Three-Monkey Mind

Knowledge acquisition through variation, lime pie edition


This is a companion followup post to Said Achmiz’s Key lime pie and the methods of rationality, henceforth abbreviated as KLPatMoR. In KLPatMoR, Said Achmiz shows people the value of SDI by using it to generate a good (key-)lime pie recipe. His quotation/distillation of SDI is as good as anything as I could generate, so I’ll replicate it here for you:

  1. Collect procedural instructions for the object of your search from many different people, saturating your field of search with a variety of recommendations.
  2. See what the instructions have in common, and also determine whether to include any non-common elements.
  3. Try out the resulting procedure, then experiment with variations on it.

KLPatMoR primarily focuses on the first 2½ parts of SDI. This post focuses on everything after the “then” — that is, “experiment with variations on it.” My purpose here is to show you that:

Background knowledge is important. If you haven’t already, you should read Said Achmiz’s Key lime pie and the methods of rationality. At the very least, you should read the recipe at the bottom. His recipe is thoroughly considered and his pictures should give you a much clearer idea of what I’m trying to accomplish.

Sources of variation in piebaking

In order to become a passable baker, I need to learn how to do things consistently in the kitchen. In order to accurately track how consistently I can bake one particular thing, I realized that I should minimize the number of things that I could change without noticing that I’d done so. This means I should:

I decided to pick one pie recipe and stick with it until I became able to consistently make good pies using that recipe. By sticking to one recipe, I’m better able to track my progress because I’m using a consistent yardstick.

One possible downside? I could get sick of key-lime pie before I get good at making it.

Well, what could I change by accident? Lots of different things, most of which I did change by accident:

Since I’m trying to make it easy to tell if I’m changing things accidentally, I’m sticking with a key-lime pie recipe for two reasons:

My tools could also affect the outcome in ways a novice like me can’t easily predict:

If I want to learn to make pies, I should have as little change as possible between pie-baking rounds. How can I do that?

Minimizing variation


When trying a recipe for the first time, I always attempt to follow it to fidelity. That way, I don’t have to wonder if I made poor substitutions. That said, there were a few choices that I made right from the get-go as to how I’d follow the recipe:

Pie One

I followed the directions. Mostly.

When I got to the “press the crust into the edges of the pie tin” step, I wasn’t sure what I needed to do, or why. I figured out that pressing diagonally into the corners will force the crust up the sides. Here’s why you want the crust to come up the sides tightly:

Here’s where I screwed up:

Misblended filling

I put the egg yolks in a bowl. Then, I put the zest, lime juice, and sweetened condensed milk in the bowl. Finally, I beat the whole thing with a hand mixer.

While the filling tasted fine, I missed an opportunity to beat the yolks and increase their volume. If I’d done that, the filling would have been fluffier and more custard-like.

Later on, I’d beat the filling extensively, but with all the other ingredients already added. This causes other problems.

Egg white in the filling

I asked Said about this and he said that traces of egg white might retard the fluffiness of the filling (this would be noticeable even before baking the pie).

Put too much sugar in the whipped cream

By AquaCalc’s calculations, a half cup of granulated sugar weighs 67 grams. I looked at the remnants of my bag of sugar and thought “there’s no way this is 67 grams. I’ll dump the whole thing in and open up my second bag.”

After the bag was empty, the scale read 140something grams. Oops.

Forgot to put vanilla in the cream

I copy down the ingredients list and instructions of every recipe I try so I can make sure I understand all the steps.

Somehow, I didn’t copy this ingredient over.

Slightly overbaked crust

The crust was well-baked. A little too well-baked; it tasted slightly burnt. I had a lot of trouble figuring out when the crust had gotten darker. The edges of the crust will darken compared to the middle, but it’s hard to tell when the center is darker than it used to be.

Baking took too long

I baked it on the second rack. Middle rack is #3. Top rack, up by the broiler, is #5.

It took about 25 minutes for the crust to seem done.

It took about 25 minutes for the filling to have one density around the outside, but another density in the middle.

Pie One recap

All that said, Pie One was easily the best lime pie I’d ever had. It was tart, but the crust and whipped cream balanced the tartness and the thick, well-baked crust balanced the fluffiness of the whipped cream and the filling. While I was anti-crust before this pie, I changed my mind: a good crust improves a pie. It doesn’t just keep the filling from sticking to the container it’s in. A good crust counterbalances the tartness of the filling and the creaminess of the topping.

The crust did not have a soggy bottom. The bottom of the crust was dampened to about halfway through, but it hadn’t wet completely through anywhere.

Pie One – Pie Two interlude: Reducing the graham-cracker amount

At this point I got ambitious. My normal graham-cracker boxes come with three packs of nine crackers. Each nine-cracker pack has 145 grams of crackers in them. Said’s recipe calls for 11 graham crackers, meaning I had to use all nine…and then two more. My graham-cracker usage would be a lot cleaner if I used a full pack instead of a full pack plus a small fraction of a second pack. What would happen if I scaled the crust recipe down by nine elevenths? Would something go wrong? I was determined to find out.

Pie Two

Pie Two was a comedy of errors. Fortunately, it still tasted good…mostly.

Comically overbaked in too much time

It took a grand total of 34 minutes before I thought the filling was done. I knew it was done because it was so overbaked it developed cracks (curdled). I figured a pie hot enough to develop cracks was past done, and I was right.

What’s more, the jiggle test wouldn’t have saved me. The temperature was so low, the pie would bake evenly throughout instead of baking from the outside in. I could’ve done the jiggle test at any time during the cooking, and the center of the filling would jiggle in the exact same way as the filling close to the edge.

I resolved to fix my stove’s temperature before baking any more pies.

Crust didn’t come up the sides as far

As mentioned in the Pie One – Pie Two interlude, I decided to use 9/11 as much crust as Said’s recipe called for. However, my crust-flattening technique wasn’t as good and I didn’t get the crust firmly up the sides in all places. This left spots where the filling touched the tin during the bake, and the filling got burnt. This burnt filling anchored the pie to the tin, and made removing slices difficult. It didn’t taste good, either, so I had to forego some of the pie edge.

One saving grace of not putting enough crust up the sides: The bottom wasn’t soggy.

Pie Two – Pie Three interlude: Oven calibration with Grands

I asked around how to calibrate my oven because the two thermometers I was using seemed to indicate that my oven’s temperature was fine. I was told that most ovens were perfectly well-calibrated right from the factory, but if I had further doubts, the best thing to do was to buy some excessively engineered popular baking good and use that to calibrate my oven. I went to the store and settled on Pillsbury Grands, a brand of buttermilk biscuit. I thought this was a good choice for several reasons:

I bought four, took them home, and inspected the instructions:


Heat oven to 350°F (or 325°F for dark or nonstick cookie sheet). Place biscuits 1 to 2 inches apart on an uncreased cookie sheet. Bake 13 to 16 minutes or until golden brown.

Grands One

I preheat the oven to 350°, stuck the Grands on my half-sheet pan on the second-from-bottom rack position, and watched them until they could pass for golden brown.

They didn’t look at all done at 13 minutes.

They didn’t look done at 16 minutes, either.

They looked fine at 19 minutes. I took them out then.

Grands Two

I moved the rack up to position #4 (of 5). These baked only incrementally more quickly.

I also noticed that eating the Grands were giving me headaches. I resolved to limit myself to one per batch and throw the rest out.

Grands Three

I adjusted the oven’s calibration to add 15°. The Grands baked in the middle rack with this setting were OKish at the 13-minute mark. They were definitely done after 17 minutes.

Pie Three

This one had the comical screwups.

Omitted sweetened condensed milk

I beat the eggs for quite a while in an attempt to get them to fluff up. I then put in the zest, and beat the mixture just enough to incorporate the zest. Then I added the lime juice and beat the mixture just enough to incorporate the lime juice. As I poured the filling into the pre-baked crust, a series of realizations crossed my mind:

This is thinner than usual…it’s not going to fill the crust like it normally does…I forgot the sweetened condensed milk!

I poured the mixture back into the mixing bowl. Surprisingly, the crust didn’t wash into the mixing bowl with the filling. The crust then sat and got soggy while I mixed the can of sweetened condensed milk into the rest of the pie filling. Back into the crust the filling went, disaster averted.

This screwup wasn’t without consequences, though. This pie was the first one to have the dreaded soggy bottom.

Now, you might wonder what the big deal is. The substance that’s soaking through the pie crust is just a millimeter or two above, right? How bad can that be?

It’s bad. Remember, part of what makes a pie great is the texture variation between different parts of the pie. In a properly-baked pie, a solid crust counterbalances the tartness and wetness of the filling. In a soggy-bottomed pie, however, the crust is sweetened and turned into mush and just isn’t very different from the rest of the pie.

Really, it’s shocking how much worse a soggy-bottomed pie is.

At this point I’m questioning the wisdom of nine-cracker crusts.


I never did see this pie when the middle had a different jiggle from the edges. I suspect that means I overbaked it. The jiggle was uniform at the 10-minute mark and the pie started to crack 14 minutes in.

Pie Three recap

At this point I realized that I’d lost my ability to tell when a pie was properly set. I checked my notes and resolved to check the pie significantly earlier. I’d been looking for a difference in doneness between the middle and the edge, and by the time I was checking the pies, the middles were also fully baked and the edges were on their way to cracking.

Oh well. Now I know what an overbaked pie looks like, and I’m pretty sure I know how to fix it.

Pie Four

I baked the crust for 11 minutes and the pie for eight. I also didn’t get to eat my first slice of the pie until 24 hours had passed. What was it like?

This pie, too, was overbaked. However, it wasn’t so overbaked that it cracked anywhere.

The base wet out in parts and there were blotches of soggy bottom. I didn’t predict this; the edges of the crust seemed perfectly baked. Evidently browned edges aren’t a good indicator of a thoroughly-baked crust.

Pie Four recap

Like the two before it, it seemed to be overbaked. However, it hadn’t gotten to the point where it had cracked. I took this one out at the eight-minute mark. I considered turning the +15° adjustment down to +10°, and eventually I did. Hopefully, for next time, this will give me a bit more lead time to catch the pie before it’s overbaked. On the other hand, Said’s recipe has different timings from mine, and that still bothers me somewhat:


For this bake, my pre-bake was longer than his but my bake was shorter than his. I’m not sure what to make of this. Here’re all the things I can think of that might cause differences between the two:

Also, here are a bunch of things that I think don’t contribute to a longer this but a shorter that:

I’m going to go for the full 11 crackers next time. I don’t want soggy bottoms anymore. While a thin crust isn’t a guarantee of a soggy bottom, it definitely reduces the error tolerance of your crust making ability. This appears to be something I can’t easily develop, and at this point, I’m not sure I want to try to develop it. I like thick graham-cracker crusts too much.

Pie Five

I baked this on a Friday afternoon, topped it Saturday morning, and ate it Saturday night.

I didn’t materially screw this one up. My less-fussy friends gave it rave reviews.

I used the full eleven crackers; this was the correct choice, and I’m going to use the full eleven crackers in the future. The filling soaked through half the height of the crust, but the crust still held firm.

Click/tap on either of the following figures to open the image in a new window.

A whole lime pie in its pie tin. It is topped with whipped cream and lime zest.
Before eating. Eighths for sharing, not fifths for testing.
The remnants of a mostly-eaten pie slice on a plate. The crust is overturned.
The underside of a slice. The crust is nice and stiff.


I wrote this to show that you can learn by trying a technique repeatedly while tweaking the technique in between runs. I also wanted to show that you can learn from tweaks even if you didn’t mean to tweak your technique. While I learned about pie-crust tolerances from my deliberate attempt to use fewer crackers, I also learned about the thin differences between pie fillings and eggy custards from my comically overbaked fillings.

So, don’t be afraid to change things around or screw things up — you might learn something.