Mole End

Things that are are easy to measure. (Well, actually, they're not, but that's another section of this blog entirely. For this moment, let us assume they are.) Things that are not, on the other hand, are quite difficult to measure.

The Mole's late mother had a lovely response to a certain class of hypothetical questions. Where others might reply "How should I know?" or "That's just silly", her rejoinder was "If you had a brother, would he like cheese?". This captures the essence of a counter-factual: what can we say about some contingent fact in a set of circumstances that are not? The resolution requires the introduction of some other premise, some hypothesis about the universe in the situation that has not transpired.

Your organization probably does not have KPIs or OKRs or even SLAs for cheese. But The Mole would lay heavy odds that you cannot throw a stone in your office without hitting someone -- ok, HR would not approve of stone-throwing and thanks to SARS-Cov-19 there's probably no one in the office -- who has at some point claimed to have delivered an initiative that saved the company some resource. And there's your cheese: the savings are relative to the expenditure that would have happened had the initiative not been executed.

Without discipline, anyone can concoct an alternative scenario that fits their ends. The savings were huge! The savings were trivial! The savings doubled our profitability! The savings were actually negative!

There is only one answer, for which The Mole is indebted to the wise Wayne Lim: don't launch the savings initiative until the stakeholders have signed off (ideally in blood) on the counter-factual. Since resource consumption is likely to be dynamic and costs are likely to be complex (e.g., comprised of fixed and variable components), the counter-factual must take the form of a model: it must be sufficiently parametrized to account for the relevant actual state of the world at the future time. With such a model in hand it is much more difficult for savings claims to be attacked Big Lebowski style.

And if you must inquire about The Mole's family: everyone loves cheese, so naturally that hypothetical brother would.

There was some surprise at the beginning of this month that Marissa Mayer was personally reviewing every candidate being hired into Yahoo.  A lot of people, particularly the ones who calculated the ratio between a typical software engineer's annual salary and the price of the 15 minutes she might spend reviewing a resume (see her offer letter if you want to do the math), questioned whether or not this was a good use of her time.  

The Mole predicts that she won't be doing it long.  She doesn't need to.  As long as hiring managers believe that their offers might be reviewed by the CEO they will take care to avoid offers that wouldn't pass muster.  The Mole knows this from experience:  at one time, he and other hiring managers in the department had to prepare offer portfolios for review by a very senior executive.  We eventually noticed that he never overruled us, and came to suspect that he didn't even read them.  But the exercise of preparing the portfolio ensured that we were certain that our offers would stand up to serious scrutiny.  The teams we built then were exceptional.

And Paul Krasner?  The Mole, in the days when he attended performance art avidly, once saw Krasner live on stage.  The most memorable part of the evening was an extended discourse on drinking one's own urine.  While playing Penn-and-Teller-esque mindgames with the audience over the actual contents of a clear plastic cup holding some golden liquid, Krasner built up to his key insight:  one can indeed achieve some amount of enlightenment by drinking one's own urine, but the enlightenment comes in the moment one is capable of deciding to do it, not from some mystical property of the liquid itself.  

And so it is with hiring at Yahoo:  the improvement will come from the changed behavior of hiring managers who will recruit as if they will be called to explain themselves to Marissa Mayer.  Her actual reviewing of the resumes is as irrelevant as Krasner's downing of his prop.

That was how he ended the evening.

Last night The Mole went to an event at the Computer History Museum where Doron Swade and Nathan Myhrvold spoke about the Difference Engine #2 that has just arrived at the museum. The machine itself astonishes, the vindication of Charles Babbage is a monumental achievement, Babbage's insights and vision are amazing. (Myhrvold's expenditures on the effort, though not revealed in detail, must have been jaw-dropping.)

The Mole cannot begin to convey the depths of wonder the Engine triggered. It will be on display at the museum for a year before taking up residence in Myhrvold's living room. Go see it!

So last week the Mole attended the SDForum SAM SIG meeting to hear Ted Neward speak on "State Management: Shape and Storage:
object/relational (O/R) impedance mismatches for persisting
objects between transient & persistent states." This is a topic that has long fascinated the Mole, which, he discovered, is another way of saying that the Mole was not the target audience. Most of the talk was about the problem; the Mole was looking for new insights into the answer.

In the car on the way home, the Mole thought some more about the issues. One of the phrases that had come into his mind during the talk was "locality of reference". It seems to the Mole that the spectrum of distances from the processor core (registers -- cache -- RAM -- disk -- offsite tape) reflects a range of concerns from extremely process-centric to extremely data-centric. An object model is very useful for computing (duh: OO, by definition, couples data with behavior); a relational model is very useful for long-term data storage. One day the Mole will draw a picture and elaborate on this.

The Mole also recommended Fabian Pascal's writings (some of which are specifically relevant) to a couple of people attending the event. The Mole believes Santayana's dictum that "those who do not remember the past are doomed to repeat it."