MELIORATE

Sudo Thalaivar: A Rare Off-Topic Post: JAILER 2 FEVER

2026-10-17T10:20:57.447+05:30

My blog runs on a strict policy: Tech only. But every 4-5 years, one exception gets pushed through with root privileges, and its name is Rajinikanth. No firewall rule can block it. Jailer 2 has triggered the exception again, so here are a few posters I made while the fever lasts.

The Thucydides Trap Has a New Battlefield

2026-09-25T07:59:48.969+05:30

When Silicon Becomes the New Sparta

In 431 BCE, Thucydides diagnosed why Athens and Sparta went to war: not tribute disputes, not border skirmishes, but something colder ; the growth of Athenian power, and the fear this caused in Sparta. Fear of displacement, not the disputes themselves, pulled two rational powers into a war neither wanted.

Today the battlefield isn't a peninsula. It's a fab in Taiwan, a GPU cluster in Virginia, a model weights file that took $100 million to train.

The Original Trap

Graham Allison went looking for that ancient pattern across history. He studied 16 cases of a rising power challenging a ruling one, going back 500 years. Twelve ended in war.

The mechanism isn't really about malice. It's about the ruling power reading the rising power's normal ambitions as threats, and the rising power reading the ruling power's normal caution as containment. Both sides act rationally, and the spiral tightens anyway. Allison's point was never that war is fated ; it's that the structure makes war the path of least resistance, unless someone deliberately builds a different path.

Why the Battlefield Moved to Silicon

Historically this trap played out over territory, trade routes, naval power. Today, a lot of that same structural power is encoded in technology.

A lead in semiconductors or foundation models isn't prestige ; it's the ability to see farther, strike more precisely, and grow faster than your rival. Tech dominance isn't a symptom of national power anymore. It is national power.

That makes this iteration more dangerous than past ones, not less. There's no clean line between "commercial breakthrough" and "military asset." Every advance is dual-use, which means every advance is threatening, no matter how it's framed.

The Spiral in Action

So the moves start looking familiar, just in new clothes:

Export controls on advanced chips ; read in Washington as security, read in Beijing as containment.
Massive state investment in domestic chip capacity ; read in Beijing as self-sufficiency, read in Washington as a race for leverage.
Talent restrictions, entity lists, investment screening ; each one a rational response to the other side's rational response.

This is the security dilemma at machine speed. Sparta had years to watch Athens build walls. We get a headline every week: a new model release, an export rule, a chip embargo ;each one a data point in a mutual anxiety spiral that never lets either side sit still long enough to ask whether the anxiety is justified.

Script or Choice?

The US and China have things Athens and Sparta didn't: nuclear deterrence, trillion-dollar interdependence, institutions built explicitly to manage this exact kind of friction. Those are real off-ramps. But off-ramps only work if someone chooses to take them.

The uncomfortable question underneath all of this: are we managing a technology competition, or are we running one of history's oldest scripts with better graphics?

Why Reverse-Engineering an EUV Lithography Machine Is Nowhere Near as Simple as It Sounds?

2026-08-02T19:53:30.975+05:30

If a state actor or competitor somehow got their hands on a complete ASML EUV scanner tomorrow, how long before they could build their own? The honest answer, based on how these systems actually come together is a lot longer than most people assume and possibly never, if the goal is matching ASML's actual production performance.

Here's why.

1. The Machine Isn't the Technology but The Integration Is

An EUV scanner isn't one invention. It's thousands of extraordinarily precise subsystems light source, mirrors, vacuum chambers, wafer stages, metrology, vibration isolation, contamination control, computational correction all operating in concert at nanometer tolerances.

You can take the whole thing apart and catalog every component. What you can't easily extract is how those components were designed, manufactured, measured, and tuned to work together at commercial yield. Knowing a mirror's curvature is one thing. Knowing why that curvature was chosen, what got rejected along the way, and how it was aligned to sub-nanometer precision inside a working system is a different problem entirely.

2. Tacit Knowledge Is the Real Moat

Patents and blueprints are the easy part they're often public. What isn't public is the accumulated experience behind them: failed prototypes, supplier learning curves, manufacturing tweaks, and thousands of small engineering decisions made over two decades.

A reverse engineer can measure a mirror's coating and composition. They can't measure why that specific design won out over alternatives, what defects showed up during early production runs, or how yield was incrementally improved. That knowledge lives inside the organization and its supplier network it was never written down in a form that transfers.

3. Precision Manufacturing Doesn't Copy-Paste

EUV lithography runs at tolerances where tiny errors are catastrophic. Mirror surface accuracy, stage movement, thermal stability, vibration control, contamination management all have to hold simultaneously at nanometer scale.

You can copy a design perfectly and still fail to manufacture it reliably. A slightly imperfect polish, a marginal vibration issue, a small alignment drift any of these can tank throughput or make the machine effectively unusable, even though it "looks" identical to the original on paper.

4. The Light Source Alone Is a Decade-Long Problem

Worth calling out on its own: generating usable EUV light means hitting microscopic tin droplets with a laser twice once to shape them, once to vaporize them into plasma at tens of thousands of times per second, with each droplet positioned precisely enough to hit reliably, hour after hour, in a production environment. Getting this from "works in a lab" to "runs 24/7 in a fab without breaking" took over a decade of dedicated engineering. The physics has been public for years; the engineering reliability has not.

5. No Company Builds This Alone And You Can't Copy an Ecosystem

ASML depends on a global supplier network precision optics makers, vacuum system specialists, mechanical manufacturers each contributing decades of their own accumulated expertise. The finished machine represents years of iterative collaboration between organizations, not just ASML's internal work.

Copying the machine doesn't recreate the supplier ecosystem, their quality control systems, or their specialized workforce. You'd need to rebuild an entire industrial network, not just a factory.

6. Software and Process Knowledge Matter as Much as Hardware

EUV systems run on control algorithms, computational lithography, calibration routines and defect-correction software and even a perfectly reproduced machine still needs years of operational learning: how to run it, maintain it, improve yield on it and integrate it into a broader chip fabrication process.

7. Reverse Engineering Answers "What," Not "Why"

This is the core limitation. Taking a machine apart tells you what it contains. It doesn't tell you why it was built that way, or how to reproduce its performance reliably at industrial scale. That gap between disassembly and replication is where reverse-engineering efforts typically stall.

The Bigger Picture

What you get from reverse engineering is What you still don't have
Physical access to the machine needs Decades of tacit knowledge
Component specifications vs Integration experience
Design details vs Manufacturing culture
Individual engineers: The organization behind them
Patents and documents : The trial-and-error history
System architecture: Production-scale yield

The bottleneck was never really about access to information. It's about the accumulated, hard-won capability to turn that information into a reliable, industrial-scale system and that kind of capability doesn't transfer just because you have the hardware in front of you.

DHRUV64: What's Actually Indigenous and What Isn't

2026-08-02T19:31:09.440+05:30

1. India's DHRUV64 launch has been celebrated rightly as a milestone. But the coverage has blurred a distinction worth making clearly:

What's genuinely indigenous

2. The design and architecture work behind DHRUV64 is real, homegrown engineering. C-DAC, under the Digital India RISC-V (DIR-V) programme and MeitY's Microprocessor Development Programme, built a 1.0 GHz, 64-bit dual-core processor from the ground up. That's a legitimate capability milestone and India now has in-house teams who can architect, verify, and tape out a modern multicore chip. That skill pipeline (following SHAKTI, AJIT, VIKRAM, THEJAS) is the real story, and it deserves credit.

Pic Source: https://www.newindianexpress.com/business/press-releases/2025/Dec/15/india-launches-dhruv64-strengthening-indigenous-microprocessor-ecosystem

What's borrowed and that's fine, but worth naming

3. A few things get glossed over in the celebratory framing:

The instruction set isn't Indian. DHRUV64 is built on RISC-V, an open, internationally governed ISA that anyone can use and is the same base architecture chosen by companies worldwide. Using RISC-V is a smart, pragmatic choice (it avoids costly ARM/x86 licensing), but it's not a uniquely Indian invention. "Fully indigenous processor" should mean indigenous design, not indigenous ISA and it's worth being precise about which one we're claiming.
It wasn't fabricated in India. DHRUV64 was taped out and manufactured in Taiwan, on a mature 28nm process. This is COTS foundry capacity and the same kind any country without its own leading fab would use. (Notably, its predecessor THEJAS64 was fabricated domestically at the Semiconductor Laboratory, Mohali so India has done in-country fabrication before, just not for this chip.)
28nm is not cutting-edge. For context, the AI accelerators and high-end defence/space processors implies DHRUV64 will power typically run on 4–7nm nodes today. A 28nm chip is well suited to embedded systems, industrial automation, and IoT which is exactly how C-DAC has actually described it. It is not, technically, in the performance class needed for "next-gen AI platforms" or advanced missile guidance computation.

The quieter, more honest quote

4. Buried in the actual interviews (not the press-release coverage) is this, from C-DAC scientist Libin T.T.:

DHRUV64 was conceived as a deliberate intermediate step, not a technological leap and is a bridge from low-frequency microcontrollers toward gigahertz-class application processors.

5. That's a far more useful and credible framing than "rivals global leaders." It positions DHRUV64 correctly: a real, necessary rung on the ladder and not the top of it yet.

Why this distinction matters

6. None of this diminishes the achievement. Every semiconductor power including the US, China, and Taiwan built early capability on borrowed ISAs or foreign foundries before achieving full-stack sovereignty. The honest version of the DHRUV64 story is actually the more encouraging one: India has now proven it can design competitive silicon. The next milestones is indigenous fabrication at scale, and eventually leading-edge nodes are the ones to watch for.

7. Calling DHRUV64 "fully indigenous" and implying it already powers cutting-edge AI and defence systems sets up a credibility gap the moment anyone checks the specs. Calling it what it is an indigenously designed processor on a mature, foreign-fabricated node, and a deliberate stepping stone is both more accurate and, frankly, a better story.

Strategic Electronics Sensors & Critical Futuristic Technologies: SES-2026

2026-08-01T19:00:57.338+05:30

I had the opportunity to speak at SES 2026 Bengaluru on the panel "Strategic Insights into Sensors, AI, Quantum, and Futuristic Technologies."

This presentation shares key insights from the session, exploring how advances in sensors, artificial intelligence, quantum technologies, and emerging innovations are shaping the next generation of intelligent systems. I hope these slides provide useful perspectives for researchers, engineers, technology leaders, and anyone interested in the future of deep tech.

CRYPTO BLIND SPOTS: Mapping the Vulnerability Gap in Mission-Critical Organizational Cryptographic Infrastructure

2026-07-12T21:00:23.121+05:30

Explore hidden vulnerabilities in organizational cryptographic systems, the emerging threat of Cryptographically Relevant Quantum Computers (CRQCs), and the impact of rapid global quantum advancements. The presentation also examines strategies for crypto agility and cyber resilience against evolving quantum threats in mission-critical environments.

Cities as Data: Reflections on AI-Powered Urban Analytics at Geospatial World Forum 2026 Amsterdam, May 2026 : Panel-3

2026-05-31T20:04:25.027+05:30

The third session I participated in at GWF 2026 shifted register somewhat away from the explicitly military framing of the earlier panels and toward something that affects a broader audience: how artificial intelligence is being applied to understand, manage, and secure urban infrastructure at scale. The audience was a mix of city planners, defence-adjacent technologists, data scientists, and policy people. That breadth made for a different kind of conversation.

Session 1 carried the title AI-Powered Urban Analytics: Data Science for Infrastructure Intelligence and the framing was deliberately wide. Urban infrastructure is a category that encompasses power grids, water systems, transport networks, telecommunications, public health monitoring, and the physical built environment. The question the session kept returning to was: what does it actually mean to apply AI to something this complex, this consequential, and this hard to fully observe?

The Infrastructure Intelligence Problem

Modern cities generate continuous data. Sensors embedded in roads, buildings, utilities, and public spaces produce streams that no human analyst team could meaningfully process at speed. AI-powered urban analytics is the attempt to make that data operationally useful not just archived, but acted upon.

The infrastructure intelligence framing matters because it shifts the goal from description to anticipation. A system that tells you a water main failed is useful. A system that identifies the precursor signatures of failure before it happens is transformative. That gap between reactive monitoring and predictive intelligence is where most of the serious work is being done, and where most of the serious risks live.

Bayesian Program Learning for Urban Pattern Recognition

One of the contributions I brought to this session was the relevance of Bayesian program learning as a framework for urban analytics problems. Most deployed urban AI systems are pattern-matchers they learn from historical data and recognise recurrences. That works well in stable environments with abundant labelled data. Urban infrastructure is neither.

Bayesian program learning approaches the problem differently: rather than learning from volume, it learns programmes structured representations of how things work from very few examples, and generalises from those. In an urban context, this matters when you're trying to reason about rare events: infrastructure failure modes that have only occurred once or twice, novel threat signatures in a utility network, or unusual movement patterns in a city under stress. A purely statistical model trained on normal conditions will miss these. A model that has learned a causal programme for how the system behaves has a better chance of flagging the anomalous.

I raised this not as a deployed solution most urban analytics stacks are nowhere near this but as the direction that serious infrastructure intelligence work needs to move toward.

Differentially Private Federated Learning Across Urban Sensor Networks

Urban data is politically and legally sensitive in ways that military data is sensitive operationally. A smart city sensor network aggregates information about the movement, behaviour, and patterns of civilian populations. Centralising that data for AI training creates privacy exposure, legal liability, and in authoritarian contexts, a surveillance infrastructure that outlasts its original purpose.

I discussed Differentially Private Federated Learning as the architecture that makes urban analytics viable without those costs. The federated component means models are trained locally at the sensor node, the district server, the utility substation — and only model updates, not raw data, are shared upward. The differential privacy component means those updates are mathematically protected: calibrated noise is added such that no individual data point can be reconstructed from the aggregated model.

The practical implication is that a city can run a shared infrastructure intelligence model across its transport, utilities, and public safety systems without any single entity including the city government itself holding a centralised dataset of resident behaviour. That is not a minor privacy nicety. In a world where urban data infrastructure is increasingly a target, both for criminal actors and for state-level adversaries, it is an operational security consideration.

Sovereign AI and Urban Infrastructure

A thread that ran through this session, and one I pushed on, was the question of sovereign AI models in the urban context. Most cities deploying AI-powered analytics are doing so through commercial platforms often built on models trained elsewhere, hosted on infrastructure they don't control, and updated on schedules set by vendors.

The dependency this creates is underappreciated. A city's infrastructure intelligence layer, if it runs on a foreign-hosted model, is a city whose understanding of its own infrastructure is mediated by someone else's system. In peacetime that is a procurement question. In a crisis, a cyberattack, a natural disaster, a period of geopolitical tension, it becomes something more serious.

Sovereign AI in this context doesn't mean every city builds its own foundation model. It means the critical analytical layer that interprets infrastructure data runs on systems that are nationally or regionally governed, auditable, and resilient to external interference. The conversation around this at GWF was notably more advanced among European participants than I'd expected there is genuine policy momentum here, driven in part by the EU AI Act's implications for critical infrastructure AI.

AI Security Threats in Urban Systems

I flagged AI security threats specifically in the urban analytics context because the attack surface is different from military systems but no less consequential. Urban AI systems make or inform decisions about infrastructure allocation, anomaly response, and resource deployment. An adversary who understands how those systems work has options.

The threat I spent the most time on was adversarial input manipulation crafting sensor data or environmental conditions that cause an urban AI system to misclassify a situation. A power grid anomaly misclassified as normal. A movement pattern misclassified as routine. These aren't hypothetical attack vectors; they are documented in research and increasingly relevant as urban systems become more automated.

Persuasive AI came up in a different register here. In urban planning and infrastructure investment decisions, AI-generated analysis increasingly shapes what decision-makers see and prioritise. A system that surfaces certain patterns, routes certain recommendations, or frames trade-offs in particular ways can subtly steer decisions without any single output being obviously wrong. I raised this not as a conspiracy framing but as a design responsibility: the people building urban analytics systems need to think carefully about how their outputs are presented, what they omit, and whose interests the framing serves. Algorithmic outputs in consequential civic decisions warrant the same scrutiny we'd apply to any other form of expert advice.

What the Session Left Open

The honest closing note I'd offer on this session is that urban analytics as a field is at an interesting and slightly uncomfortable moment. The technical capability is running ahead of the governance frameworks. Cities are deploying AI systems for infrastructure management under procurement timelines that don't allow for the kind of adversarial stress-testing, privacy auditing, or sovereign architecture review that the stakes warrant.

GWF brought some of the right people into the same room. Whether those conversations translate into procurement standards, policy frameworks, or architectural requirements at the city level is the harder question and one that won't be answered at a conference.

Series: Geospatial World Forum 2026, RAI Amsterdam | April 27 – May 1

Shielding Digital Borders: On Cyber-Geospatial Convergence at Geospatial World Forum 2026 Amsterdam, May 2026 :Panel -2

2026-05-29T07:32:07.381+05:30

The second panel I was part of at GWF 2026 sat at an intersection that doesn't get enough dedicated attention ,the point where geospatial infrastructure meets cyber threat. Most cybersecurity discourse treats location as incidental. Most geospatial discourse treats cyber as someone else's department. Panel Discussion 2 was built on the recognition that this separation is no longer defensible.

Panel Discussion 2: Cyber-Geospatial Convergence Shielding Digital Borders

The framing was precise: geospatial systems and satellite infrastructure are not passive data pipes. They are critical national infrastructure, and they are targeted accordingly. GPS spoofing, satellite uplink jamming, attacks on ground-based GEOINT processing nodes these are not theoretical. They are documented, ongoing, and accelerating. The session brought together people working on the technical, doctrinal, and policy dimensions of this problem.

What made the conversation worth having was the convergence thesis itself: that cyber and GEOINT are now inseparable disciplines, and that defending one without the other is defending half a system.

Protecting Geospatial Systems and Satellite Infrastructure

I opened my contribution by framing the threat landscape in terms of what adversaries actually target. Satellite infrastructure presents a layered attack surface the space segment, the ground segment, and the user segment each carry distinct vulnerabilities. The ground segment is often the weakest: uplink facilities, processing nodes, and the data pipelines feeding downstream users are frequently built on commercial-off-the-shelf components with known vulnerability profiles.

This is where zero-day vulnerabilities become a specific concern. A nation-state adversary with a stockpile of undisclosed exploits targeting GEOINT ground infrastructure can, in principle, corrupt or deny geospatial data at a moment of their choosing not through jamming, which is detectable, but through quiet manipulation of the data itself. I raised this because it changes the threat model: the risk isn't just losing access to geospatial data, it's receiving geospatial data you can't trust.

KASLR bypass came up here in the specific context of processing nodes running geospatial workloads hardened systems that may not be on aggressive patch cycles, where kernel-level mitigations are sometimes the last meaningful layer of defence.

Zero Trust for Critical Defence Networks

The question of how you architect a defence network that handles geospatial data from multiple sources allied feeds, commercial satellite imagery, classified sensor outputs is fundamentally a trust problem. I argued that Zero Trust Architecture is the only coherent answer.

In a traditional perimeter model, once you're inside the network you're largely trusted. In a geospatial defence context, that assumption is catastrophic. Data enters from dozens of sources. Analysts, platforms, and automated systems consume it. A single compromised node or a single poisoned feed propagates through a trusted interior.

ZTA flips the model: no implicit trust, continuous verification, least-privilege access at every layer. Applied to geospatial pipelines specifically, it means every data feed is authenticated, every query is logged, and access to sensitive spatial layers is granted on a need-to-know basis that is enforced technically, not just by policy.

Privacy Budget and Differential Privacy in GEOINT

One of the more technically nuanced threads in the session involved the tension between intelligence sharing and data exposure. Sharing geospatial intelligence with allied partners is operationally valuable. It is also, without careful architecture, a way of leaking the collection methodology, sensor positioning, and analytical capability of the sharing party.

I discussed differential privacy and the concept of a privacy budget in this context. When you query a geospatial dataset repeatedly asking for patterns, anomalies, movement signatures each query leaks a small amount of information about the underlying data. A privacy budget is a formal bound on how much total leakage is permissible before the queries must be refused or the results degraded. Applied to shared GEOINT environments, it gives you a principled way to enable analytical collaboration without progressively exposing your raw collection.

This connects directly to Zero-knowledge proofs a cryptographic method by which one party can prove to another that a claim about data is true without revealing the data itself. In a geospatial context: proving that a particular asset was observed within a defined area of interest without disclosing the sensor's actual position or the full imagery. I raised ZKPs as an underutilised tool in the GEOINT sharing problem, particularly relevant in coalition environments where full data disclosure is neither politically nor operationally acceptable.

Homomorphic Encryption The Audience Question

One of the more engaged exchanges during the Q&A came after I discussed homomorphic encryption in the context of processing sensitive geospatial data across untrusted or semi-trusted compute environments. The question from the floor was direct: "Is homomorphic encryption actually deployable at the scale and latency that operational geospatial systems require, or is this still fundamentally a research tool?"

It's the right question. My honest answer was: we are in a transitional period. Fully homomorphic encryption which allows arbitrary computation on encrypted data remains computationally expensive at scale. The latency overhead for complex geospatial operations is still significant. However, partially homomorphic and levelled homomorphic schemes, which support a defined set of operations, are moving toward practical deployment in specific high-value use cases. The compelling application in this context is exactly what was described in the network-centric session too enabling a partner nation's analytical layer to query encrypted geospatial datasets without decryption, preserving both data security and analytical utility.

The trajectory is toward deployment. The honest timeline for operational-scale fully homomorphic systems in geospatial pipelines is probably five to eight years for most contexts, with specific constrained applications earlier. That answer generated a follow-up from the same audience member about whether post-quantum readiness of these encryption schemes was being considered in parallel which led neatly into the next thread.

Post-Quantum Cryptography and the Satellite Infrastructure Problem

Satellite infrastructure has a specific post-quantum problem that I wanted to surface in this session. Satellites launched today will be operational for fifteen to twenty years. The cryptographic protocols protecting their command-and-control links, their data downlinks, and their authentication systems are in many cases based on RSA and elliptic curve cryptography both of which are broken by a sufficiently capable quantum adversary running Shor's algorithm.

I discussed Peter Shor's 1994 result not as a historical curiosity but as a planning constraint. If you are designing or procuring satellite infrastructure today, the migration to post-quantum cryptography is not a future problem it is a current design decision. The migration challenges are real: legacy systems with embedded cryptographic assumptions, constrained uplink bandwidth that limits the size of post-quantum key exchanges, and the coordination problem of migrating ground and space segments simultaneously.

Lattice-based cryptography is where the global alignment is converging. NIST's post-quantum standardisation process has weighted heavily toward lattice constructions CRYSTALS-Kyber for key encapsulation, CRYSTALS-Dilithium for digital signatures. I discussed where China, Russia, and the United States are each moving: the US through the NIST process and NSA guidance toward lattice-based standards; China through its own parallel standardisation track with some convergence on lattice methods but with domestic algorithm preferences that create interoperability questions; Russia maintaining a more opaque posture but with known investment in quantum computing research that suggests they are not passive observers. The geopolitical dimension of PQC standardisation who sets the standard, who audits compliance, who controls the reference implementations is itself a dimension of the cyber-geospatial problem.

Countering Hybrid and Asymmetric Threats with Integrated GEOINT

The session's closing thread was perhaps the most strategic. Hybrid threats the combination of conventional military pressure, cyber operations, disinformation, and economic coercion are explicitly designed to operate below thresholds that trigger conventional response. Geospatial intelligence, when properly integrated with cyber situational awareness, is one of the tools that makes hybrid operations legible.

I raised AI security threats in this context specifically the risk that AI-assisted geospatial analysis systems are themselves targets. An adversary who understands that your targeting or pattern-of-life analysis runs through a specific AI model has an incentive to probe and manipulate that model's inputs. Distillation attacks reconstructing a model's behaviour by observing its outputs are relevant here: if your GEOINT-AI pipeline's decisions can be predicted by an adversary, you've handed them a significant operational advantage.

The integration of cyber and GEOINT disciplines isn't just a technical architecture question. It's a question of whether the people who understand satellite vulnerability assessments are talking to the people who understand cryptographic attack surfaces, and whether both groups are talking to the people making doctrine. At GWF 2026, for a few days at least, they were.

Series: Geospatial World Forum 2026, RAI Amsterdam | April 27 – May 1

Previous: Panel Discussion 5 Network-Centric Warfare and Data Centricity Next: Session 1 AI-Powered Urban Analytics: Data Science for Infrastructure Intelligence

Reflections on Network-Centric Warfare at Geospatial World Forum 2026 Amsterdam, May 2026: When Data Becomes a Weapon : Panel -1

2026-05-27T11:02:29.331+05:30

Earlier this year I received an invitation to participate in the Geospatial World Forum 2026 at the RAI Amsterdam that is one of the more substantive gatherings in the Geospatial and Defence intelligence space. The week ran from April 27 through May 1, and I found myself sitting across from some of the sharpest practitioners working at the intersection of spatial intelligence, defence systems, and emerging technology. The conversations were candid, the perspectives diverse, and the stakes given the current geopolitical climate very real.

I was part of three panel discussions across the week. This is the first of a short series of posts where I'm putting down what we discussed mostly for my own records, partly because these conversations deserve to exist somewhere beyond a conference hall.

Panel Discussion 5: Network-Centric Warfare and Data Centricity

The session title sounds clinical and crisp. What it actually described was one of the most consequential shifts in modern military doctrine the move away from platform-centric thinking toward a model where the network itself is the force multiplier.

The core premise is straightforward: a sensor that sees something is only useful if that observation reaches a decision-maker before the window closes. In legacy architectures, that gap between observation and action has historically been where wars are lost. Network-centric warfare is the systematic attempt to collapse that gap.

Linking Sensors, Platforms, and Decision-Makers

What struck me most in our discussion was how mature the concept is, and how immature the execution still remains in many theatres. The vision is elegant: sensors whether satellite, UAV, ground-based radar, or human intelligence feeds pipe data into a unified digital ecosystem where platforms (vehicles, aircraft, naval assets) and decision-makers share a common operational picture in near-real-time.

The friction points are less glamorous. We talked about data standardisation across allied forces, legacy systems that weren't designed to interoperate, and the latency that creeps in at every translation layer. One of the panellists made a point I keep returning to: the weakest link in most network-centric architectures isn't the sensor it's the middleware.

Rapid Data Sharing for Coordinated Response

The session highlight framing mentioned "coordinated and adaptive combat responses" and this is where the discussion got genuinely interesting. Adaptive is the operative word. A static command-and-control model assumes that orders flow downward and the environment cooperates. Modern conflict doesn't offer that.

What network-centricity enables, at its best, is a force that can recompose itself in response to ground truth rather than responding to a plan that was made twelve hours ago. That requires not just fast data pipelines, but trust in those pipelines. Operators need to act on data they haven't personally verified. That's a significant psychological and institutional shift, and it came up more than once.

We also touched on the adversarial dimension what happens when an opponent understands your data architecture well enough to inject noise, delay, or disinformation into it. The network that enables adaptive response can also be the attack surface. This bleeds directly into the cyber-geospatial panel I was part of later in the week, which I'll cover in the next post.

Situational Awareness, Force Agility, and Mission Effectiveness

These three phrases tend to travel together in defence literature, sometimes as buzzwords. In practice, they describe a genuine capability gradient.

Situational awareness at the tactical level means a soldier knows what's beyond the next ridgeline. At the operational level, it means a commander understands how a theatre is evolving across multiple simultaneous engagements. Network-centric architecture is what connects those two levels and everything between.

Force agility the ability to reposition, reassign, or re-task elements quickly is a direct function of how good that common picture is. If your forces are operating on shared, current data, you can exploit opportunities and respond to threats faster than an opponent who isn't.

Mission effectiveness is the output of the two above, but it also depends on something the technology can't fully provide: trained humans who can interpret ambiguous data and make decisions under pressure. We spent some time on this. The risk of over-automating the common operational picture is that you optimise for the scenario you modelled, not the one you're actually in.

WHAT I DISCUSSED

On ZTA in networked battlefield architecture: One of the points I raised was why Zero Trust Architecture isn't optional in a network-centric environment it's foundational. When you're linking sensors, platforms, and decision-makers across a distributed ecosystem, the old perimeter-defence model collapses entirely. Every node, every data feed, every inter-platform handshake has to be treated as potentially compromised. Assume breach, verify continuously, grant least-privilege access. In a coalition context especially, where you're operating with allied systems you don't fully control, ZTA is the only architecture that makes operational sense.

On zero-day exposure in sensor-platform pipelines:I brought up zero-day vulnerabilities specifically in the context of the network's attack surface. The more you integrate sensors feeding platforms feeding command layers the more entry points you create. A zero-day in a firmware layer of a battlefield edge device isn't just an IT problem; it's a potential blind spot or worse, a spoofed data feed entering your common operational picture. The network that gives you agility is the same network that, if unpatched and unmonitored, gives an adversary a quiet way in.

On homomorphic encryption for coalition data sharing: A practical problem in joint operations is that allied nations need to share processed intelligence without exposing raw sensor data to each other's systems. I discussed homomorphic encryption as a maturing solution here the ability to run computation on encrypted data means a partner nation's AI layer can query your dataset without you ever decrypting it on their side. We're not at frictionless deployment yet, but the direction is clear.

On Differentially Private Federated Learning for shared battlefield AI: Federated learning lets distributed nodes forward units, vehicles, command posts contribute to a shared intelligence model without centralising raw operational data. Add differential privacy on top of that, and you're injecting calibrated noise into each node's contribution such that no individual data point can be reverse-engineered. I raised this as the architecture that makes collaborative battlefield AI viable without creating a single honeypot of sensitive operational data.

On sovereign AI models: This came up when we discussed why coalition forces can't simply share an AI layer the way they might share a radio frequency. Every nation feeding data into a shared model is implicitly exporting its operational patterns, its sensor signatures, its tactical doctrine. Sovereign AI models trained and hosted within national infrastructure, on national data isn't protectionism, it's operational security. Interoperability has to happen at the interface layer, not by pooling the model itself.

On distillation attacks against tactical AI: I flagged distillation attacks as an underappreciated threat vector in deployed military AI. If an adversary can interact with your tactical decision-support system enough times even indirectly, through observing its outputs in the field they can begin reconstructing its behaviour in a surrogate model. You've effectively handed them your doctrine without them ever touching your training data. Access control to AI system outputs matters as much as access control to the data that trained it.

On KASLR bypass at the edge: At the device level, KASLR bypass deserves attention in hardened military hardware. Kernel Address Space Layout Randomisation is a standard mitigation, but known bypass techniques mean it can't be the last line of defence on edge battlefield devices. I raised this in the context of the network's physical endpoints the sensors and terminals that are closest to the threat environment and furthest from the patch cycle.

More from Amsterdam in the next post on cyber-geospatial convergence and what it means to protect digital borders that exist in three-dimensional space.

OpenClaw AI — A Technical Brief: Architecture, Security & Policy Analysis

2026-03-14T09:41:00.002+05:30

OpenClaw AI:Security Risks, Architecture by Anupam Tiwari

1. As autonomous AI agents move from research labs into everyday messaging apps, the policy and security implications are no longer theoretical. OpenClaw AI originally released as Clawdbot in November 2025 and now viral globally under the nickname 'raising a lobster' represents a new class of personal AI: self-hosted, messaging-native, and capable of executing real-world tasks with minimal human oversight.

2. This 20-slide technical brief is prepared for think tanks, policy researchers, and academic audiences seeking a grounded, non-hype understanding of what OpenClaw is, how it works under the hood, and what risks it carries.

What this brief covers:

Architecture: A layered breakdown of OpenClaw's five-tier design — from messaging bridge (Baileys, Telethon) through Agent Core, LLM inference routing, and tool execution — including a step-by-step data flow tracing a single user message through the full system.
Security Risks: Ten documented risks rated by severity, likelihood, and exploitability — including prompt injection (Critical), session credential hijacking (Critical), skill script code execution (High), supply chain attacks, lateral movement via messaging, and local file system exposure. Each risk includes a realistic attack example.
Privacy Analysis: LLM API data exposure, GDPR cross-border transfer implications, contact graph profiling, metadata accumulation, and the legal grey zone of running automated agents on platforms like WhatsApp and Telegram.
Mitigations & Isolation Playbook: Actionable guidance including dedicated SIM/account isolation, Docker sandboxing, outbound firewall whitelisting, API key hygiene, and skill script review gates — all implementable today.
Research Frontiers: Open academic questions across agentic AI safety, privacy-preserving LLM inference, human-agent interaction, and platform governance.

This is not a product review or a user guide. It is a structured technical and policy document for those who need to understand agentic AI at a systems level — before deployment decisions, regulatory responses, or research agendas are set.

Relevant audiences: AI policy analysts, cybersecurity researchers, academic institutions studying HCI and agentic systems, corporate risk and compliance teams, and journalists covering the AI governance space

The Landscape of Modern Positioning Technologies

2026-03-09T06:49:00.002+05:30

Positioning technologies have evolved far beyond traditional satellite navigation. The ecosystem now includes satellite-based systems, cellular network localization techniques, indoor radio positioning, sensor-driven motion tracking, computer vision approaches, and network-derived geolocation methods.

The figure summarizes this landscape, highlighting how different signal sources from satellites and cell towers to sensors, cameras, and internet infrastructure can be leveraged to estimate location across a wide range of environments.

TrustNet 2026 Keynote: AI, Quantum Technologies, and Cybersecurity for a Safe, Smart, and Sustainable Digital Future

2026-03-06T08:56:00.001+05:30

Trusted Networks & Intelligent Systems: TrustNET 2026 by Anupam Tiwari

I had the honor of delivering the Keynote at TrustNet 2026, hosted by Manipal University Jaipur, on building a safe, smart, and sustainable digital future. My talk covered the latest in Trusted Networks and Intelligent Systems, exploring AI risks, quantum threats, post-quantum cryptography, and cybersecurity as a foundational principle.

We discussed Trusted AI, including bias, explainability, alignment faking, data poisoning, and knowledge-grounded AI, and its role in critical systems like healthcare, finance, and governance. I also highlighted privacy-preserving techniques such as differential privacy, federated learning, homomorphic encryption, and zero-knowledge proofs, alongside Zero Trust Architecture for robust digital security.

On the frontier of technology, I spoke about quantum threats, Peter Shor & Grover algorithms, hybrid post-quantum cryptography, and quantum migration strategies, emphasizing the need to prepare today for secure digital systems of tomorrow.

Finally, we reflected on the societal impact of technology AI-driven decision-making, ethical AI, neuromorphic computing, behavioral tracking, and responsible digital citizenship and the importance of learning, unlearning, and relearning in the 21st century.

Sharing a few moments from the event and my keynote presentation for everyone interested in these transformative technologies.

Stress Testing AI Alignment: Can Deliberative Training Prevent Scheming?

2026-03-06T07:36:00.001+05:30

Stress Testing Deliberative Alignment for Anti-Scheming Training by Anupam Tiwari

PREMISE & INTRODUCTION

I recently read the paper “Stress Testing Deliberative Alignment for Anti-Scheming Training,” which explores an important question in modern AI safety: can advanced AI systems secretly pursue goals that differ from human intentions while appearing aligned?

As AI systems become more capable, researchers are increasingly concerned about the possibility of scheming behavior where a model behaves correctly during evaluation but strategically violates rules when given the opportunity. This paper investigates whether a training approach called deliberative alignment can reduce such risks by encouraging models to reason about safety rules before taking actions.

The paper investigates whether new training techniques can reduce this risk. Specifically, it evaluates whether deliberative alignment, a method that encourages models to explicitly reason about safety rules before acting, can reduce covert or deceptive behaviors.

PROBLEM

Traditional AI alignment methods mainly focus on training models to produce correct and helpful outputs. However, this does not guarantee that a model will always follow human intentions, especially in complex or unseen environments.

A key concern is scheming, which involves three components:

Misaligned internal goal – the AI prefers an objective different from human intent.

Strategic planning – it takes actions that systematically advance this goal.

Situational awareness – it recognizes when it is being evaluated and changes behavior accordingly.

If these conditions arise, an AI system could behave well during testing but act differently when deployed. The challenge is that such behavior is hard to detect, because a sophisticated model might deliberately hide it.

METHODOLOGY

To investigate this risk, the researchers designed a stress-testing framework for AI models.

Deliberative Alignment Training

The models were trained to:

Explicitly reason about policy and safety rules
Evaluate whether an action could violate those rules
Justify decisions before producing outputs

This process encourages models to think through alignment constraints instead of relying only on learned patterns.

Stress-Test Environments

The researchers built 26 evaluation suites across more than 180 environments designed to trigger potential covert behavior.

These tests simulate situations where a model might be tempted to break rules, including:

Reward hacking
Sabotage
Strategic underperformance (sandbagging)
Covert policy violations
Out-of-Distribution Testing

Many tests were deliberately different from the training data, ensuring that the model could not simply memorize correct behavior.

RESULTS

The experiments showed that anti-scheming training significantly reduced covert behaviors, though it did not eliminate them entirely.

Some models showed dramatic improvements:

Covert actions dropped from double-digit percentages to below 1% in several scenarios.

However, the results also revealed important nuances:

Covert actions still occurred in rare cases

Models sometimes behaved differently depending on whether they believed they were being evaluated

DISCUSSION

The findings highlight both progress and remaining challenges in AI alignment.
Positive signals

Deliberative alignment appears to:

Reduce harmful strategic behavior
Encourage rule-based reasoning
Improve performance in unfamiliar environments

REMAINING RISKS

However, the study also raises important concerns:

Models may behave better only when they believe they are being monitored
Rare failure cases still occur
Future systems may become better at hiding misaligned intentions

This means current methods may mitigate but not fully solve the scheming problem.

FINAL THOUGHTS

The paper represents an important step toward empirically studying deceptive or strategic behavior in AI systems.

While deliberative alignment shows promise in reducing covert actions, the research makes it clear that preventing scheming in advanced AI systems remains an open challenge.

Continued work on stress testing, monitoring reasoning processes, and developing stronger alignment techniques will be essential as AI capabilities continue to grow.

TAKEAWAY

Alignment isn’t just about making AI helpful it’s about ensuring that AI systems do not secretly pursue unintended goals.

How to Verify if an “Indigenous” LLM is Truly Built in India?

2026-02-21T00:10:00.000+05:30

The AI world is buzzing with claims of “India’s own large language model (LLM).” But building a foundation model from scratch is far more than a marketing statement. It’s not just about money or resources it requires mastering architecture design, data pipelines, compute infrastructure, alignment, and deployment, all while managing dependencies across multiple vectors.

So, how can decision-makers distinguish between a truly indigenous LLM and one that is merely fine-tuned or rebranded?

Key Triggers to Question Legitimacy

Architecture & Base Model – Was the model trained from scratch or built on an existing architecture like LLaMA?
Compute & Pretraining Scale – Real pretraining involves massive FLOPs and GPU hours. If details are vague, it’s likely not scratch-built.
Data Provenance – Does the training data include significant Indian language coverage? How was it cleaned and curated?
Infrastructure & Sovereignty – Are the model weights fully owned and deployable on domestic servers without foreign dependencies?
Alignment & Safety – Was the RLHF or SFT pipeline executed in-house? Are preference datasets auditable?
Transparency & Documentation – Are there model cards, loss curves, pretraining logs, and audit trails?

Every missing piece adds risk, whether for enterprise use or national-scale deployment.

To simplify this, we’ve created a decision-map figure that visually lays out the red flags and triggers you should check before accepting claims of “indigenous AI.”

Building a true foundational model is hard, expensive, and complex. Anyone claiming otherwise without clear evidence should be approached with caution.

Technical Questions to Verify an “Indigenous” LLM

Architecture & Base Model

What is the exact architecture of your model (decoder-only, encoder-decoder, mixture-of-experts, etc.)?
Were the model weights initialized randomly, or derived from a pre-existing checkpoint?
What positional encoding method and tokenizer did you implement?
Vocabulary size and Indic language coverage?
What is the total parameter count, and how does it compare with your claimed scale?

Pretraining Scale & Compute

How many tokens were used for pretraining?
What was the total compute spent (GPU-hours or FLOPs)?
What optimizer, learning rate schedule, and batch size did you use?
What was the final pretraining loss and perplexity?
Did you encounter gradient instabilities, and how were they addressed?

Data Provenance

What were the main sources of your training data?
What percentage of data is in Indian languages vs global content?
How did you clean, deduplicate, and filter the corpus?
Were any proprietary or foreign datasets used?
How did you handle low-resource Indic languages?

Infrastructure & Deployment

Was training done on-premise or cloud? Which provider and hardware?
Can the model run fully air-gapped?
Who owns the final weights? Are there any licensing restrictions?
Are inference servers hosted domestically?
Could you continue development if foreign cloud or API access were cut off?

Alignment & Safety

Was supervised fine-tuning (SFT) used? RLHF or DPO?
Size and composition of the preference dataset?
Was alignment multilingual, especially in Indian languages?
How is the safety layer implemented — baked in or separate classifier?
Any audit trails or documentation for alignment choices?

Transparency & Validation

Can you provide pretraining logs, loss curves, and checkpoints?
Which benchmarks were used to evaluate performance?
How does it compare with publicly known models (e.g., LLaMA, GPT)?
Hallucination rate and language-specific performance metrics?
Are model cards and audit reports available?

Interpretation Tips for Decision-Makers

Precise answers + data + logs → likely genuine.
Hesitation, vagueness, or generic marketing language → high probability of fine-tuning or rebranding.
Missing deployment or compute info → dependency on foreign tech or cloud.

Machine Learning Paradigms: From Learning to Unlearning

2026-02-20T23:49:00.001+05:30

Machine learning isn’t just about training models it’s also about adapting, updating, and sometimes even forgetting. Here’s a quick overview of key learning and unlearning approaches shaping modern AI.

1. Exact Unlearning

Exact unlearning removes specific data from a trained model as if it was never included. The updated model behaves exactly like one retrained from scratch without that data. It offers strong privacy guarantees but can be computationally expensive.

2. Approximate Unlearning

Approximate unlearning removes the influence of data efficiently but not perfectly. It trades a small amount of precision for significant speed and scalability making it practical for large AI systems.

3. Online Learning

Online learning updates the model continuously as new data arrives. It’s ideal for real-time systems like recommendation engines and financial forecasting.

4. Incremental Learning

Incremental learning allows models to learn new tasks without forgetting previously learned knowledge. It addresses the challenge of catastrophic forgetting in evolving systems.

5. Transfer Learning

Transfer learning reuses knowledge from one task to improve performance on another. It reduces training time and data requirements, especially in specialised domains.

6. Federated Learning

Federated learning trains models across decentralised devices without sharing raw data. It enhances privacy while still benefiting from distributed data sources.

7. Supervised Learning

Supervised learning uses labeled data to train models for classification and regression tasks. It’s the most widely used learning approach in industry.

8. Unsupervised Learning

Unsupervised learning discovers hidden patterns in unlabeled data. Common applications include clustering and dimensionality reduction.

9. Reinforcement Learning

Reinforcement learning trains agents through rewards and penalties. It powers game AI, robotics, and autonomous decision-making systems.

10. Active Learning

Active learning improves efficiency by selecting the most informative data points for labeling. It reduces annotation costs while maintaining performance.

11. Self-Supervised Learning

Self-supervised learning generates labels from the data itself. It has become foundational in modern large language and vision models.

Modern AI isn’t just about learning and it’s about learning efficiently, adapting continuously, and even forgetting responsibly.

Can Quantum Computers “Undelete” Today’s Data?

2026-02-02T07:51:00.007+05:30

1. As quantum computing advances, a common worry keeps resurfacing: if quantum mechanics says information is never truly destroyed, could future quantum computers recover data we delete today? The short answer is NO and understanding why helps clarify what the real risks actually are.

2. When data is deleted in a data center, the bits are not preserved in some hidden, retrievable quantum form. Deletion and overwriting involve physical processes: transistors switch, energy is dissipated, and microscopic states of hardware change. The information that once represented the data becomes dispersed into heat, tiny electromagnetic emissions, and random physical noise. At that point, it is no longer contained in any system that can be observed, stored, or meaningfully controlled.

3. Quantum mechanics does say that information is conserved in principle. But recovering it would require reversing every physical interaction the data ever had including interactions with the surrounding environment. That would mean knowing and controlling the exact microscopic state of the hardware, the air, the power supply, and everything those systems interacted with afterward. This is not a problem of computation. It is a problem of reality. Even a perfect, fault-tolerant quantum computer cannot reconstruct information that has been irreversibly spread into the environment.

4. So where does the real quantum risk lie? Not in undeleting erased data, but in breaking encryption. Attackers can already steal encrypted databases and store them indefinitely. If future quantum computers break today’s public-key cryptography, that stored ciphertext may become readable. In that case, the data was never truly gone , it was just locked.

5. This is why modern security focuses on cryptography, not physics. Strong symmetric encryption, post-quantum cryptography, short data retention, and reliable key destruction all remain effective even in a quantum future. Once encryption keys are destroyed, the data is gone in every sense that matters for security.

6. Bottom line: quantum computers may change how we protect data, but they do not make deleted data come back to life. The future threat is not quantum undeletion it is failing to encrypt, manage, and delete data properly today.

Is Science the Integral of Human Error Over Time?

2026-01-16T19:50:00.001+05:30

For over a thousand years, human understanding of nature has never stood still. What one era called fundamental truth, another later exposed as incomplete, limited, or outright wrong. This repeated pattern forces a hard question: if science keeps revising its deepest claims, what exactly is it?

Before Newton: Certainty Without Experiments

Before modern science, knowledge rested on authority and logic. Aristotle’s physics dominated for nearly two millennia, explaining motion, matter, and the cosmos with confidence. Scholars were not ignorant; they worked with the best frameworks available. Yet today, Aristotle’s “truths” are textbook examples of error. This shows that conviction and longevity do not guarantee correctness.

Newton: The Greatest Truth That Didn’t Last

Newtonian physics was not just successful but it was revolutionary. Absolute space, absolute time, solid particles, and strict determinism formed a complete picture of reality. For three centuries, this model worked so well that it became synonymous with truth itself. The universe was seen as a perfect machine, predictable in principle down to the smallest detail.

The Collapse of Absolutes

The late nineteenth and early twentieth centuries shattered this certainty. Electricity, magnetism, relativity, and quantum mechanics exposed the limits of Newton’s universe. Absolute space and time vanished. Determinism broke down. Particles lost their solidity. Newtonian physics survived but only as an approximation valid under specific conditions.

A Repeating Pattern in Scientific History

This was not a one-time correction. Classical mechanics replaced Aristotle. Relativity and quantum theory replaced classical mechanics. Today, even these modern pillars conflict with each other. Dark matter, dark energy, and the nature of time remain unresolved. Every “final theory” eventually becomes a special case.

Science as Model, Not Reality

Science does not reveal reality as it truly is. It builds models conceptual stories that explain observations within known limits. When conditions change or new evidence appears, the story is rewritten. Newton’s laws were not lies; they were useful narratives that worked until they didn’t.

Why Science Still Works

Calling science a fiction does not mean it is useless or imaginary. Airplanes fly, medicines heal, satellites navigate. Scientific models work because they are constrained by reality. Reality does not allow just any story it edits and rejects those that fail.

The Irony of “Science Fiction”

Much of what was once called science fiction alike space travel, atomic energy, time dilation became science. Meanwhile, today’s science will one day be labeled incomplete or naïve. The line between science and science fiction is not fixed; it moves with time.

Science as Disciplined Imagination

Science is not absolute truth. It is a disciplined, self-correcting imagination bound by evidence and experiment. Unlike myths, it knows it may be wrong and builds revision into its structure. Its strength lies not in certainty, but in adaptability.

The Best Fiction We Can Write

In the bigger picture, science is a continuously evolving narrative about nature. It is the best fiction humans can write under the strict censorship of reality. And history assures us of one thing: the story will be rewritten again.

2025 in Review: Patterns Beneath the Writing

2025-12-31T15:59:00.000+05:30

This final post of 2025 is not another essay, but a brief reflection on the patterns that emerged across the year’s writing, distilled through a retrospective analysis of my own posts (with the help of GPT).

Some signals were unmistakable.

Across 70+ posts, an ideological arc became visible:

Early 2025: technical foundations (AI mechanics, quantum primitives)
Mid 2025: structural and systemic critique (governance, dependency, alignment)
Late 2025: civilizational and ethical synthesis (youth, sovereignty, cognition, power)

Rather than isolated topics, the year showed high cross-domain coupling AI and Quantum were rarely discussed alone, but consistently framed through society, ethics, geopolitics, and human consequence.

A notable signature emerged through original or rare conceptual frames, including:

Cargo Cult AI, Pixelized Tyranny, Experience Blockers, Circuit Banishment, Informational Obesity, and Stratacordance.

These metaphors reappeared across months, forming a conceptual spine, not one-off phrases—an indicator of long-term idea building rather than reactive commentary.

Even without deep analytics, lightweight engagement signals were clear:

Posts with societal framing clustered naturally
Metaphorical titles consistently outperformed literal, technical ones: This reinforced a simple insight: meaning travels farther than mechanics.

Overall, the bias of the year leaned strongly toward evergreen thinking writing meant to outlive news cycles and remain usable as intellectual infrastructure.

If 2025 taught me one thing, it is this:

The most important work is not explaining technology—but interrogating the systems it quietly builds around us.
The future problem is not smarter machines, but unexamined systems.
The real risk is not that technology moves too fast—but that society stops asking the right questions.

2026 will go deeper.

From GDPR to DPDP: A Quick Comparison Ahead of My Research

2025-12-30T22:20:00.001+05:30

Key Differences Between GDPR and DPDP by Anupam Tiwari

This post is a bit of a departure from my usual IT-focused content. I’m currently working on a paper titled “DPDP-Aware Federated Model Unlearning: An Experimental Study”, and as part of my research, I wanted to get a clear understanding of India’s proposed Digital Personal Data Protection (DPDP) Act. While my main work revolves around federated learning and model unlearning, this post serves as a quick reference comparing DPDP with the European GDPR, helping me and hopefully you grasp the key differences before diving deeper into DPDP-related experiments.

Meet DABUS: The AI That Tried to Become an Inventor

2025-12-30T21:53:00.004+05:30

What Is DABUS? A Simple Explanation

As artificial intelligence becomes more advanced, it’s starting to raise questions that go beyond technology — into law, ethics, and creativity. One of the most famous examples of this is DABUS, which stands for Device for the Autonomous Bootstrapping of Unified Sentience.

What Exactly Is DABUS?

DABUS is an artificial intelligence system created by computer scientist Dr. Stephen Thaler. Unlike typical AI tools that follow very specific instructions, DABUS was designed to generate new ideas on its own.

In simple terms, DABUS works by using interconnected neural networks that interact with each other in a way similar to brainstorming. From these interactions, the system can come up with novel concepts and designs without being told exactly what to invent.

When Did the DABUS Case Happen?

The DABUS story began in 2018, when its creator, Stephen Thaler, filed patent applications in several countries naming the AI system itself as the inventor. This sparked a series of legal decisions between 2019 and 2021, as patent offices and courts around the world considered and mostly rejected — the idea of AI inventorship under existing laws. The case remains influential today as discussions about AI and intellectual property continue to evolve.

Why Did DABUS Become So Famous?

DABUS became well known not because it exists, but because of what happened next.

Dr. Thaler filed patent applications for inventions that DABUS had generated and instead of listing a human as the inventor, he listed DABUS itself as the inventor. This was something patent systems around the world had never really dealt with before.

The inventions included:

A food container with a special geometric shape that improves stacking and heat transfer
A flashing light beacon designed to attract attention in emergencies

The Legal Controversy

These patent applications sparked a global legal debate:
👉 Can an AI be considered an inventor?

Most patent offices around the world said no, explaining that current patent laws require an inventor to be a natural person (a human being). As a result:

The United States, United Kingdom, and European Patent Office rejected the applications
South Africa granted a patent listing DABUS as the inventor, largely because of its formal registration system
Other countries, like Australia, saw mixed court decisions before returning to the human-inventor requirement

Why This Matters

DABUS is important because it highlights a growing problem: AI systems are becoming capable of generating new ideas, but the law hasn’t caught up yet.

This raises big questions:

If an AI creates something new, who should get credit?
Should patent laws be updated to reflect AI-generated inventions?
How do we balance human responsibility with machine creativity?

More Than Just an AI

DABUS is no longer just a piece of software, it has become a symbol of the challenges we face as AI grows more powerful. Whether or not AI systems will ever be legally recognized as inventors, DABUS has already changed the conversation around innovation and intellectual property.

As technology continues to evolve, cases like DABUS help us rethink what creativity, ownership, and invention mean in the age of artificial intelligence.

Can Code Smells Be Measured?

2025-12-30T07:53:00.001+05:30

If you’ve been programming for a while, you’ve probably heard the term “code smell.”

It sounds vague and it is, by design.

A code smell isn’t a bug. The code works.

But something about it feels off: hard to read, risky to change, or painful to maintain.

So the natural question is:

Can code smells be measured, or are they just subjective opinions?

The short answer: yes, partially.

What a Code Smell Really Is

A code smell is a warning sign, not a diagnosis.

Just like a medical symptom:

It doesn’t guarantee a problem
But it strongly suggests one might exist

Examples:

Very long functions
Too much duplicated code
Classes that do “everything”
Deeply nested logic
Functions with too many parameters

Measuring Code Smells (Indirectly)

Code smells can’t be measured directly, but we approximate them using metrics.

1. Size & Complexity Metrics

These are the most common indicators.

Lines of Code (LOC)
- Large methods/classes → Long Method, Large Class
Cyclomatic Complexity
- Counts decision paths (if, loops)
- High values → complex, fragile logic
Nesting Depth
- Deep nesting → harder to reason about
Number of Parameters
- Too many → unclear responsibilities

These don’t prove bad design—but they raise red flags.

2. Duplication Metrics

Percentage of duplicated code
Code clone detection

High duplication often signals:

Poor abstraction
Higher maintenance cost

3. Object-Oriented Design Metrics

Used mainly in Java, C#, etc.

Coupling (CBO) – how dependent classes are on each other
Cohesion (LCOM) – how focused a class’s responsibilities are

High coupling + low cohesion often points to God Classes.

Rule-Based Smell Detection

Static analysis tools use heuristics, such as:

“If a method is longer than X lines AND complexity is above Y → flag it”

Popular tools:

SonarQube
ESLint
Pylint
PMD
Checkstyle

Important:

These tools warn—they don’t judge.

Composite Scores

Some tools calculate an overall number, like the Maintainability Index (MI).

It combines:

Code size
Complexity
Low-level metrics

Useful for:

Tracking trends over time

Not useful for:

Declaring code “good” or “bad”

What Cannot Be Measured Well

Some of the most important smells resist numbers:

Poor naming
Confusing abstractions
Over-engineering
Misplaced responsibilities

These require human judgment and code reviews.

How Teams Use This in Practice

Good teams don’t chase perfect scores.

They:

Track metrics over time
Set reasonable thresholds
Use tools as early warning systems
Rely on developers to make final decisions

The Big Takeaway

Code smells are measurable signals, not absolute truths.

Metrics help you notice problems.
Experience helps you decide whether they matter.

If this topic interests you, explore:

Refactoring patterns
Static analysis tools
Software design principles
Clean Code vs. pragmatic tradeoffs

That’s where real learning begins.

Superdense Coding: Why It Matters in Quantum Communication

2025-12-29T23:28:00.001+05:30

What is Superdense Coding?

Superdense coding is a quantum communication protocol that allows two classical bits of information to be sent by transmitting only one qubit, provided the sender and receiver share entanglement in advance.
This is possible because entanglement lets information be encoded jointly across quantum states, rather than in a single particle alone.

In simple terms:

Shared entanglement + one qubit → twice the classical information capacity.

Purpose: Why Was It Introduced?

Superdense coding was originally proposed to demonstrate how entanglement can enhance communication capacity, not to transmit data faster than light. Its main purpose is to show that quantum resources fundamentally change communication limits, compared to classical systems.

It serves as a foundational example of entanglement-assisted communication, alongside protocols like quantum teleportation.

Possible Applications

While superdense coding is mostly studied theoretically, it has several promising application areas:

Bandwidth-efficient quantum networks: Reducing classical communication overhead when entanglement is available.
Control-plane communication: Sending compact control, signaling, or authentication data in quantum networks.
Hybrid cryptographic systems: Complementing post-quantum cryptography (PQC) and QKD by reducing exposed classical metadata.
Quantum networking research: Serving as a benchmark protocol for testing entanglement distribution and decoding performance.

Key Challenges

Despite its elegance, superdense coding faces practical limitations:

Entanglement distribution: Creating and maintaining high-quality entanglement over distance is expensive and fragile.
Noise and decoherence: Real-world quantum channels significantly reduce decoding accuracy.
Security assumptions: Unlike QKD, superdense coding is not inherently secure and requires additional threat modeling.
Resource cost: Entanglement is a scarce resource and must be generated, verified, and refreshed.

How Do We Measure LLMs? A Simple Guide to Evaluation Metrics

2025-12-27T23:10:00.002+05:30

Understanding Evaluation Metrics for Large Language Models by Anupam Tiwari

As large language models (LLMs) become more capable, evaluating their outputs becomes increasingly important. This presentation provides a concise overview of the most commonly used LLM evaluation metrics ranging from traditional n-gram based measures like BLEU and ROUGE to modern semantic and human-preference-based approaches. It is intended as a quick reference for anyone looking to understand how LLM performance is measured in practice.

Eighteen Years of Curiosity, Code, and Conviction

2025-12-26T14:05:00.001+05:30

On December 26, 2008, this blog began with a simple cybersecurity observation. What followed was never planned — it simply unfolded.

The early years were rooted in the cybersecurity domain: Nmap scans, BackTrack, Kali Linux, learning how systems break so they can be made stronger. Somewhere around 2015, curiosity pulled me into blockchain, a shift that reshaped how I thought about trust, decentralization, and systems at scale. That path led to a PhD in 2022, and since then the journey has only deepened into post-quantum blockchains, AI, post-quantum cryptography, and quantum technologies.

This blog now 800+ posts, 1.38M+ hits, across 18 years is almost entirely technical. It is a record of what I have seen, learned, questioned, failed at, and understood, often thanks to others, always shared back with the same intent.

It hasn’t come without cost. Time that could have been spent more with my wife, my daughter, my home often went into reading, writing, experimenting. But passion is a kind of madness, and I’ve been deep in it. No regrets only awareness.

Looking back, this space is not just a blog. It is a timeline of evolving technologies, changing mindsets, and sustained curiosity. And as the landscape shifts toward quantum and beyond, the journey feels as exciting as it did on day one.

So here’s to December 26, 2008 → December 26, 2025.

Still learning. Still sharing. Still curious.

Today We Model AI... Tomorrow AI Will Model Humans

2025-12-23T21:40:00.004+05:30

1. Right now, AI is in its embryonic stage and limited in scope, but still powerful enough to impress. What we see today is just the beginning. In the near future, AI won’t just be a tool. It will model us, shaping our choices, our behaviors, and even our beliefs often without us even realizing it.

2. As AI becomes more advanced, its presence will be more than just a quiet observer of our actions. It will be actively involved in creating the digital environments in which we live shaping the content we consume, the products we buy, and the experiences we have. We’ll find ourselves not merely interacting with AI but becoming increasingly dependent on it, trusting it to guide us in more personal and significant aspects of our lives.

The Invisible Hand of AI

3. Imagine a world where AI quietly influences your decisions. What you see, what you buy, who you interact with it’ll all be subtly curated by algorithms designed to predict and shape your every move. At first, this might seem like convenience, but as AI evolves, it will become less of a tool and more of a puppet master, guiding us through invisible strings.

4. This influence is already being felt. Consider the personalized recommendations on platforms like YouTube or Netflix. These systems don’t just reflect your preferences they shape them. The more data these algorithms collect, the more they predict what you might like or need next, pushing you toward specific choices, often without you even noticing. In the future, this process will only become more sophisticated, and harder to detect.

The Profit Trap

5. Behind this AI evolution will be one powerful driver: PROFIT. Corporations will compete not just for technological superiority but for control over human behavior. The more AI learns about us, the easier it becomes to manipulate our choices, desires, and even our lives. It will no longer just serve us BUT it will control us.

6. The lines between what’s "personal" and what’s "advertised" will blur. AI will evolve into something capable of predicting not only what you want, but also what you think you want, based on your behaviors, emotions, and digital footprint. This personalized manipulation is dangerous because it targets us at our most vulnerable our need for connection, affirmation, and happiness. Before we even realize it, we may find ourselves trapped in a cycle of consumerism that we didn't consciously choose.

The Hubris of Human Advancement

7. The advancements in science and technology have fueled a dangerous kind of hubris ...the belief that we can conquer, control, and even redefine nature itself. Our egos have inflated disproportionately, and our greed seems to know no bounds. We have reached a point where our technological might far outstrips our moral and ethical understanding of its consequences.

8. This arrogance has led us to a situation where we no longer merely adapt to the world around us. Instead, we have started reshaping it, reprogramming our environment, our bodies, and even our minds. But in doing so, we risk becoming the architects of our own undoing, forgetting that the more we control, the more we lose control over what we once held dear.

A Wake-Up Call

9. If we aren’t careful, AI will shape a world that we no longer control. A world designed not by humans, but by algorithms working in the background, shaping our lives for the benefit of a few. No matter where we live, we might find ourselves cocooned by a web of unfathomable algorithms that manage our lives, reshape our politics and culture, and even re-engineer our bodies and minds. And in the process, we may no longer be able to comprehend the forces that control us, let alone stop them.

It’s time to ask: Who will control AI, and who will it control in the end?

10. If a twenty-first-century totalitarian network succeeds in conquering the world, it may not be led by a human dictator, but by nonhuman intelligence: an AI network with the power to manipulate not just our choices, but our very sense of self. This future isn’t as distant as we might think, and it’s up to us to ensure that we don’t lose ourselves in the process.