Intellectually Curious

A deep dive into one of psychology’s classic visual tricks—the Ebbinghaus illusion. We explore how the brain constructs size, what it reveals about perception and action pathways, and how genetics, development, and culture shape our susceptibility. From early debates about vision to cross-cultural findings that emerge in childhood, this episode shows why seeing isn’t just about photons—it's a...

An in-depth look at HealthBench, the open-source benchmark for safe, effective healthcare AI. We explore how 5,000 multi-turn clinical chats are scored by 262 physicians across 60 countries on 48,562 criteria, covering accuracy, communication, context, and instruction following. We also review early results (GPT-3.5 Turbo ~16%, GPT-4 ~32%, O3 ~60%, and the surprising Nano outperforming a larger mo...

In this Science Corner episode, we unpack the method of loci—the memory palace technique. Learn what it is, why it taps into our brain’s spatial and memory systems, and the neuroscience behind its effectiveness. We’ll walk you through practical steps to build your own memory palace, discuss high-level results from memory athletes and students, and share caveats to keep in mind for learners of all...

We explore A000222—the coefficients of the Menage-Hit polynomials that count, for n couples around a circular table with alternating genders, how many arrangements have exactly k adjacent couples. From the zeros at indices 0 and 1 to the rapid factorial growth ~ (2/e^2) n!, we trace what the numbers mean and why they matter in combinatorics. We discuss the prime-divisibility identity found by Van...

In this Deep Dive we follow 5 under the map that replaces a number by the sum of the squares of its decimal digits. The sequence goes 5, 25, 29, 85, 89, and then falls into an eight-term cycle: 89, 145, 42, 24, 16, 37, 58, and back to 89. It never reaches 1, so 5 is not a happy number; it cycles instead. We discuss how this relates to the OEIS entry A000221, its connections to A0003132 and A00079,...

We cut through the hype around AI to explain why solid data modeling is the bedrock of reliable AI today. This episode covers shifting data quality into the operational layer, the difference between AI in analytics versus real‑time ops, grounding LLMs with structured context, retrieval + graphs, knowledge graphs, and the rise of AI agents in a polyglot data world. Note:   This podcast was AI-gener...

In this Science Corner episode, we unpack elaborative encoding—the process of linking new information to what you already know to create durable memories. We explore why it works, the cognitive mechanisms behind it, and practical strategies—imagery, association, memory palaces, peg systems, and PAO. You'll get concrete exercises for remembering names, handling abstract concepts, and expanding...

We unpack what a system is, why the whole is more than the sum of its parts, and how boundaries, feedback, and emergence drive biology, technology, and society. From cybernetics to modern complexity, this episode reveals the core ideas and practical value of systems thinking. Note:   This podcast was AI-generated, and sometimes AI can make mistakes.  Please double-check any critical information. S...

What if the promise of 'zero friction' tech is just friction in disguise? This episode digs into the three worlds of friction—the frictionless digital realm, the creaking physical world that underpins our travel and infrastructure, and the curated spaces where friction is fashionable or outsourced. We trace how effort is conserved but relocated—from do-it-all apps to overwhelmed air traf...

On The Deep Dive, we tackle harmonic regression—the art of decomposing time-series seasonality with sine and cosine terms. We’ll walk through the linear regression setup, how to add trends or adapt to count data via GLMs, and how the method can capture annual, quarterly, or daily cycles. Finally, we discuss its limits—namely the symmetric shape—and look at practical fixes like two-step extensions...

An introduction to A000220, counting asymmetric (identity) trees with n nodes. We unpack what symmetry means for trees, why the first nonzero term appears at n = 7, and how the sequence connects to the broader family of tree counts via generating functions. We explore the role of the rooted-tree generating function A004111, the subtraction of symmetry contributions, and the resulting asymptotic gr...

We unpack self-regulated learning (SRL): metacognition, strategic action, and motivation; how learners plan, monitor, and adapt to reach goals; and why a growth mindset matters. We'll compare key models (Zimmerman’s cyclical phases and Winn & Hadwin’s task-focused phases), explore sources of regulation (active, dynamic, and interest-driven), and share practical strategies like journaling,...

In this episode, we unpack Voyager—the embodied AI that learns on the fly inside Minecraft. We break down the architecture (action agent driven by LLMs, a skill library of executable JavaScript, and the curriculum and critic agents), and explain how feedback loops turn actions into new skills. We also cover how to set it up locally, the mods and tooling required, and what Voyager’s approach signal...

A deep dive into the sequence A000219, the number of planar partitions. We explore McMahon’s theorem linking box-restricted planar partitions to lozenge tilings of a hexagon with a hyperfactorial formula, and the recent dual McMahon results counting tilings outside a shamrock core. We’ll see how S-chord hexagons, half-integer hyperfactorials via the gamma function, and Kuo’s graphical condensation...

We explore the simple rule of summing squares of digits and how iterating it reveals happy numbers in base 10 and beyond. Formalizing with f_{2,b}(n), we trace examples like 13 and 4, examine cycles, and discuss arithmetic dynamics across bases. We'll see why there are infinitely many happy numbers in any base, how density behaves without settling to a single value, and what makes bases like...

We explore the classic A000218, the rule that replaces a number with the sum of the squares of its decimal digits. Starting from 3, the path is 3 → 9 → 81 → 65 → 61 → 37 and then settles into the eight-term cycle 37 → 58 → 89 → 145 → 42 → 20 → 4 → 16 → 37. We’ll connect this to happy and unhappy numbers, Porgis’s 1945 result, and generalizations to sums of other powers and to different bases. Note...

In this Science Corner episode on the Deep Dive, we explore a bold idea: might the universe itself be rotating? We examine the Hubble tension—the mismatch between early- and late-universe measures of the expansion rate—and how a small global rotation, or related ideas like a self-similar dark fluid, could alter cosmic history. We trace concepts from Gödel to the possibility of our universe inside...

In this Deep Dive we explore OEIS A000217, the triangular numbers T_n = n(n+1)/2. We trace their roots from the early geometry of the Pythagoreans to Gauss, and survey the many ways they appear in counting, geometry, and algebra. We’ll highlight core identities like T_n + T_{n+1} = (n+1)^2, plus a suite of combinatorial interpretations (handshakes, domino sets, etc.). We’ll also peek at generating...

We unpack a groundbreaking approach called the Perception Encoder (PE), a single, scalable model trained with global vision-language contrastive learning on images and videos. Learn how PE surprisingly learns task-relevant features for OCR, object detection, depth estimation, and tracking without task-specific pretraining. We break down the training recipe, important ablations (progressive resolut...

We explore how the simple rule “replace a number by the sum of the squares of its decimal digits” behaves when starting from 2. Following the path 2 → 4 → 16 → 37 → 58 → 89 → 145 → 42 → 20 → 4, the sequence falls into a fixed eight-term cycle. We unpack the idea of offset in OEIS entries, how A000216 relates to the digit-sum rule and to A08079 (the cycle starting at 4), and how a compact, PRR-like...

In May 1945, as Europe collapsed, an unlikely coalition formed at Castle Itter in Austria: American GIs, disillusioned German officers, and VIP French prisoners fighting side by side against the Waffen-SS. This episode traces how this improbable defense came together, who the players were, and what the sources—from interviews and archives to modern histories—reveal about loyalty, courage, and the...

An accessible tour of Fermat numbers F_n = 2^{2^n} + 1 (A000215): from Fermat’s prime guesses to Euler’s disproof with F5 = 4294967297 = 641 × 6700417. We discuss why any prime factor must satisfy special congruence conditions, the pairwise coprimality of Fermat numbers, and the big open questions about infinitely many Fermat primes and related arithmetic curiosities. Note:   This podcast was AI-g...

A deep dive into how Lyft forecasts demand and driver supply across hyper-local geographies (geohash-6) every five minutes, with predictions updated each minute. We explore the trade-offs between fast, updating time-series models and deeper neural networks, why latency and engineering costs matter, and how data characteristics like spatial correlation and geography shape model choice. Learn how th...

We unpack the Zeigarnik Effect—the memory bias that makes unfinished tasks linger and urge you to finish them. From Kurt Lewin and Bluma Zeigarnik’s classic experiments to Maria Ovsienkina’s motivational angle, we explore why incomplete goals stay salient, when the effect shows up, and how modern design uses it—from progress bars to onboarding checklists. Practical takeaways cover study breaks, op...

A concise introduction to the Box-Cox data transformation. Learn what it is, why it's useful, and how the lambda parameter shapes the transformation; how maximum likelihood selects the best value toward normality; plus practical tips, checks, and caveats (positive data requirements, outliers, and post-transform diagnostics). Real-world contexts—from manufacturing to finance—illustrate how thi...