Today on FAIC: Why did I build a combinator-inspired transformation pass in the first place? Because it's fun! Also, other reasons. ehttps://ericlippert.com/2023/02/23/bean-machine-retrospective-part-8/
Today on FAIC: After that brief digression about combinatory logic, we'll start using some of those ideas to solve compiler problems. ericlippert.com/2023/02/08/b…
Today on FAIC: Finishing up my quick introduction to combinatory logic with the starling; next time we'll see how this inspired some of the design for my Bean Machine compiler. ericlippert.com/2023/02/03/t…
Today on FAIC: a follow-up to my previous post exploring the question "*intuitively*, why can a bird without a fixpoint and a mockingbird not co-exist in a compositional forest?" What's the connection between those three concepts? ericlippert.com/2023/01/30/t…
Today on FAIC: I'm interrupting my series on Bean Machine to talk briefly about combinatory logic; it'll all make sense eventually I'm sure. ericlippert.com/2023/01/17/t…
Today on FAIC: How does Bean Machine treat a call to random variable function as the abstract notion of "a random variable" instead of a distribution object, or a sample from that distribution? By using one simple trick!
ericlippert.com/2022/12/20/b…
Today on FAIC: A brief digression to discuss toasters, network configuration and Alexa -- and to congratulate some former colleagues. ericlippert.com/2022/12/14/i…
Today on FAIC: three years ago I was given the challenge of extracting a Bayesian network from a Bean Machine model written in Python. ericlippert.com/2022/12/07/b…
Today on FAIC: introducing Bean Machine Graph, which trades a "mechanism domain" API and a restricted set of data shapes and operators for improved inference performance over Bean Machine proper.
ericlippert.com/2022/12/05/b…
Today on FAIC: A quick recap of the Bean Machine probabilistic programming language. Next week we'll start looking at the compiler. ericlippert.com/2022/12/02/b…
I was briefly on loan to JMW's team a few years ago; happy to see that this work is getting much deserved recognition. It is an interesting and hard problem to solve at FB scale!
One of the teams I support at Meta is focused on improving the experience of using SQL at Meta. We've just published a blog post about some of our work: engineering.fb.com/2022/11/3…
My team at Meta -- which lowers costs and improves decision making of other teams -- has been disbanded as a cost cutting measure, part of the 11K employees laid off yesterday.
I'm fine, and not looking for work right now. I am however very concerned for my team.
I know it's a tough market out there but if anyone has leads on positions for smart data scientists with some compiler experience who solve hard problems at scale, please do reply here; thanks!
I am installing Windows 11 right now and I immediately noticed a small but important improvement in the installation script. It now says "please keep your computer on". Previous updates said "do not turn off your computer". This is an improvement in two ways.
First, it's simply more pleasant and polite to ask someone to do something instead of making a curt demand. Second, and even better, this now follows Eric's Rule Of Walking On Docks.
When I'm walking on a slippery dock holding my keys I never tell myself (or others) "don't drop your keys". I always tell myself "hold on to your keys". Visualize the desired outcome, not the unwanted outcome!