Experimental design for non-compliance, and Perl arcania
We learn a little experimental design for when subjects opt out of the treatment branch we have planned for them. We also learn about an extremely un-useful Perl feature.
Social constructions and high rates of scientific false positives
We take a brief look at how something seemingly-technical like "selling out" an event is a social construction, and we reason from basic principles how much of published science might be false positives.
Memory limits, the north pole is moving, and standard deviations
We learn a systemd trick and some history around the perception of physics. We also think about whether rote memorisation of standard deviations is a useful way to get a sense of variation, and if this in turn leads to better statistical literacy.
Statistical literacy and saving lives for $1,000,000
We hypothesise about something called statistical literacy, which may or may not be a developmental milestone. We also learn what it takes to save a life.
Entropic Thoughts Premium Newsletter #2
We get three links to good articles, we learn about talking empathetically, handling children, what reinsurance is, and get some forecasting tips.
The Basean Loom, and arguing for self-review
We learn how we can combine conditional probabilities into an implied base rate. Separately, we learn that self-review is about as efficient as peer review, although less effective.
War and Peace and days and years
We get a review of War and Peace, and we learn how many days there are in a year.
Chi-squared from fundamentals and the Trial of the Pyx
We figure out how the mysterious Chi-squared test comes about, and learn how we could have illicitly profited from a very important English government position in the 1300s, had we had our modern knowledge then.
Estimate sample size, good software, and lines of code
We learn how to estimate sample size from an intuitive idea of effect size, we learn how to write good software, and we measure software complexity in lines of code.
Nix, statistical analysis of build times, and congestion as demand
We look at how one can use Nix to replace some container usage, we statistically analyse build times to get failure rate out, and we learn a useful perspective on congestion.
Entropic Thoughts Premium Newsletter #1
Transitive cancellation, paper cost of digital, startup survival rates, Amazon's way of working backwards, Palahniuk on writing, data structures in Haskell, forecasting 2025.
Books galore! Savage Money and Audit Cultures
We learn a lot about value systems and how they affect what people do, and we look at a particular kind of bureaucracy that comes from collecting evidence of self-monitoring for compliance with regulations. (Whew, that's complicated!)
Emacs supercharging other programs, Ada is great, and more space software!
We learn how we can use Emacs to supercharge other programs that are suboptimally charged on their own. We learn how Ada excels at low-level programming, and we reflect on one of the things that are really cool about being in software: that we get exposed to the guts of other industries.
Why p values should be distrusted and why one should dig deeper than the immediate problem
We briefly work through why the p value specifies the lower bound for the risk of false positives, look at a common – and silly – source of error beyond that, and then
Supercomputer message-passing, forecasting retrospective, and signs of expertise
We discover how forecasting is going, and learn a little about how to use MPI for supercomputing. We again dip our toes into one of my favourite subjects: research around expertise.
Probability-generating functions, the effect of surprise, and premium newsletter
We'll learn about probability-generating functions, characteristic functions, the element of surprise, and the ability to subscribe as a paying reader of this newsletter!
Watching out for false positives, folding left and right
We do a little basic Bayesian analysis to determine in advance how to react to difficult real-world situations.
Randomising decisions, running Linux on hard mode, and killing astronauts
We learn a little about how to run Linux complicatedly in a VM with Windows as a hypervisor, we use randomness to decide better (or at least faster, which often translates to better), and we look at the risk of space programmes.
Feedback is necessary for expertise
We learn how to recognise fields in which experts can exist, and succinctly describe distributed systems.
The world did not end in 1983, good tagging, and charging for time
On September 26th, we celebrated Petrov day, in memory of Stanislav Petrov who did absolutely nothing at just the right time. In this edition, we also learn how to automate tag quality judgments, and discuss pricing strategies for consulting.
Micromorts! Children get sick! Deep learning made waves!
A very quick look at the effect of deep learning since 2012, some plots on how often children get sick (and whether they acquire meaningful immunity from being pre-exposed to disease), and a note about measuring and comparing risks.
Code reviews are awesome! And boy is Python growing
Correctly drawing bar charts when the buckets into which responses fall are unequal in size leads to fast conclusions that get lost in the noise otherwise. But most of this newsletter is about one of the greatest practices in software development: code reviews.
More on bubble sort, and (separately) software architecture
Where we misunderstand an argument in favour of bubble sort, talk about types of abstraction in a collaborative card game implementation, and read some 1884 science!
Sample unit engineering, AI prompting, and poker
Where we learn some prompt engineering, some sample unit engineering (which is how to count things extra cleverly) and some non-engineering: a little about betting in poker.
The fatal flaw of story points, plus a note on wildland firefighting
We lament the lack of verifiable statements in business, embodied through story points. We also learn how wide a mineral dirt road needs to be to have a reasonable chance of slowing down or stopping a wildfire!
When bubble sort is not terrible, and clever physics experiments
We learn the one case when bubble sort is surprisingly not a mistake, and also about how troublesome it could be back in the early 1900s to figure out fundamental physical properties of matter.
Traffic flow fundamentals, MVP scope and timing
Where we get an opportunity to look at how I write flashcards, and learn about how minimal an MVP really is – and when is the right time to build one!
ACX prediction context halfway review, LLMs, and questions at work
We look at the status of the ACX 2024 prediction contest. We also learn how to invoke LLMs particularly conveniently from the command line to ask stupid questions, and what one should consider before answering questions at work.
Bootstrapping like it's 1908 and the Apollo Guidance Computer
How new techniques are re-hashes of old, about this newsletter, and some fun facts about the Apollo guidance computer.