Quote for the day:
“Great minds discuss ideas; average minds discuss events; small minds discuss people.” -- Eleanor Roosevelt
Who Owns End-of-Life Data?
Enterprises have never been more focused on data. What happens at the end of
that data's life? Who is responsible when it's no longer needed? Environmental
concerns are mounting as well. A Nature study warns that AI alone could generate
up to 5 million metric tons of e-waste by 2030. A study from researchers at
Cambridge University and the Chinese Academy of Sciences said top reason
enterprises dispose of e-waste rather than recycling computers is the cost.
E-waste can contain metals, including copper, gold, silver aluminum and rare
earth elements, but proper handling is expensive. Data security is a concern as
well as breach proofing doesn't get better than destroying equipment. ...
End-of-life data management may sit squarely in the realm of IT, but it
increasingly pulls in compliance, risk and ESG teams, the report said. Driven by
rising global regulations and escalating concerns over data leaks and breaches,
C-level involvement at every stage signals that end-of-life data decisions are
being treated as strategically vital - not simply handed off. Consistent IT
participation also suggests organizations are well-positioned to select and
deploy solutions that work with their existing tech stack. That said, shared
responsibility doesn't guarantee seamless execution. Multiple stakeholders can
lead to gaps unless underpinned by strong, well-communicated policies, the
report said.How AI is Disrupting the Data Center Software Stack
Over the years, there have been many major shifts in IT infrastructure – from
the mainframe to the minicomputer to distributed Windows boxes to
virtualization, the cloud, containers, and now AI and GenAI workloads. Each
time, the software stack seems to get torn apart. What can we expect with GenAI?
... Galabov expects severe disruption in the years ahead on a couple of fronts.
Take coding, for example. In the past, anyone wanting a new industry-specific
application for their business might pay five figures for development, even if
they went to a low-cost region like Turkey. For homegrown software development,
the price tag would be much higher. Now, an LLM can be used to develop such an
application for you. GenAI tools have been designed explicitly to enhance and
automate several elements of the software development process. ... Many
enterprises will be forced to face the reality that their systems are
fundamentally legacy platforms that are unable to keep pace with modern AI
demands. Their only course is to commit to modernization efforts. Their speed
and degree of investment are likely to determine their relevance and competitive
positioning in a rapidly evolving market. Kleyman believes that the most
immediate pressure will fall on data-intensive, analytics-driven platforms such
as CRM and business intelligence (BI). AI Improves at Improving Itself Using an Evolutionary Trick
The best SWE-bench agent was not as good as the best agent designed by expert
humans, which currently scores about 70 percent, but it was generated
automatically, and maybe with enough time and computation an agent could evolve
beyond human expertise. The study is a “big step forward” as a proof of concept
for recursive self-improvement, said Zhengyao Jiang, a cofounder of Weco AI, a
platform that automates code improvement. Jiang, who was not involved in the
study, said the approach could made further progress if it modified the
underlying LLM, or even the chip architecture. DGMs can theoretically score
agents simultaneously on coding benchmarks and also specific applications, such
as drug design, so they’d get better at getting better at designing drugs. Zhang
said she’d like to combine a DGM with AlphaEvolve. ... One concern with both
evolutionary search and self-improving systems—and especially their combination,
as in DGM—is safety. Agents might become uninterpretable or misaligned with
human directives. So Zhang and her collaborators added guardrails. They kept the
DGMs in sandboxes without access to the Internet or an operating system, and
they logged and reviewed all code changes. They suggest that in the future, they
could even reward AI for making itself more interpretable and aligned.Data center costs surge up to 18% as enterprises face two-year capacity drought
Smart enterprises are adapting with creative strategies. CBRE’s Magazine
emphasizes “aggressive and long-term planning,” suggesting enterprises extend
capacity forecasts to five or 10 years, and initiate discussions with providers
much earlier than before. Geographic diversification has become essential. While
major hubs price out enterprises, smaller markets such as São Paulo saw pricing
drops of as much as 20.8%, while prices in Santiago fell 13.7% due to shifting
supply dynamics. Magazine recommended “flexibility in location as key, exploring
less-constrained Tier 2 or Tier 3 markets or diversifying workloads across
multiple regions.” For Gogia, “Tier-2 markets like Des Moines, Columbus, and
Richmond are now more than overflow zones, they’re strategic growth anchors.”
Three shifts have elevated these markets: maturing fiber grids, direct renewable
power access, and hyperscaler-led cluster formation. “AI workloads, especially
training and archival, can absorb 10-20ms latency variance if offset by 30-40%
cost savings and assured uptime,” said Gogia. “Des Moines and Richmond offer
better interconnection diversity today than some saturated Tier-1 hubs.”
Contract flexibility is also crucial. Rather than traditional long-term leases,
enterprises are negotiating shorter agreements with renewal options and
exploring revenue-sharing arrangements tied to business performance.Fintech’s AI Obsession Is Useless Without Culture, Clarity and Control
what does responsible AI actually mean in a fintech context? According to PwC’s
2024 Responsible AI Survey, it encompasses practices that ensure fairness,
transparency, accountability and governance throughout the AI lifecycle. It’s
not just about reducing model bias — it’s about embedding human oversight,
securing data, ensuring explainability and aligning outputs with brand and
compliance standards. In financial services, these aren’t "nice-to-haves" —
they’re essential for scaling AI safely and effectively. Financial marketing is
governed by strict regulations and AI-generated content can create brand and
legal risks. ... To move AI adoption forward responsibly, start small. Low-risk,
high-reward use cases let teams build confidence and earn trust from compliance
and legal stakeholders. Deloitte’s 2024 AI outlook recommends beginning with
internal applications that use non-critical data — avoiding sensitive inputs
like PII — and maintaining human oversight throughout. ... As BCG highlights, AI
leaders devote 70% of their effort to people and process — not just technology.
Create a cross-functional AI working group with stakeholders from compliance,
legal, IT and data science. This group should define what data AI tools can
access, how outputs are reviewed and how risks are assessed.Is Microsoft’s new Mu for you?
Mu uses a transformer encoder-decoder design, which means it splits the work
into two parts. The encoder takes your words and turns them into a compressed
form. The decoder takes that form and produces the correct command or answer.
This design is more efficient than older models, especially for tasks such as
changing settings. Mu has 32 encoder layers and 12 decoder layers, a setup
chosen to fit the NPU’s memory and speed limits. The model utilizes rotary
positional embeddings to maintain word order, dual-layer normalization to
maintain stability, and grouped-query attention to use memory more efficiently.
... Mu is truly groundbreaking because it is the first SLM built to let users
control system settings using natural language, running entirely on a mainstream
shipping device. Apple’s iPhones, iPads, and Macs all have a Neural Engine NPU
and run on-device AI for features like Siri and Apple Intelligence. But Apple
does not have a small language model as deeply integrated with system settings
as Mu. Siri and Apple Intelligence can change some settings, but not with the
same range or flexibility. ... By processing data directly on the device, Mu
keeps personal information private and responds instantly. This shift also makes
it easier to comply with privacy laws in places like Europe and the US since no
data leaves your computer.
Is It a Good Time to Be a Software Engineer?
AI may be rewriting the rules of software development, but it hasn’t erased the
thrill of being a programmer. If anything, the machines have revitalised the joy
of coding. New tools make it possible to code in natural language, ship
prototypes in hours, and bypass tedious setup work. From solo developers to
students, the process may feel more immediate or rewarding. Yet, this sense of
optimism exists alongside an undercurrent of anxiety. As large language models
(LLMs) begin to automate vast swathes of development, some have begun to wonder
if software engineering is still a career worth betting on. ... Meanwhile, Logan
Thorneloe, a software engineer at Google, sees this as a golden era for
developers. “Right now is the absolute best time to be a software engineer,” he
wrote on LinkedIn. He points out “development velocity” as the reason. Thorneleo
believes AI is accelerating workflows, shrinking prototype cycles from months to
days, and giving developers unprecedented speed. Companies that adapt to this
shift will win, not by eliminating engineers, but by empowering them. More than
speed, there’s also a rediscovered sense of fun. Programmers who once wrestled
with broken documentation and endless boilerplate are rediscovering the creative
satisfaction that first drew them to the field.
Dumping mainframes for cloud can be a costly mistake
Despite industry hype, mainframes are not going anywhere. They quietly support
the backbone of our largest banks, governments, and insurance companies. Their
reliability, security, and capacity for massive transactions give mainframes an
advantage that most public cloud platforms simply can’t match for certain
workloads. ... At the core of this conversation is culture. An innovative IT
organization doesn’t pursue technology for its own sake. Instead, it encourages
teams to be open-minded, pragmatic, and collaborative. Mainframe engineers have
a seat at the architecture table alongside cloud architects, data scientists,
and developers. When there’s mutual respect, great ideas flourish. When legacy
teams are sidelined, valuable institutional knowledge and operational stability
are jeopardized. A cloud-first mantra must be replaced by a philosophy of “we
choose the right tool for the job.” The financial institution in our opening
story learned this the hard way. They had to overcome their bias and reconnect
with their mainframe experts to avoid further costly missteps. It’s time to
retire the “legacy versus modern” conflict and recognize that any technology’s
true value lies in how effectively it serves business goals. Mainframes are part
of a hybrid future, evolving alongside the cloud rather than being replaced by
it.
Why Modern Data Archiving Is Key to a Scalable Data Strategy
Organizations are quickly learning they can’t simply throw all data, new and
old, at an AI strategy; instead, it needs to be accurate, accessible, and, of
course, cost-effective. Without these requirements in place, it’s far from
certain AI-powered tools can deliver the kind of insight and reliability
businesses need. As part of the various data management processes involved,
archiving has taken on a new level of importance. ... For organizations that
need to migrate data, for example, archiving is used to identify which essential
datasets, while enabling users to offload inactive data in the most
cost-effective way. This kind of win-win can also be applied to cloud resources,
where moving data to the most appropriate service can potentially deliver
significant savings. Again, this contrasts with tiering systems and NAS
gateways, which rely on global file systems to provide cloud-based access to
local files. The challenge here is that access is dependent on the gateway
remaining available throughout the data lifecycle because, without it, data
recall can be interrupted or cease entirely. ... It then becomes practical to
strike a much better balance across the typical enterprise storage technology
stack, including long-term data preservation and compliance, where data doesn’t
need to be accessed so often, but where reliability and security are crucial.
The Impact of Regular Training and Timely Security Policy Changes on Dev Teams
Constructive refresher training drives continuous improvement by reinforcing
existing knowledge while introducing new concepts like AI-powered code
generation, automated debugging and cross-browser testing in manageable
increments. Teams that implement consistent training programs see significant
productivity benefits as developers spend less time struggling with unfamiliar
tools and more time automating tasks to focus on delivering higher value. ...
Security policies that remain static as teams grow create dangerous blind spots,
compromising both the team’s performance and the organization’s security
posture. Outdated policies fail to address emerging threats like malware
infections and often become irrelevant to the team’s current workflow, leading
to workarounds and system vulnerabilities. ... Proactive security integration
into development workflows represents a fundamental shift from reactive security
measures to preventative strategies. This approach enables growing teams to
identify and address security concerns early in the development process,
reducing the cost and complexity of remediation. Cultivating a security-first
culture becomes increasingly important as teams grow. This involves embedding
security considerations into various stages of the development life cycle. Early
risk identification in cloud infrastructure reduces costly breaches and improves
overall team productivity.
/articles/staff-plus-strategic-thinking/en/smallimage/thumbnail-1749200507294.jpg)
