My Link Log

A collection of links that are worth sharing

First published: 9/11/19

Last updated: 18/03/24

A while back, I created a bookmark classification utility. Bookmarks have always been an integral part of my web-surfing experience and through this post, I intend to share some of my favorite bookmarks. This list will be updated in the future. If you find any link to be dead, try inserting https://web.archive.org/ ahead of the URL.

• Speech and Language Processing (3rd ed. draft): An excellent book dealing with computational linguistics written by Dan Jurafsky and James Martin.
• Things You Should Never Do, Part I: Joel Spolsky talks about how one should rewrite code in an organization.
• Minichess: Contains a lot of minichess variants. Creating one is a great way to learn to program.
• Seat 61: A train enthusiast sharing his various rail experiences from around the globe.
• How to speak: In Prof. Patrick Winston's own words - The talk helps people do a better job in lectures, theses defenses, and job talks. If you like the talk, I highly recommend you to check out his 6.034 - Introduction to AI course along with his home and favorites page.
• Two-bit history: I love reading about the history of computers, programming languages, and software. This is an excellent blog to read about such stuff. The author discusses topics like Version Control, Semantic Web, Relational Model, RSS feeds, and much more.
• List of Turing Lectures: My favorite is Ken Thompson's Reflections on trusting trust, 83.
• What overlooked class of tools should a self-taught programmer look into: An intuitive HN discussion about CS tools one should explore.
• You (probably) don’t need ReCAPTCHA: If you are running a small enterprise/project, you can refer to this blog for alternatives to Google's ReCaptcha.
• HTTP Security Headers - A Complete Guide: A handy reference!
• On Chomsky and the Two Cultures of Statistical Learning: Written by Peter Norvig, this debate has fascinated me for the past month. To see things from a different perspective, I highly recommend reading The Norvig - Chomsky debate and Norvig vs. Chomsky and the Fight for the Future of AI.
• Talking about Peter Norvig, I thoroughly enjoyed reading the following three essays of his: Teach Yourself Programming in Ten Years, How to Write a Spelling Corrector, and Solving Every Sudoku Puzzle.
• Bug in Mars Pathfinder mission's rover: An experience worth sharing in every Operating Systems class.
• The Chinese Room Argument: A follow-up would be to refer to an HN discussion regarding what does it mean for a machine to "understand"?
• List of Advanced Data Structures: A great resource from Stanford's CS166.
• When not to use a regex: How powerful can simplicity be? As a beginner, this was an eye-opener.
• Two of George Orwell's finest works: 1984 (another link) and Animal Farm, are in Public Domain in Australia and Canada respectively.
• Surprisingly popular: A decision-making concept worth exploring.
• Teach Yourself Computer Science: I particularly like the reference materials recommended in the link.
• What are some useful computer-related technical skills I can learn within a day?: This link contradicts what Peter Norvig said in his essay. Nevertheless, as a beginner, this Quora post helped me dig deeper into many different concepts.
• Case Studies from MIT Sloan: I really enjoyed Sony's Battle for Video Game Supremacy.
• History of Undecidability: A series of videos from Computerphile with Prof. Brailsford explaining various undecidable problems throughout mankind's journey, ending with Turing's Undecidable Problem.
• Tetris: Tetris game being analyzed in depth.
• How Shazam works: The explanation here is in-depth. The article speculates working as described in Shazam's original paper.
• Why explore space? : A beautifully worded letter written by Ernst Stuhlinger explaining the importance of space exploration. His letter tries to justify why space-research is critical to any country's growth.
• The economics of Robinson Crusoe: A great article dealing with financial independence.
• Cambridge in Colour: Fascinating photography-based resources.
• Software Engineering SE, Security SE, and Mathematics SE: All of these StackExchange sites have top questions worth exploring!
• Lecture Notes - Computer Networks: While I strongly prefer the textbook Computer Networks: A Top-Down Approach, the lecture notes compiled here are a handy reference. For a quick introduction to Computer Networks, the essays: How Does the Internet Work? and A broad intro to networking is a good starting point.
• The Website Obesity Crisis and Legends of the Ancient Web: Two talks given by Maciej Ceglowski. The former counters the trend of increasing web-page sizes and bloatware. The latter describes the history of radio and how it went on to be used as a propaganda tool, drawing a parallel between the emergence of the Internet and radio. Dan Luu's Web bloat complements the first talk.
• Designing Windows 95’s User Interface: This article details the methodology behind designing Windows 95's UI/UX and the birth of Windows' start button, taskbar, help window, and printer wizard.
• Modern SAT solvers: fast, neat and underused (part 3 of N) and SAT Solving -- The Basics: Two intuitive blog-posts detailing modern SAT solver's working (DPLL, clause learning, and two watch literals).
• You and Your Research: A thought-provoking talk given by Richard Hamming. It encourages you to be an outlier.
• Epigrams on Programming: Curated by Alan Perlis, an epigram is a remark expressing an idea in a clever and amusing way. My favorite is: Fools ignore complexity. Pragmatists suffer it. Some can avoid it. Geniuses remove it.
• Why I Keep a Research Blog: This blog-post resonates with my motive for maintaining a blog. Technical writing and teaching can help bring clarity to our thoughts and ideas.
• Some Useful Probability Facts for Systems Programming: As the author describes it - a bunch of rough-and-ready probability rules of thumb that are deeply related and have many practical applications when designing systems.
• A brief history of CSS until 2016: I read this while playing Mozart, loved it! It describes the making of CSS, the impact it had on web-browsers, and the evolution of web-fonts.
• The Last Question: A short story by Isaac Asimov. Beautifully crafted, this short story perfectly complements 2001: A Space Odyssey.
• Page Weight Matters: Authored by Chris Zacharias, this blog-post explains how reducing page weight opened a whole new door for Youtube. It complements the views presented by Maciej Ceglowski and Dan Luu.
• How much land does a man need?: Mahatma Gandhi once said - We have sufficient for everybody's needs, not for greed. Gandhi revered Leo Tolstoy and this short story from Tolstoy perfectly captures the essence of the quote above. It is simple yet powerful.
• George Orwell's Reviews and Profiles: A hidden jewel! George Orwell cogently expresses his thoughts on many luminaries of our time like - Dickens, Gandhi, and Tolstoy.
• I, Pencil: This essay by Leonard Read describes the manufacturing process a pencil has to undertake. The first person narrative is captivating.
• One Man's View of Computer Science: Richard Hamming's 1968 Turing Lecture emphasizes the engineering aspect of Computer Science, and discusses the importance of mathematics and ethics in the field. Furthermore, his book - The Art of Doing Science and Engineering - contains fascinating chapters (3-5) on the history of computers. He taught a capstone course regarding the same for graduate students at NPS, Monterey, California.
• 6.005: Software Construction: Taught in spring 2016, this MIT OCW course contains thorough readings on Software Engineering topics such as testing, code review, debugging, and specifications.
• Applying to Ph.D. Programs in Computer Science: A well-detailed guide for undergraduates applying to various graduate programs. In this paper, Prof. Mor Harchol-Balter expresses how most professors review the applications and what factors can make an application stand out. After drafting my Statement of Purpose, the self-review process mentioned by the Online Writing Lab at Purdue University was particularly helpful.
• How I Learned To Stop Worrying And Debug Other People's Code: While seeking answers for a fundamental question - How to explore large open-source codebases? - I came across this thought-provoking blog-post. This post was a major inspiration for the MLH OSS Guide.
• Mining of Massive Datasets: Co-authored by Jure Leskovec, Anand Rajaraman, and Jeff Ullman (Stanford). I regard this as the most concise textbook on Big Data Analysis. Moreover, the authors taught an online course complementing the topics mentioned in this book.
• Letter to the Patent Office From Professor Donald Knuth: A fascinating letter written by Donald Knuth to the Patent Commissioner objecting to the policy of granting software patents.
• How to Read a Paper: I have found the three-pass approach particularly helpful.
• The Missing Semester of Your CS Education: Formal undergraduate education is usually devoid of many helpful topics such as using version control, working with a command-line environment, shell scripting, debugging, and profilers. This course by MIT (student-run classes) makes a commendable effort to fill in these voids.
• The Math behind Neural Networks - Backpropagation: A step-by-step guide deriving backpropagation equations for a 3-layer feed-forward neural network architecture.
• How 0-based indexing was born?: Contains explanation from Martin Richards (creator of BCPL) on the use of 0-based indexing in BCPL, its transition to C, and describes its necessity in an era of IBM computers.
• Math For Programmers: This intriguing blog-post motivates a "surveying" based approach to learning mathematics for programmers. Furthermore, one of the books the author mentions about - Discrete Mathematics and Applications by Kenneth Rosen - is one of the finest textbooks I have read during my CS undergraduate.
• Cormen’s Rules of Usage: Prof. Thomas Cormen (Co-author of CLRS) used to give a graduate-level talk at Dartmouth on academic writing. This paper provides a list of 39 such rules to improve the readability of an academic text.
• What's your favorite elegant/beautiful algorithm?: A Hacker News gem which inspires everyone to explore fascinating topics such as BitTorrent, Anagram matching, magic squares, Stable Marriage problem, watch literals (which I have mentioned above in an eariler point) and many many more.
• If ... then ... else had to be invented!: Takes readers on a fascinating adventure revolving around the formation of "else if" statements, starting with Eniac, crossing Whirlwind, Fortran, COBOL, and ending with Algol 60.
• On holy wars and a plea for peace: In Gulliver's Travels, certain subjects of Lilliput rebelled against the king and broke their eggs at the big end (Big-Endians). While the others did so from the little end (Little-Endians). In this paper, Danny Cohen takes inspiration from Jonathan Swift and introduces big and little endianness to computer science.
• How tail call optimization works: From this blog-post - "Tail call optimization happens when the compiler transforms a call immediately followed by a ret into a single jmp". This blog-post details about tail call optimization using call, ret, and jmp statements.
• Why C language decided to start counting arrays from zero: Mike Hoye describes how Dr. Martin Richards (creator of BCPL) made this decision and how it transcended down to its successor - C. Furthermore, he describes a funny anecdote involving time sharing systems, processor cycles, MIT, the President of IBM, and yatches.
• The Untold Story of Silk Road, Part 1 (Part 2): A breathtaking story involving Ross Ulbricht's silk road and the FBI. My friend, Kishore, had the following fascinating comments - "The man leading two lives, espousing purity of thought, complete freedom, total anonymity, yet his platform demanded the deanonymization of those who worked under him, and his grip gradually became more and more totalitarian as he wished to enact sweet justice on those who opposed him."
• Deterministic Debugging with rr: Discusses record-and-replay debugging (also known as time-travel debugging) using Mozilla's rr tool which records all of the nondeterministic parts such as signals, file system interactions, context switches, etc. Interestingly, Michael Chastain had developed a similar tool (mec-replay) in 1995.
• Gustav Kirchhoff's paper describing current and voltage laws: As mentioned in Donald Knuth's TAOCP (volume 1), Gustav Kirchhoff was the first to use free trees to find a set of fundamental cycles in an electrical network and this paper contains the earliest references of trees. In 1972, W. W. Bartley III published a paper on Lewis Carroll's lost book on logic, which contained the Method of Trees - the earliest use of truth trees to prove or refute a logical formula.
• Bradman and the value of a start: This Reddit post sheds more light into Don Bradman's batting style by comparing his starts to some of the other legendary batsmen.
• A Good Part-of-Speech Tagger in about 200 Lines of Python: In 2013, Matthew Honnibal challenged NLTK's then POS tagger by constructing a leaner and more accurate tagger using an Averaged Perceptron. Intriguingly, in 2015, NLTK changed their default implementation to the one detailed in this blog-post.
• Mikhail Chigorin v James Mortimer (Paris 1900): With an absolutely bonkers opening, this obscure chess game is one of my favorites.
• The secret rules of the Internet: Buni and Chemaly discusses how moderation policies have historically been shaped and regulated by the Internet giants. For further reading, Prof. Paul Barrett's Who Moderates the Social Media Giants? is a compelling read.
• The Fortran 1 compiler: Driven by a need to produce efficient object code, some fascinating optimization techniques it discusses are symbolic registers (for intermediate form of code), operator precedence, dead-code elimination, loop optimizations, and control-flow graph creation followed by a Monte Carlo-based register allocation.
• Seven insights into Queueing Theory: This essay explains how the response time (service time + wait time) and utilization can be used to infer service center's load using queueing theory principles.
• How recursion got into programming: a tale of intrigue, betrayal, and advanced programming-language semantics: It describes how recursion got into Algol 60 and the contributions of Peter Naur, F.L. Bauer, Edsger Dijkstra, and A. van Wijngaarden.
• Contextual Word Representations: A Contextual Introduction: Provides an intuitive understanding of word embeddings.
• Light Years Ahead - The 1969 Apollo Guidance Computer: A computer which weighed 32 kgs and had 55 watts of power consumption, 5600 3-input NOR gates, 4 KB RAM, and 72 KB ROM helped mankind reach the moon. In this video, Robert Wills articulates the software engineering principles used for the Apollo Guidance Computer and narrates the breathtaking 1201 and 1202 program alarms encountered by Neil Armstrong and Buzz Aldrin minutes before the lunar touchdown.
• What the world will be like in a hundred years (1922): An article from 1922 by W.L. George which paints a suprisingly accurate picture of 2022. He discusses topics such as ubiquity of air travel, transition of railroads to rail freight transportation, leap from coal to renewable energy, realism in movies, women empowerment, skyscrappers, attempts at birth control, and the continuity of the war-peace cycle. My favorite quote from the essay is - mankind never tears anything down completely to build up something else; it erects the new while retaining the old.
• How to Generate Random Colors Programmatically: I love aesthetically pleasing color schemes. In this blog-post, Martin Leitner-Ankerl describes how we can map colors from RGB space to HSV space and use the golden ratio to create a diverse color scheme across the whole color spectrum. I've created a Python library (okcolor) following his technique..
• A Face Is Exposed for AOL Searcher No. 4417749: An interesting story I came across in my Privacy class. In 2006, AOL had released around 20M search queries from 650k users with the goal of allowing researchers to understand search patterns. For privacy reasons, they had anonymized the records by removing user identifiers. However, NYT was successfully able to de-anonymize user with ID 4417749 using their search queries. The end result was a class action lawsuit being filed against AOL.
• The Designer's Notebook: How to Write Sports Commentary: I've always been fascinated by the idea of automatically generating realistic sports commentaries. In this post, Ernest Adams describes his experience doing the same at Electronic Arts for the Madden line of American football games. Its interesting to hear the nuances that come into play - like when to use action-by-action commentary and when to introduce color commentary.
• Richard Hipp speaks out on SQLite: The creator of SQLite discusses the complexity of SQLite's codebase, how it is more than a key-value store, how SQLite got inspired by FAA's DO-178B testing guidelines, and how fuzzy testing has improved their code quality. My favorite part is when he describes how testing the machine level such that every branch instruction has been taken at least once has helped his team iterate faster.
• The Exceptional Beauty of Doom 3's Source Code : This link discusses the quality of doom 3's source code and how they achieved it. If you scroll down to the bottom, there is a comment from John Carmack. The fun parts are - “Unified Parsing and Lexical Analysis” and “Minimal Templates”. In the latter, the author discusses how the creators re-wrote all necessary STL functions!
• Game Engine Black Book Wolfenstein 3D: As I understand, Wolf3D was one of the earliest examples of ray casting in games. This book by Fabien Sanglard dives into the source code of Wolf3D and explains how ray casting was used (start at section 4.7).
• Allegory, Realism, and Vermeer's use of the camera obscura: I came across this paper in my Computer Graphics class. Philip Steadman describes how Johannes Vermeer used a camera obscura to bring about his magnificent attention to detail. If this was indeed the case, it makes one wonder just how difficult it must have been for Vermeer to trace and paint directly onto the canvases without making mistakes. My favorite part in the paper is when Steadman describes the glass orb in “The Allegory of Faith” which allegedly provides a small glimpse of the camera obscura.
• Four decades of computer graphics : Another paper I came across in my Computer Graphics class. It describes the developments in Computer Graphics from the 60s to 90s and foreshadows key future developments like VR. Its fascinating that the Whirlwind project in 60s had a quarter of an acre of electronics and used light pens to display information. The 80s certainly seemed like a fun decade for CG - CAD/CAM being used in engineering design, better software for businesses and graphics art, the emergence of devices such as the mouse, sensor gloves, and color printers, and improved standards in the graphics community.
• Special Effects for Star Trek II (The Genesis Demo): Yet another paper I came across in my Computer Graphics class. Written by Alvy Ray Smith, this paper describes how the Wrath of Khan became the first feature film to contain a sequence created entirely with computer graphics, and how an incredible team at LucasFilms carried forward this task. It is fascinating to see how they were pushing the boundaries in 1982 - aiming to have no antialiasing, using fractal techniques to reduce the need to manually enter the data, creating innovative camera movements using n-dimensional cubic splines, and designing a texture mapper for spheres. Fun fact, towards the end of the paper, Ray Smith mentions the word Pixar, which he describes as a machine being created by LucasFilms to increase the speed of computing of the frames by 2x to 3x magnitude!
• Lightmap: A neat data-structure which is used to cache surface lighting information that can be multiplied with a diffuse texture map when needed.
• Manuel Blum's advice to a beginning graduate student : The closest to my heart is his advice on the power of writing. He mentions how the only difference between a FA (finite automaton) and a TM (Turing machine) is that the TM, unlike the FA, has paper and pencil. So without writing, we are all reduced to a finite automaton!
• How training puzzles are generated: James Clarke discusses how Lichess.org heuristically automates the process of generating their mate-in-N and material advantage puzzles.
• Why Philosophers Should Care About Computational Complexity: Parts of this paper by Scott Aaronson (like section 4.3 - can humans solve NP-complete problems efficiently? ) makes you wonder what problems in real-life are philosophically NP.
• Graveyard hashing: Tombstones have been used in linear-probing hash-tables as a marker to suggest that an item was on an index but has been deleted since. However, it is known that primary clustering can occur with such a strategy. In their 2021 paper, Michael Bender, Bradley Kuszmaul, and William Kuszmaul discusses a variant of linear-probing called graveyard hashing that eliminates primary clustering (under certain conditions). This in-turn improves data locality. Their strategy involves "artificially increasing the number of tombstones placed in an array until they occupy about half the free spots. These tombstones then reserve spaces that can be used for future insertions."
• Jason Rahman's blog: Delayed Branch: During my internship at Facebook, Jason's code reviews were incredibly informative - they were mostly about optimizations in C++ which can improve performance. Software Performance: Generic Tricks discusses how avoiding copies, reformulating common operations like idiv, restricting the use of shared pointers, and explicitly providing a branch's likelihood can improve performance. Software Performance: System Optimizations goes into some specifics on what FB's storage team is doing for optimizations - avoiding system calls in hot data path and minimizing system calls behind abstraction layers (the example of doing async logging using MPSC queue is interesting!), thread pinning, using polling instead of interrupts (io_uring) or routing interrupts to a small set of non-data path cores, and using huge pages to lower TLB misses.
• The APL Programming Language Source Code: It amazes me how I never heard about APL till now. The idea of functions executing everything on the right seems natural. Each function in APL is either a monadic (one argument) or a dyadic (two arguments). They don't have any formal loops and rely on the iota function (⍳) for looping. The link contains a mind-bending APL program to compute prime numbers upto R - (~T∊T∘.×T)/T←1↓⍳R
• When Splines Were Physical Objects : Describes how splines were drawn before CAD came into the picture.
• Cache-oblivious Mesh Layouts: Yet another paper I came across in my Computer Graphics class. It is about designing algorithms to construct cache-oblivious layouts of polygonal meshes. Cache-oblivious layouts are different from cache-aware layouts as they do not consider the cache parameters. It was interesting to see that the number of cache misses at runtime can be computed using the edge span distribution (based on the runtime access pattern) and the cache miss ratio function, and that a local permutation can help find an optimal edge span distribution. I also enjoyed the part of the paper where they used Multilevel minimization to approximate an NP-hard problem of finding all the local permutations. If the paper appears dense, I recommend checking out the slides from their SIGGRAPH talk.
• Borůvka's algorithm : A surprisingly simple and beautiful minimum spanning tree algorithm, which should be taught in every algorithms class - alongside Prim's and Krushkal's algorithms. It was designed by Otakar Borůvka in 1926 to construct an electrical network connecting several cities using the least amount of wire. This algorithm is described in great detail by Jeff Erickson.
• Twenty questions: A 19th century rule-book that describes the format of a game called twenty questions. It introduces the concept of umpires (who resolve any dispute) and captains (an official spokesperson). Interestingly, their version did not have a constrain on the type of guess word. However, every Sunday it was mandatory for the participants to pick an object, person, or thing mentioned in the Bible.
• Persistent Homology: An Introduction and a New Text Representation for Natural Language Processing: This short paper from Xiaojin Zhu is a great way to explore Group Theory, Simplicial Homology, Persistent Homology, and apply their basic principles to solve a fun NLP problem - classifying an author's age (into child, adolescent, or adult) based on their writing style.
• Reconstructing Turing's "Paper Machine": This article from ChessBase describes a chess algorithm written by Alan Turing and David Champernowne in 1948. To the best of my knowledge, it is the oldest existing chess algorithm. It introduces the notion of quiescence search ("Captures had to be followed up at least to the point where no further capture was immediately possible. Check and forcing moves had to be followed further.") and game trees - for chess. Remarkably, their relative values for chess pieces were quite similar to the ones later discussed by I. A. Horowitz (in 1951) and Bobby Fisher (in 1972). Their evaluation function also took king safety, piece mobility, piece safety, castling, pawn advancement, and mates/checks into consideration. The 2-ply variant seems to have a Lichess rating of 1600 (for blitz and rapid games).
• The topologist's world map: A unique way to describe a world map that is topologically equivalent to a normal map.
• Regulating Data: Regulating tech monopolies has always been a contentious issue. The vast amount of personal data that these companies possess about their users has raised concerns about privacy and security, leading to discussions around regulation. The traditional anti-trust remedies involve breaking up their businesses, which is certainly less desirable from a consumer experience perspective. This blog post from Raghuram Rajan suggests that users should instead own the data they generate, and that this data should be stored by a decentralized entity that is independent of the firms that use the data. Aggregated user data can then be partitioned into different buckets (such as advertising and academic research), which can, in turn, be exchanged with different businesses in exchange for their services. While this idea is certainly not new, I like the concise nature of his blog post and how he uses the Unified Payments Interface (UPI) as an example to strengthen his argument.
• Guide to making a CHIP-8 emulator: Walk before you can run is an old idiom that applies to many things in life. Suprisingly enough, it also applies to coding emulators. CHIP-8 is one of the simplest emulators to program, with just a handful of instructions. And it is often used as a stepping stone to building more complex emulators, such as NES. This blog post from Tobias Langhoff, along with a technical reference guide from Thomas Greene, is a great way to get started with CHIP-8.
• Contextual Word Representations: A Contextual Introduction: This article by Noah Smith provides a nice introduction to word embeddings.
• Encounter with the Woz: In this fun story, the author describes a meeting with Steve Wozniak back in 2006 at a Barnes & Noble. It turns out that Wozniak once pulled a prank during the 70s by creating a fake advertisement for Zaltair, a fictitious computer that was supposedly the successor to the MITS Altair. This new computer was built around the Zilog Z80 processor instead of the Intel 8080. Wozniak used some of the worst ads he could find for wording, and then he distributed the brochures at the West Coast Computer Faire. The author recreated the brochure and was able to get Wozniak to sign it. What's more, Wozniak was carrying laser-cut stainless steel business cards, which he handed out at the event.
• A.I.: Apparently, one of the earliest examples of reinforcement learning came from a project by Marvin Minsky called SNARC (Stochastic Neural Analog Reinforcement Computer). Although Minsky's 1952 paper on SNARC appears to be lost, he explained its architecture in a 1981 interview with Jeremy Bernstein, a theoretical physicist and a staff writer for The New Yorker.
  Relevant part from the interview:

  As an undergraduate, Minsky had begun to imagine building an electronic machine that could learn. He had become fascinated by a paper that had been written, in 1943, by Warren S. McCulloch, a neurophysiologist, and Walter Pitts, a mathematical prodigy. In this paper, McCulloch and Pitts created an abstract model of the brain cells—the neurons—and showed how they might be connected to carry out mental processes such as learning.
  "The one I then envisioned would have needed a lot of memory circuits. There would be electronic neurons connected by synapses that would determine when the neurons fired. The synapses would have various probabilities for conducting. But to reinforce 'success' one would have to have a way of changing these probabilities. There would have to be loops and cycles in the circuits so that the machine could remember traces of its past and adjust its behavior. I thought that if I could ever build such a machine I might get it to learn to run mazes through its electronics— like rats or something. I didn't think that it would be very intelligent. I thought it would work pretty well with about forty neurons. Edmonds and I worked out some circuits so that —in principle, at least—we could realize each of these neurons with just six vacuum tubes and a motor."
  Minsky's machine was certainly one of the first electronic learning machines, and perhaps the very first one. In addition to its neurons and synapses and its internal memory loops, many of the networks were wired at random, so that it was impossible to predict what it would do. A "rat" would be created at some point in the network and would then set out to learn a path to some specified end point. First, it would proceed randomly, and then correct choices would be reinforced by making it easier for the machine to make this choice again—to increase the probability of its doing so. There was an arrangement of lights that allowed observers to follow the progress of the rat—or rats. "It turned out that because of an electronic accident in our design we could put two or three rats in the same maze and follow them all," Minsky told me. "The rats actually interacted with one another. If one of them found a good path, the others would tend to follow it. We sort of quit science for a while to watch the machine. We were amazed that it could have several activities going on at once in its little nervous system. Because of the random wiring, it had a sort of fail-safe characteristic. If one of the neurons wasn't working, it wouldn't make much of a difference —and, with nearly three hundred tubes and the thousands of connections we had soldered, there would usually be something wrong somewhere. In those days, even a radio set with twenty tubes tended to fail a lot. I don't think we ever debugged our machine completely, but that didn't matter. By having this crazy random design, it was almost sure to work, no matter how you built it."

• This is how the cover art of Defender of the Crown (1986) was made: Mysteries are not just confined to the world of law and murder; they also transcend to the world of history as well. In this blog post, Jimmy Wilhelmsson goes on a detective mission to identify the artist behind the cover art of Defender of the Crown, a 1986 game from Cinemaware. His journey leads him - and us - to an intriguing conclusion. Another interesting internet mystery that's worth exploring is: Who made these circles in the Sahara?
• The Alien Voice: Alvin Lucier's North American Time Capsule 1967: In 1961, the earliest demonstration of computer speech synthesis occurred at Bell Labs using an IBM 704. The computer performed a rendition of the song Daisy Bell (Bicycle Built for Two). Interestingly, Arthur C. Clarke, who happened to be present at the Bell Labs demonstration, later referenced it in 2001: A Space Odyssey, where the HAL 9000 computer sings the same song to Dave Bowman just before being deactivated. This essay by Christoph Cox further discusses this by describing the Vocoder technology, what inspired Homer Dudley to build it in 1928, and how years later, Alan Turing and his colleagues modified it to create a secure speech system called SIGSALY for World War II. It also discusses the emergence of computer-generated music.
• Brr: This blog is written by an anonymous individual who was deployed to Antarctica in 2022-23 to work on IT projects. Throughout a series of blog posts, they describe their experiences of living in Antarctica. They cover topics such as laundry, voting while stationed in Antarctica, sending postal mail, and how ATMs work. Additionally, the blogger highlights the Beer Can, a unique infrastructure at Amundsen-Scott South Pole station that connects their living zone to their critical infrastructure zone, located fifty feet below the surface.
• The Cab Ride I'll Never Forget: I don't have much to say about this blog post besides the fact that it's an absolute must-read. It shows how seemingly small acts of kindness can lead to life-changing experiences.
• The Byte magazine archive: This archive, spanning from 1975 to 1994, contains a plethora of hidden gems. For instance, their February 1992 issue features a section on Archie, one of the earliest internet search engines.
  Relevant part from the February 1992 issue:

  For many people, particularly programmers and engineers, the Internet means "info- booty": shareware and freeware source code, documents, graphics, and data sets available by file transfer downloads and from E-mail servers. Sites like UUNET and The World each have several gigabytes' worth of publicly available archives. These are but two of the hundreds of sites with archives accessible via these methods. Even admitting a fair amount of redundancy among archives, it still adds up to about 100 gigabytes, and new sites and offerings are coming on-line every day.
  With so many different archives, it can be hard to figure out where (and at what network address) to access the items you want. If you don't know what you want beyond compilers or CP/M applications, it's even more overwhelming.
  The "archie group" at McGill University (Montreal, Quebec, Canada) has one solution to the problem: archie (archive without the v), the Internet Archive Server Listing Service (for access, see reference 2). Archie is a central database of information about Internet-accessible archive sites, plus server programs that provide access by telnet, anonymous file transfer protocol (FTP), E-mail, and the Prospero distributed computer system.

  Mondo 2000 and High Frontiers are other technology-focused retro magazines worth exploring.
• Alphabet Chess: While its merits are debatable, this is certainly a one-of-a-kind way to play chess.
• Deletion from hash tables without tombstones: Two of the most common collision resolution strategies are chaining and open addressing. In open addressing, linear probing is widely used. Furthermore, when deleting an element in a linear probing scheme, tombstones are often used to mark the deleted element. However, this strategy wastes memory and increases the frequency of expensive rehashing. This blog post by attractivechaos proposes a clever deletion strategy without tombstones. When deleting an element, the strategy moves the next element to the deleted location if it was pushed away by the deleted element due to hash collision. This process recurses until an empty location is reached. Isn't that marvelous? attractivechaos has shown that this strategy can work surprisingly well in practice. I am curious to compare it with graveyard hashing.
• 3D printing a full-size kayak: In this blog post, Nathan Rooy discusses an interesting project he worked on during COVID-19 lockdown. One day, I would love to 3D print an entire chess set.
• Droste effect: A visual recursion! M. C. Escher's Prentententoonstelling (loosely translated to Print Gallery or Print Exhibition) is a breathtaking work of art that showcases this fascinating effect. The Mathematical Structure of Escher's Print Gallery discusses more about how Escher created this masterpiece.
• Theorem of the Day: Robin Whitty discusses a variety of well-known and lesser-known theorems as succinctly as possible.
• AI's Instagram Problem: Andrew Ng shares great words of wisdom on taking inspiration from emerging technologies instead of doubting the value of our own work. I hope to remember and apply this as I advance in my career.
  My favorite part is:

  Someone else's sculpted physique does not take away from your beauty. And the emergence of a hot new technology doesn't mean your current project isn't also valuable, assuming it's technically sound, has a reasonable expectation of impact, and isn't made obsolete by newer technology (which doesn't happen very often). Projects of all shapes and sizes can be wonderful, and what's buzzy today is only one of the many things that will prove valuable in the future.

• This is still a motherfucking website: Shows how websites can be made aesthetically pleasing with minimal use of CSS.
• Car Problems: Shows how even insane-looking problems can sometimes be real.
• Advice to a Beginning Graduate Student: For many years now, this talk from Manuel Blum has held a special place in my mind and has greatly influenced my reading and writing habits. This is thanks to a beyond fascinating philosophically mathematical argument he makes:
  Is writing elevating our powers to that of a Turing machine?

  You are all computer scientists. You know what FINITE AUTOMATA can do. You know what TURING MACHINES can do. For example, Finite Automata can add but not multiply. Turing Machines can compute any computable function. Turing machines are incredibly more powerful than Finite Automata. Yet the only difference between a FA and a TM is that the TM, unlike the FA, has paper and pencil. Think about it. It tells you something about the power of writing. Without writing, you are reduced to a finite automaton. With writing you have the extraordinary power of a Turing machine.

  Additionally, there are some other gems in there as well.
  One involving Bertram Kostant's literacy skills.

  I remember a professor of Mathematics at MIT, name of BERTRAM KOSTANT, who would keep his door open whenever he was in his office, and he would always be at his desk writing. Writing. Always writing. Was he writing up his research? Maybe. Writing up his ideas? Maybe. I personally think he was reading, and writing what he was reading. At least for me, writing what I read is one of the most enjoyable and profitable ways to learn hard material.

  And the power of combining random access with books.

  Books are not scrolls. Scrolls must be read like the Torah from one end to the other. Books are random access -- a great innovation over scrolls. Make use of this innovation! Do NOT feel obliged to read a book from beginning to end. Permit yourself to open a book and start reading from anywhere. In the case of mathematics or physics or anything especially hard, try to find something anything that you can understand. Read what you can. Write in the margins. (You know how useful that can be.) Next time you come back to that book, you'll be able to read more. You can gradually learn extraordinarily hard things this way.

• Create random images in Python: A beautiful algorithm by Renato Budinich to create random images using a two-step procedure. The first step involves taking a random gray-valued mask image and applying the path-finding procedure of the Easy Path Wavelet Transform (EPWT) to it. This path-finding technique is greedy and aims to obtain a succession of vectorized pixel values with minimal variation. It does this by always picking the next point (among the available neighbors) that yields the minimum absolute value difference in pixel values. This approach helps it select regions with similar pixel values before making a jump to pixels with significantly lighter or darker regions. Once a random path of length N is obtained, a random color map is applied, with each point k in the path mapped to the k/N point in the color map. Et voilà! A gorgeous image emerges.
• Null Island Is One of the Most Visited Places on Earth. Too Bad It Doesn’t Exist: Null Island is a fantasy island located at the point where the equator meets the prime meridian. This post explains how it became a dumping ground for geocoding errors.
• Unix and Beyond: An Interview with Ken Thompson: A 1999 interview with Ken Thompson, where he discusses the design of the Plan 9 and Inferno operating systems, the history of Unix, and his hobby of collecting jukebox music. Years later, at SCaLE 20x, Thompson gave a talk on his jukebox. Having worked on radio generation software myself, I can strongly resonate with his hobby. Thompson also intersperses some great wisdom in his interview, such as the importance of orthogonal thinking - it's always good to take an orthogonal view of something. It develops ideas.
• Bitboards and Connect Four: Prof. Dominikus Herzberg explains how a Connect Four game can be represented and operated using a bitboard.
• Martin Gardner's 1969 column, "Mathematical Games": This was the April edition, in which Gardner had laid out eight problems based on logic and probability. A central theme amongst many of them was that the problem involved some sort of game - such as chess, name cards, and word games. To call these problems intriguing would be an understatement - there's an uncrossed knight's tour problem, a variant of chess where the first person to check the opponent's king wins, a variant of the urn problem, and, a word game. I have provided an analysis on this word game in one of my blogs.
• Play Tic-Tac-Toe with Knuth: This little blog post from Russ Cox has amazed me for a long time. This year, I finally got the opportunity to go over TAOCP Volume 4A, hoping to understand Knuth's algorithm. The algorithm discusses how optimal tic-tac-toe moves can be represented using Boolean logic.
  He starts by picturing two 3x3 grids - one for the X’s and another for the O’s. This setup introduces 18 boolean variables - x1 to x9 for the X's, and o1 to o9 for the O's.
  Now, creating a full-scale truth table for these 18 variables means we would be looking at 2^18, or 262,144 rows. However, it turns out only 4520 of those are legal inputs. So, the question becomes, how do we find patterns in this boolean chain to build our tic-tac-toe strategies.
  Knuth simplifies it by considering only the following strategies - (a.) winning - when putting an X in a cell wins the game - like when two cells in any line already have X's. (b.) blocking - when putting an X in a cell stops the other player from winning - like when two cells in any line have O's. (c.) forking - when putting an X in a cell opens up multiple winning moves - you store all the winning line possibilities (horizontal, vertical, and diagonal lines - all 9) in a set, and for each winning line (say {i, j, k}), you place an X on i, leave k empty, and see if a move on j creates two win chances. (d.) defending - similar, when putting an X in a cell stops the other player from creating multiple win chances. (e.) And just making any legal move - when a cell does not have an X or an O.
  The priority order of moves follows the order above. There's also a bit of priority within the legal moves - like the middle spot is best since it's part of more winning lines, while the edges are the last choice.
• Knightmare: A DevOps Cautionary Tale: A bone-chilling tale of a financial services firm that faced a $460 million loss in 45 minutes due to a software bug.
• Ghost Leg: I believe I am done using rock, paper, scissors after finding out about this elegant algorithm. Ghost leg essentially boils down to a sequence of swaps on a set of indices, where each horizontal line (leg) represents a swap between two adjacent indices at a time t. The final order of these indices after all swaps have been applied gives us a permutation of the original set.
  Furthermore, the Wikipedia page describes what a "prime" ghost leg is - as there are an infinite number of ghost legs representing a particular permutation, those ghost legs have a kind of equivalence. Among those equivalent ghost legs, the ones which have smallest number of legs are called "prime". Fascinatingly, there is a relationship between ghost legs and bubble sort. Bubble sort is a sorting algorithm that sorts an array by repeatedly swapping adjacent elements that are in the wrong order. This effectively "bubbles up" the largest unsorted element to its correct position in each iteration through the array. If you think of a ghost leg's end configuration as a shuffled sequence, then the process of adding legs to revert it back to the initial sorted state with the minimum number of moves would be analogous to bubble sort. In other words, each leg would represent a swap that bubble sort would perform. There is also an algorithm to parallelize a ghost leg; however, I leave that up to the curious reader to explore. Hint: red-black balls.
• John Graham-Cumming's keynote on "The Great Railway Caper": John Graham-Cumming narrates the fascinating story of the 1951 business computer LEO I, one of the world's first, which was handed a big data problem by the British Rail: to calculate the shortest distance between each of their 5,500 railway stations. What follows is an engineering feat that spanned over 9 months - the discovery of Dijkstra's algorithm years before Dijkstra published it, optimizations to distribute the data processing tasks, and the fascinating use of punch cards. Tim Greening-Jackson has also documented this story in his paper on LEO I and the BR job.
• John Calhoun's blog: My favorites include his experience interviewing at Apple in 1995, after which he worked there for over two decades, and his SystemSix project.
• Why Do Movies Move?: In this post, Alvy Ray Smith discusses how we perceive motion in films despite them consisting only of individual, still frames shown in quick succession. He explores motion blur and how computationally generated animations recreate it.
• The TX-2 Computer and Sketchpad: In this article, Ivan Sutherland recollects his experience working with the TX-2 computer and developing Sketchpad for it.
• A Parabel: A 1973 Parabel narrated by Dijkstra that involves a train toilet problem. Dijkstra's EWD manuscript archive is fascinating; I encourage exploring his trip reports, such as this one. Furthermore, his three golden rules for successful scientific research are very well put.
• Publishing my first game using Pico-8: A fun blog from Ben Boyter. It brings new life to the conventional rock, paper, scissors game. One day, I'll make my own Pico-8 game.
• The Mystery of the Bloomfield Bridge: There is a pedestrian bridge that crosses I-494 just west of the Minneapolis Airport, connecting Bloomington to Richfield. Hundreds of thousands of cars drive under it, but why is it there, especially in an area that is neither particularly walkable nor connected to any establishments? In this blog post, the author investigates this seemingly irrelevant yet surprisingly fun internet mystery.

←