𝐮𝐩𝐝𝐚𝐭𝐞𝐬 𝐛𝐲 𝐣𝐨𝐢𝐧𝐢𝐧𝐠 𝐭𝐡𝐞 WhatsApp Channel now! 👈📲
𝑭𝒐𝒍𝒍𝒐𝒘 𝑶𝒖𝒓 𝑺𝒐𝒄𝒊𝒂𝒍 𝑴𝒆𝒅𝒊𝒂 𝑷𝒂𝒈𝒆𝐬 👉 Facebook, LinkedIn, Twitter, Instagram