aawalton

Erlang has done live updates to code for 30 years. Definitely possible.

Umm, he’s missing the obvious explanation that confirmed cases lag actual cases? Confirmed cases are still going up because testing is going up, regardless of whether infection is going up.

I understood this to mean flattening the curve is important, but not enough. Containment is the most critical intervention, but also the one that requires effective government intervention (South Korea).

Article implies that the average person had a 2% chance of drowning in a given year. If that was true, 64% of people would down by the agree of 50. Basic math doesn't add up. A 2% annual risk of death is catastrophic compared to normal life risks:

Khan Academy (https://www.khanacademy.org/math) is excellent. I recommend working through the exercises as far as you can and then re-learning concepts grade by grade. I go through this process every few years just to stay fluent.

Measuring coding contribution is hard. The most robust attribution models calculate the difference between the success of the project with and without the contribution, but in practice, we almost never know how successful the project would have been without the contribution. The closest we can get are isolated A/B tests, where we can measure the value of a specific code change in terms of business outcomes.

Measuring the volume of changes in any form is problematic, since it’s possible that the programmers who add the most code are actually making the code worse by accumulating technical debt faster than business value (instead of refactoring so that important changes can be made simply).

MUMPS (aka M) was actually my first professional programming language at Epic Systems, the tech giant of the Healthcare space. It’s contemporary with C (which is also still used), and has several features which were unheard of at the time.

1. M bundles a reasonably efficient database with the language itself. Every variable supports jagged hierarchical arrays (which we usually call objects in other languages). These were implemented as balanced binary search trees, which means it’s difficult to write slow operations (at the cost of slightly sub-optimal performance for everything).

2. M was designed for massive server/client operations, handling concurrent use and access by tens of thousands of users out of the box.

3. M was incredibly terse. We laugh about the horror stories now, but when it was developed, every byte mattered, so single letter keywords and white-space-award syntax were features, not bugs.

As a result of these strengths, M solved real problems for large organization, managing data and concurrency at scale, with a simple storage paradigm that prioritizes good enough performance all of the time.