What's on My Mind

Notes, curiosities, and working margins.

This page is looser than the main site. It is where I can keep track of research notes, methods I am learning, side interests, half-formed ideas, and technical material that does not quite belong in a formal academic profile but still says something real about how I think.

Browse the Notebook

Follow a line of thought.

Choose a theme to gather related notes. Some ideas belong to more than one intellectual thread.

Showing all notes.

Works in Progress

I am currently thinking primarily about the resilience of road infrastructure and, more specifically, about how that resilience can be quantified in a meaningful and usable way. This line of work feels important to me because it asks not only how roads fail, but how they endure, adapt, and recover under changing environmental and service conditions.

Virginia Tech has been a very good place to pursue these questions. I am grateful for the opportunity to work here, and I have especially benefited from the guidance of Professor Deb Mishra, whose clarity, sharpness, and communicative style make him an excellent mentor and a deeply thoughtful person to learn from. In some ways, he reminds me of my previous supervisor, Professor Dharamveer Singh, who was also exceptionally sharp, intelligent, and full of promise.

A Technical Note

Ways of Reading Asphalt Binder

  1. SARA Fractionation (Iatroscan)

    Function: Separates asphalt into four chemical families based on solubility and polarity: Saturates, Aromatics, Resins, and Asphaltenes.

    What it gives us: The colloidal index. It tells us how well the asphaltenes, the hard and brittle parts, are dispersed in the maltenes, the softer and oilier parts. This helps predict whether a binder will remain stable or phase-separate and become brittle.

  2. FTIR (Fourier Transform Infrared Spectroscopy)

    Function: Measures how the binder absorbs infrared light at different wavelengths to identify specific functional groups.

    What it gives us: Oxidation indices, especially carbonyl and sulfoxide peaks. It is the primary tool for measuring how much a binder has aged or rusted at a molecular level because of heat and oxygen.

  3. GPC / SEC (Gel Permeation Chromatography)

    Function: Acts like a molecular sieve that separates molecules by size.

    What it gives us: The molecular weight distribution. It identifies the formation of large molecular size particles, which increase as asphalt ages, and it is also excellent for confirming the presence of polymers such as SBS.

  4. DSC (Differential Scanning Calorimetry)

    Function: Measures the amount of heat absorbed or released by the binder as it is heated or cooled.

    What it gives us: The glass transition temperature \(T_g\) and wax content. It identifies the temperature at which the binder turns from a flexible material into a brittle, glassy state, which is critical for cold-weather performance.

  5. CHNS (Elemental Analysis)

    Function: Completely combusts a tiny sample to determine the mass percentage of carbon, hydrogen, nitrogen, and sulfur.

    What it gives us: The elemental profile. A high sulfur content, for example, often points to specific crude oil sources, including those from the Middle East, while the carbon-to-hydrogen ratio can indicate the degree of aromaticity.

  6. NMR (Nuclear Magnetic Resonance)

    Function: Uses a strong magnetic field to observe the environment of hydrogen or carbon atoms.

    What it gives us: Aromatic and aliphatic carbon content. It provides a much more detailed skeleton map of the molecules than FTIR, allowing researchers to see how the carbon chains are actually structured.

  7. XRF (X-Ray Fluorescence)

    Function: Bombards the sample with X-rays to cause the emission of characteristic secondary X-rays from metals.

    What it gives us: Trace metal detection. It is used to find zinc, copper, or molybdenum, which can act as fingerprints for recycled engine oil bottoms or other prohibited additives.

  8. AFM (Atomic Force Microscopy)

    Function: Scans a physical probe over the surface of the binder to map its topography at the nanoscale.

    What it gives us: Micro-morphology, often in the form of bee structures. It shows the distribution of wax and asphaltenes on the surface, which relates to how the binder will adhere to aggregates.

  9. Fluorescence Microscopy

    Function: Uses high-intensity UV light to make certain materials glow.

    What it gives us: Polymer dispersion quality. Since most polymers used in asphalt glow under UV light while the bitumen remains dark, this test shows whether the polymer is blended evenly or forming large, inefficient clumps.

Current Notes

Machine Learning

Convolutional neural networks as an entry point into modern machine learning.

Machine learning has changed dramatically in recent years, especially with the rise of artificial neural networks. These biologically inspired computational models now outperform many earlier forms of artificial intelligence across a wide range of learning tasks.

Among the most influential neural network architectures is the convolutional neural network, or CNN. CNNs are particularly powerful for image-based pattern recognition, and their relatively clear structure makes them one of the most accessible ways to begin working with modern neural networks.

At this stage, my interest in CNNs is mainly directed toward understanding how the architecture developed, how it is being used in current work, and what kinds of image-recognition problems it handles especially well.

I am reading on CNNs at the moment because I expect to use them in my current work, and I want to build that understanding from the basics upward rather than treating them as a black box.

A useful starting point, and a gentle introduction for readers who want an entry into CNNs, is arXiv:1511.08458.

Why is Perelman awesome?

Grigori Perelman

Portrait of Grigori Perelman

A figure who remains very much on my mind is Grigori Perelman, whose intellectual courage, depth, and refusal of spectacle continue to feel extraordinary. He represents, to me, a rare model of what it means to pursue ideas with absolute seriousness and without compromise.

I am also planning to read Perfect Rigor, which feels like a natural way to stay with his work and the strange clarity surrounding his life.

Book: Perfect Rigor: A Genius and the Mathematical Breakthrough of the Century (link)

Current Curiosity

Rayleigh Distribution and the Stress-Strain Curve

I want to spend time on the Rayleigh distribution and the stress-strain curve, especially as a way of thinking more carefully about variability, response, and the shape of material behavior under loading. What interests me here is not only the mathematics itself, but also the possibility of seeing statistical description and mechanical interpretation in a more connected way.

A Thought

Rheology, Theology, and the Deborah Number

Illustration of Deborah number showing short and long timescale deformation

There is a strange parallel between rheology and theology through the Deborah number. The biblical line often paraphrased as “the mountains flowed before the Lord” takes on an entirely different force once one remembers that, given enough time, even what appears solid may begin to move. In that sense, the Deborah number feels like more than a technical ratio; it becomes a reminder that permanence is often a matter of timescale. Howard A. Barnes and his colleagues discuss this beautifully in An Introduction to Rheology, which is a book I love and keep returning to.

A Mathematical Aside

Erdos Number

Erdos number is quite interesting, partly because it turns collaboration into something almost topological. It is a playful concept, but it also says something serious about how mathematical work travels through people, papers, and shared intellectual lineages.

Formally, if we think of a collaboration graph \(G=(V,E)\), where each author is a vertex and an edge joins two authors who have written a paper together, then the Erdos number of an author \(v\) is simply the graph distance from \(v\) to Paul Erdos:

\[ \operatorname{Erdos}(v)=d_G(v,\mathrm{Erdos}). \]

In that sense, a whimsical academic fact becomes a clean expression of connectedness inside a scholarly network. It also reminds me of graph theory, which fascinates me. In that spirit, Network Science by Albert-László Barabási is a wonderful read.

A Mathematical Aside

Poisson Distribution

The Poisson distribution has a kind of austere elegance. It is one of the cleanest ways to describe how often something happens when events arrive rarely, independently, and with a steady average rate. It feels modest at first, but it is one of those formalisms that quietly appears everywhere.

If \(X\) counts the number of events in a fixed interval and \(\lambda\) is the average rate, then the basic model is

\[ \Pr(X=k)=\frac{e^{-\lambda}\lambda^k}{k!}, \qquad k=0,1,2,\dots \]

The beauty of it is that one number, \(\lambda\), controls both the expected count and the variance. In that sense, the Poisson distribution is a compact way of saying: randomness can still have shape, and even scattered events can obey a disciplined pattern.

A Mathematical Aside

Graph Theory

I am quite genuinely infatuated with graph theory. There is something deeply satisfying about the way it reduces structure, relation, and complexity to a language that is both austere and expansive. A graph can look minimal on paper, \(G=(V,E)\), and yet carry within it connectivity, flow, distance, centrality, cycles, spanning trees, and entire geometries of interaction.

At the most basic level, if \(V\) is a finite vertex set and \(E \subseteq \{\{u,v\}:u,v\in V,\;u\neq v\}\), then a great deal already follows. One may define the degree of a vertex \(v \in V\) by

\[ \deg(v)=\lvert\{u\in V:\{u,v\}\in E\}\rvert, \]

and the graph distance between two vertices \(u,v\in V\) by

\[ \begin{aligned} d_G(u,v)=\min\{\ell:\;&\exists\text{ a path }\\ &u\leadsto v\text{ of length }\ell\}. \end{aligned} \]

From there the whole subject seems to open: connectedness becomes a statement that \(d_G(u,v) < \infty\) for every pair of vertices, a tree becomes a connected graph with \(|E|=|V|-1\), and an adjacency matrix \(A=(a_{ij})\) encodes the graph in a way that lets combinatorics and linear algebra speak to one another through

\[ a_{ij}= \begin{cases} 1, & \text{if } \{v_i,v_j\}\in E,\\ 0, & \text{otherwise.} \end{cases} \]

What I love here is not only the formalism, though the formalism is beautiful, but the fact that graphs keep reappearing everywhere: collaboration networks, infrastructure systems, transport problems, diffusion, reliability, and the movement of ideas themselves. Graph theory feels to me less like a chapter of mathematics and more like a durable way of seeing.

A Physical Reverie

Dirac and the Other Half

Dirac trusted an equation far enough to let it describe matter that no one had yet seen. His union of quantum mechanics and special relativity,

\[ \bigl(i\gamma^\mu\partial_\mu-m\bigr)\psi=0, \]

carried solutions that seemed to point beyond the known world. Rather than discard them as mathematical inconvenience, Dirac followed their logic toward the positron and antimatter.

In that sense, Dirac conceived half of the material universe in his mind before experiment revealed it. It remains one of the loveliest examples of disciplined imagination: an equation insisting that nature was larger, stranger, and more symmetrical than anyone had observed.

A Reading Note

On the Importance of Reading

I keep returning to the importance of reading, not only as an academic obligation but as a way of enlarging inner life, sharpening judgment, and developing seriousness of thought. One of the most memorable reminders of this comes from Werner Herzog, whose relentless emphasis on reading feels both demanding and liberating.

His repeated insistence to read, read, read says something essential: intellectual work is not sustained by information alone, but by long companionship with books, voices, and ideas. He says this in the context of filmmaking, but it applies just as strongly to research, where depth of thought is often inseparable from depth of reading. This compilation of Herzog on reading is something I find genuinely awesome.

An Eclipse, A Verdict

Eddington and the Bent Light

Einstein's general theory of relativity made a precise and audacious prediction: the Sun's gravity should bend passing starlight, shifting the apparent position of a star near the solar limb by

\[ \Delta\theta=\frac{4GM_{\odot}}{c^2b}\approx1.75^{\prime\prime}. \]

During the total solar eclipse of 29 May 1919, Arthur Eddington photographed stars from Príncipe while a companion expedition observed from Sobral, Brazil. With the Sun briefly hidden, the sky became a laboratory: the stars appeared displaced by the curvature through which their light had travelled.

The results announced by Frank Dyson, Eddington, and Andrew Crommelin were consistent with Einstein's prediction and became the celebrated first observational confirmation of general relativity. I find the scene irresistible: an eclipse made the geometry of spacetime visible, and an idea formed in equations left a measurable mark upon the sky. The Royal Astronomical Society preserves an excellent account of the eclipse.

A Scenario Worth Stress-Testing

AI 2040: A Future Delayed on Purpose

The AI Futures Project's AI 2040: Plan A is not presented as its authors' most likely forecast. It is a positive policy scenario: slow the race to superintelligence, make frontier research inspectable, and preserve human control long enough to build a credible safety case.

Read AI 2040
  1. 2029

    Make the deal

    The United States and China agree to a verified slowdown and declare the compute that could sustain frontier development.

  2. 2030

    Brake the loop

    Fully automated AI research would trigger a rapid takeoff; the agreement prevents that loop while useful inference remains available.

  3. 2035

    Hold the line

    Development pauses near top-human-expert capability, creating time for alignment, oversight, and institutional learning.

  4. 2040

    Conditional handoff

    Scaling resumes toward superintelligence only after confidence in safety is judged stronger than the risks of waiting.

Verify the compute

Chip accounting, mutual declarations, audits, inference-only controls, and secured research clusters make the slowdown enforceable without relying on trust alone.

Open the research

Algorithms, experiments, code, and safety cases become open to scrutiny, while dangerous model weights and sensitive data remain protected.

Titrate capability

Scale toward the strongest AI that remains confidently controllable, then slow almost to a halt rather than racing through an intelligence explosion.

The five pressures Plan A confronts

  • Loss of control
  • Concentrated power
  • Great-power conflict
  • Labor displacement
  • Catastrophic misuse

What interests me most is the discipline of scenario scrutiny. Instead of saying that governance will somehow catch up, the proposal has to survive dates, incentives, verification limits, institutional weakness, and the possibility of defection. Whether or not one accepts Plan A, that is a serious way to think about a future arriving faster than our institutions can comfortably reason about it.

Are You Up?

A small numbers game.

There is a hidden number between 1 and 31. You have six attempts. I'll tell you if the guess should go higher or lower.

Six attempts remain.

Back to the top