Daily Tech Digest - October 22, 2025


Quote for the day:

"Good content isn't about good storytelling. It's about telling a true story well." -- Ann Handley



When yesterday’s code becomes today’s threat

A striking new supply chain attack is sending shockwaves through the developer community: a worm-style campaign dubbed “Shai-Hulud” has compromised at least 187 npm packages, including the tinycolor package that has 2 million hits weekly, and spreading to other maintainers' packages. The malicious payload modifies package manifests, injects malicious files, repackages, and republishes — thereby infecting downstream projects. This incident underscores a harsh reality: even code released weeks, months, or even years ago can become dangerous once a dependency in its chain has been compromised. ... Sign your code: All packages/releases should use cryptographic signing. This allows users to verify the origin and integrity of what they are installing. Verify signatures before use: When pulling in dependencies, CI/CD pipelines, and even local dev setups, include a step to check that the signature matches a trusted publisher and that the code wasn’t tampered with. SBOMs are your map of exposure: If you have a Software Bill of Materials for your project(s), you can query it for compromised packages. Find which versions/packages have been modified — even retroactively — so you can patch, remove, or isolate them. Continuous monitoring of risk posture: It's not enough to secure when you ship. You need alerts when any dependency or component’s risk changes: new vulnerabilities, suspicious behavior, misuse of credentials, or signs that a trusted package may have been modified after release.


Cloud Sovereignty: Feature. Bug. Feature. Repeat!

Cloud sovereignty isn’t just a buzzword anymore, argues Kushwaha. “It’s a real concern for businesses across the world. The pattern is clear. The cloud isn’t a one-size-fits-all solution anymore. Companies are starting to realise that sometimes control, cost, and compliance matter more than convenience.” ... Cloud sovereignty is increasingly critical due to the evolving geopolitical scenario, government and industry-specific regulations, and vendor lock-ins with heavy reliance on hyperscalers. The concept has gained momentum and will continue to do so because technology has become pervasive and critical for running a state/country and any misuse by foreign actors can cause major repercussions, the way Bavishi sees it. Prof. Bhatt captures that true digital sovereignty is a distant dream and achieving this requires a robust ecosystem for decades. This isn’t counterintuitive; it’s evolution, as Kushwaha epitomises. “The cloud’s original promise was one of freedom. Today, when it comes to the cloud, freedom means more control. Businesses investing heavily in digital futures can’t afford to ignore the fine print in hyperscaler contracts or the reach of foreign laws. Sovereignty is the foundation for building safely in a fragmented world.” ... Organisations have recognised the risks of digital dependencies and are looking for better options. There is no turning back, Karlitschek underlines.


Securing AI to Benefit from AI

As organizations begin to integrate AI into defensive workflows, identity security becomes the foundation for trust. Every model, script, or autonomous agent operating in a production environment now represents a new identity — one capable of accessing data, issuing commands, and influencing defensive outcomes. If those identities aren't properly governed, the tools meant to strengthen security can quietly become sources of risk. The emergence of Agentic AI systems make this especially important. These systems don't just analyze; they may act without human intervention. They triage alerts, enrich context, or trigger response playbooks under delegated authority from human operators. ... AI systems are capable of assisting human practitioners like an intern that never sleeps. However, it is critical for security teams to differentiate what to automate from what to augment. Some tasks benefit from full automation, especially those that are repeatable, measurable, and low-risk if an error occurs. ... Threat enrichment, log parsing, and alert deduplication are prime candidates for automation. These are data-heavy, pattern-driven processes where consistency outperforms creativity. By contrast, incident scoping, attribution, and response decisions rely on context that AI cannot fully grasp. Here, AI should assist by surfacing indicators, suggesting next steps, or summarizing findings while practitioners retain decision authority. Finding that balance requires maturity in process design. 


The Unkillable Threat: How Attackers Turned Blockchain Into Bulletproof Malware Infrastructure

When EtherHiding emerged in September 2023 as part of the CLEARFAKE campaign, it introduced a chilling reality: attackers no longer need vulnerable servers or hackable domains. They’ve found something far better—a global, decentralized infrastructure that literally cannot be shut down. ... When victims visit the infected page, the loader queries a smart contract on Ethereum or BNB Smart Chain using a read-only function call. ... Forget everything you know about disrupting cybercrime infrastructure. There is no command-and-control server to raid. No hosting provider to subpoena. No DNS to poison. The malicious code exists simultaneously everywhere and nowhere, distributed across thousands of blockchain nodes worldwide. As long as Ethereum or BNB Smart Chain operates—and they’re not going anywhere—the malware persists. Traditional law enforcement tactics, honed over decades of fighting cybercrime, suddenly encounter an immovable object. You cannot arrest a blockchain. You cannot seize a smart contract. You cannot compel a decentralized network to comply. ... The read-only nature of payload retrieval is perhaps the most insidious feature. When the loader queries the smart contract, it uses functions that don’t create transactions or blockchain records. 


New 'Markovian Thinking' technique unlocks a path to million-token AI reasoning

Researchers at Mila have proposed a new technique that makes large language models (LLMs) vastly more efficient when performing complex reasoning. Called Markovian Thinking, the approach allows LLMs to engage in lengthy reasoning without incurring the prohibitive computational costs that currently limit such tasks. The team’s implementation, an environment named Delethink, structures the reasoning chain into fixed-size chunks, breaking the scaling problem that plagues very long LLM responses. Initial estimates show that for a 1.5B parameter model, this method can cut the costs of training by more than two-thirds compared to standard approaches. ... The researchers compared this to models trained with the standard LongCoT-RL method. Their findings indicate that the model trained with Delethink could reason up to 24,000 tokens, and matched or surpassed a LongCoT model trained with the same 24,000-token budget on math benchmarks. On other tasks like coding and PhD-level questions, Delethink also matched or slightly beat its LongCoT counterpart. “Overall, these results indicate that Delethink uses its thinking tokens as effectively as LongCoT-RL with reduced compute,” the researchers write. The benefits become even more pronounced when scaling beyond the training budget. 


The dazzling appeal of the neoclouds

While their purpose-built design gives them an advantage for AI workloads, neoclouds also bring complexities and trade-offs. Enterprises need to understand where these platforms excel and plan how to integrate them most effectively into broader cloud strategies. Let’s explore why this buzzword demands your attention and how to stay ahead in this new era of cloud computing. ... Neoclouds, unburdened by the need to support everything, are outpacing hyperscalers in areas like agility, pricing, and speed of deployment for AI workloads. A shortage of GPUs and data center capacity also benefits neocloud providers, which are smaller and nimbler, allowing them to scale quickly and meet growing demand more effectively. This agility has made them increasingly attractive to AI researchers, startups, and enterprises transitioning to AI-powered technologies. ... Neoclouds are transforming cloud computing by offering purpose-built, cost-effective infrastructure for AI workloads. Their price advantages will challenge traditional cloud providers’ market share, reshape the industry, and change enterprise perceptions, fueled by their expected rapid growth. As enterprises find themselves at the crossroads of innovation and infrastructure, they must carefully assess how neoclouds can fit into their broader architectural strategies. 


Wi-Fi 8 is coming — and it’s going to make AI a lot faster

Unlike previous generations of Wi-Fi that competed on peak throughput numbers, Wi-Fi 8 prioritizes consistent performance under challenging conditions. The specification introduces coordinated multi-access point features, dynamic spectrum management, and hardware-accelerated telemetry designed for AI workloads at the network edge. ... A core part of the Wi-Fi 8 architecture is an approach known as Ultra High Reliability (UHR). This architectural philosophy targets the 99th percentile user experience rather than best-case scenarios. The innovation addresses AI application requirements that demand symmetric bandwidth, consistent sub-5-millisecond latency and reliable uplink performance. ... Wi-Fi 8 introduces Extended Long Range (ELR) mode specifically for IoT devices. This feature uses lower data rates with more robust coding to extend coverage. The tradeoff accepts reduced throughput for dramatically improved range. ELR operates by increasing symbol duration and using lower-order modulation. This improves the link budget for battery-powered sensors, smart home devices and outdoor IoT deployments. ... Wi-Fi 8 enhances roaming to maintain sub-millisecond handoff latency. The specification includes improved Fast Initial Link Setup (FILS) and introduces coordinated roaming decisions across the infrastructure. Access points share client context information before handoff. 


Life, death, and online identity: What happens to your online accounts after death?

Today, we lack the tools (protocols) and the regulations to enable digital estate management at scale. Law and regulation can force a change in behavior by large providers. However, lacking effective protocols to establish a mechanism to identify the decedent’s chosen individuals who will manage their digital estate, every service will have to design their own path. This creates an exceptional burden on individuals planning their digital estate, and on individuals who manage the digital estates of the deceased. ... When we set out to write this paper, we wanted to influence the large technology and social media platforms, politicians, regulators, estate planners, and others who can help change the status quo. Further, we hoped to influence standards development organizations, such as the OpenID Foundation and the Internet Engineering Task Force (IETF), and their members. As standards developers in the realm of identity, we have an obligation to the people we serve to consider identity from birth to death and beyond, to ensure every human receives the respect they deserve in life and in death. Additionally, we wrote the planning guide to help individuals plan for their own digital estate. By giving people the tools to help describe, document, and manage their digital estates proactively, we can raise more awareness and provide tools to help protect individuals at one of the most vulnerable moments of their lives.


5 steps to help CIOs land a board seat

Serving on a board isn’t an extension of an operational role. One issue CIOs face is not understanding the difference between executive management and governance, Stadolnik says. “They’re there to advise, not audit or lead the current company’s CIO,” he adds. In the boardroom, the mandate is to provide strategy, governance, and oversight, not execution. That shift, Stadolnik says, can be jarring for tech leaders who’ve spent their careers driving operational results. ... “There were some broad risk areas where having strong technical leadership was valuable, but it was hard for boards to carve out a full seat just for that, which is why having CIO-plus roles was very beneficial,” says Cullivan. The issue of access is another uphill battle for CIOs. As Payne found, the network effect can play a huge role in seeking a board role. But not every IT leader has the right kind of network that can open the door to these opportunities. ... Boards expect directors to bring scope across business disciplines and issues, not just depth in one functional area. Stadolnik encourages CIOs to utilize their strategic orientation, results focus, and collaborative and influence skills to set themselves up for additional responsibilities like procurement, supply chain, shared services, and others. “It’s those executive leadership capabilities that will unlock broader roles,” he says. Experience in those broader roles bolsters a CIO’s board résumé and credibility.


Microservices Without Meltdown: 7 Pragmatic Patterns That Stick

A good sniff test: can we describe the service’s job in one short sentence, and does a single team wake up if it misbehaves? If not, we’ve drawn mural art, not an interface. Start with a small handful of services you can name plainly—orders, payments, catalog—then pressure-test them with real flows. When a request spans three services just to answer a simple question, that’s a hint we’ve sliced too thin or coupled too often. ... Microservices live and die by their contracts. We like contracts that are explicit, versioned, and backwards-friendly. “Backwards-friendly” means old clients keep working for a while when we add fields or new behaviors. For HTTP APIs, OpenAPI plus consistent error formats makes a huge difference. ... We need timeouts and retries that fit our service behavior, or we’ll turn small hiccups into big outages. For east-west traffic, a service mesh or smart gateway helps us nudge traffic safely and set per-route policies. We’re fans of explicit settings instead of magical defaults. ... Each service owns its tables; cross-service read needs go through APIs or asynchronous replication. When a write spans multiple services, aim for a sequence of local commits with compensating actions instead of distributed locks. Yes, we’re describing sagas without the capes: do the smallest thing, record it durably, then trigger the next hop. 

No comments:

Post a Comment