Tessendorf Plate

I’ve been deliberately-quiet about what I do day-to-day professionally so far. I’m not sure why; I think most of it is a desire to maintain humility and to head off dismissal of the effort that I like putting into this project (“Oh, well, you’re a visual effects artist, so you probably just put these photos in the computer and computer them up, and how hard can that be?”). When I started this project, I was working as a computer graphics artist at Weta Digital in New Zealand. Prior to that, I worked for a while at the now-defunct ESC Entertainment on two of the Matrix films (as well as the ever-popular Catwoman), the now-defunct Big Idea Productions (makers of the “Veggie Tales” children’s animated series…I sang in one of the Silly Songs), and taught for a while at Notre Dame. Currently I’m at Pixar.

“What I do” tends to vary per-project, but my interests are spread amongst shading, lighting, rendering, and “effects” work. Effects (or FX, as it’s oft-abbreviated) usually entails things that are difficult to hand-animate or model: natural phenomena like explosions, rain, dust, or funky things like ‘rips in the fabric of space-time’ (as I got to work on for Jumper, an otherwise pretty-terrible movie). One of the most captivating FX tasks for me to play with, however, is water.

Water is an interesting phenomenon to try to represent mathematically on a computer; ‘what water looks like’ is a product of many variables. Water reflects and refracts its environment, and can carry participating media that scatters light as it bounces around inside a water volume. Water can churn and froth, can split apart and rejoin itself, and interacts with other objects in interesting ways. There’s no ‘correct’ way to make water in computer graphics, which is why it’s so interesting to me; every time I sit down to work on a shot that involves water, I try it a different way, seeking to get the best result with the tools I have at my disposal (which, in a software-driven industry, are constantly changing).

One of the ‘easiest’ representations of water is an open ocean. I say “easy” because an open ocean with no objects interacting with the water surface allows us to shortcut or sidestep many problems that are otherwise difficult to deal with. Back in the late 90′s, a guy named Jerry Tessendorf theorized that we could take a set number of wavelengths and use them to represent an open ocean.

To get all supernerd about it: there exist in the world several data-gathering “sea state” mechanisms — ways of mathematically-describing “this is how the ocean looks right now”. These are real-world mechanisms…sometimes an array of sea buoys are used, other cases use satellite data, etc. The data captured by these methods is then collected into a “wave spectrum”, which is basically a collection of wavelengths that describe what the ocean surface is doing at a given moment. We can take these spectra and perform an inverse Fourier transform on them to yield a pattern that, when applied to a patch of mesh in a 3d package, can yield something that resembles an ocean.


While this approach to representing an ocean has been refined and improved over the last decade, the general idea remains the same. “Tessendorf waves” are almost invariably what you’re seeing when you watch any film that leverages computer graphics to represent an ocean. King Kong, Avatar, Tintin, Brave, Cars 2…these are films I’ve been fortunate enough to work on that all feature this approach to representing oceans.

I’ve never lost my interest in trying to make my own servicepieces for this project, but the road for doing so is notably slower than for cooking. I’ve been taking lots of classes as I can afford to, to learn things like machining, welding, woodworking, etc. in an effort to piece techniques together to understand how I might craft a servicepiece myself. I recently read about services such as Shapeways and Ponoko that offer users the ability to 3d print an idea in ceramic. This idea is particularly interesting to me, as you have to use a computer as the input to a 3d printed piece. Because I spend so much time representing reality on computers, I wanted to see how I could explore this approach as a way of making a servicepiece. And because I have a long and intimate interest in working with water, using Tessendorf waves as a source of inspiration seemed like a unique way to make it my own.

I started roughing out my idea with zero regard for the design constraints inherent to 3d printed ceramic. The FX department at Pixar makes heavy use of Houdini as a tool for generating various effects, but I didn’t know it before coming here 3 years ago. I still have a lot to learn, so I decided to use it as my prototyping suite, forcing myself to get to know it better. I really love Montgatina’s Tabula plate series, so I started with it as a reference. My first question for myself had to do with the scale of waves I wanted to involve.



On the left, I’m representing an “annulus“; when making an ocean for a given shot in a film, I only care about water that the camera sees, so I often leverage a shape that makes sense when viewed from the camera, but viewed from other angles looks a bit funky. In the center, I’m playing with more of a ‘mid-scale’ ocean, like what you might see standing at the bow of a giant cruiseliner. On the right, I’m getting extremely close to the water, as if I’m bending over the edge of a rowboat and nearly touching it.

Of the three variations, I liked the aesthetics of the rightmost one. It felt fundamentally more beautiful to me — elegant, cleaner and (as Sarah pointed out) way nicer to scrape a fork across. The other ones I imagined to be annoying to actually eat off of.

I had some questions about the amount of detail I want to see at that scale, so I cloned off several versions of the rightmost one, adjusting things like wave height and the overall ‘resolution’ of the wave (i.e. how much fine detail we see). At lower resolutions (seen on the left) the plate started to look less ‘watery’ to me, and at higher resolutions (on the right), the whole pattern looked more like a mountain range that I was viewing from afar rather than water. The centermost one felt the best to me, so I forged on with that one.


I imagined this design being nice and big…15-17″ or so, and maybe 3″ high. I tweeted these renders to Shapeways and Ponoko, asking them their thoughts on how reasonable this design was. Ponoko responded soon to tell me that it was basically awful. It’s way too big than what can be 3d printed, and structurally was unlikely to survive the process. Shapeways pointed me to this outline of the design guidelines for printing in 3d, which I realized I probably should have consulted first. I read it roughly, paying closest attention to the bounding box specifications: these made me realize I was going to be creating a much smaller plate than I’d originally imagined.

I spent a few days vacillating about this; on the one hand, I really, really love the above design. But I worried the only way to make it was to hand-sculpt it, which took away some of the poetry I find in literally printing the work I do every day. A friend suggested I could CNC mill this shape from something like wood, make a mold from it, then slipcast porcelain from this mold. This sounded like a possibility, but also meant quite a bit more expense (having a mold made from a shape can cost over $500, and that didn’t account for the time and money I needed to spend to teach myself how to do all these steps). I’m not ruling out this approach (I’d love to learn how to do this, in fact), but I thought it might also be worthwhile designing within the constraints of the 3d printing process, both to learn and to appreciate what it has to offer.

So, I redesigned my plate. I did so from an informed position, knowing at least how I wanted the wave pattern to look. After a couple tries, I landed on an idea like this, measuring about 7″ square.


I tried a few variations; I compared what the straight edges felt like in contrast to allowing the natural contours of the Tessendorf pattern to come through.



Shapeways/Ponoko also offer glazes in black as well as white. I think a black plate would be pretty badass, but decided to opt for white for the first attempt.

A Tessendorf pattern displaces a surface both positively and negatively; a flat plate gains both mounds (at the wave crests) and ‘bowls’ (at the wave troughs). I felt strongly that I wanted my plate to have both; I really love the idea of being able to plate sauces or liquids in the troughs of the wave patterns, rather than only have mounds to work on. This meant the plate had to be fairly elevated, to allow for downward displacement. It’s not obvious from these renders, but this design featured a flat bottom (if you flip the plate over, the surface you see is flat, rather than inversely-contoured to match the top surface), which I would learn later has significant ramifications.



Shapeways charges by surface area for 3d printed ceramic, while Ponoko charges by volume. I uploaded this design to both to compare pricing. For this design, Shapeways yielded a price about $50 cheaper than Ponoko, coming in at around $120 total. I placed my order, then sat back to wait for the robots to do my bidding.

Three days later, I got an email from Shapeways; they had found issue with my design, and had rejected it. Rather than refunding my money, they gave me a ‘store credit’. They said they could give me a refund if I threw a big fit (they worded it more nicely than that, but that’s how I read it), but they wouldn’t move forward with printing this design…I needed to fix it.

The problem was in the ‘bowl’ area of the plate: the negatively-displaced top surface came within 3mm of the flat bottom surface, yielding a wall thickness of 3mm. The minimum wall thickness for a portion of printed ceramic is noted in the design guidelines as 6mm, but Shapeways recommended 8-10mm would be more appropriate for the plate.

Going back and  rereading the design guidelines more carefully, I took note of the wall thickness minimum and maximum. Wall size needs to be between 6mm and 15mm, with an avoidance of sharp thickness transitions. I realized that if I lifted the top surface so that my too-thin area was at least 8mm thick, I would end up with areas that were way thicker than 15mm (because, recall, the bottom of the plate was flat).


These design guidelines felt inscrutable to me, and I was feeling frustrated. I had emailed Martin several years ago asking questions about working with porcelain, and he had recommended a book for me to read up about it. I went back through and read it more carefully over several days, and did a lot of reading about how ceramics are fired (something that, until now, I hadn’t bothered to research much).

Ceramics (be they clay or forms of porcelain) have a water content at the time they’re sculpted. 3d printed ceramic is no different; small dots of clay are deposited in a designed shape by the 3d printer’s nozzle, and are held together with a binding agent. While the clay is wet, a piece is said to be in a ‘green’ state. It’s left to dry for a period of time; the faster it dries (e.g. by heated air), the more of a differential forms between the dried clay and the wet clay underneath it. If this differential is too high, the clay can warp or crack. This phenomenon isn’t unique to clay; I’ve experienced it when dehydrating ingredients too fast. The Yuba I cooked recently, in fact, did this; the yuba was a big ball when I put it in the dehydrator, and I tried accelerating the rate of dehydration for one batch by turning up the heat of the dehydrator. The outer surface dried and tried to shrink, but the interior wasn’t dry yet, so the exterior cracked open like a layer of paint. Clay can do the same.

To reduce the potential of this happening, one can seek to reduce the opportunities for large humidity gradients; e.g. by keeping the wall thickness no smaller than 6mm and no larger than 15mm. The idea is that you want to let water evaporate from the clay at a steady, constant rate. When the clay is fired (to permanently harden it), if any water still exists in the clay, it turns to steam, again causing cracking or warping as the steam tries to escape.

Learning about this fascinated me, and I spent several more days closely-inspecting all the plates and bowls we have in the house. Almost all of them exhibited this exact property; I measured points all around Martin’s plates I’ve bought over the years, as well as several Montgatina bowls, plates we own from Crate and Barrel, and even some Hadley crockery from my mom. All of it features very smooth, soft transitions from wide to narrow measurements, with no individual part drifting too far outside the measurements noted. I suspect different materials have different drying rates, and therefore allow for slightly-different design constraints, but for the most part the fundamentals hold true here.

Bearing this in mind, I started reworking my design again, paying much more attention to the contours of the plate and how thickness was distributed throughout it. I realized that solving my thickness problem would mean contouring the bottom of the plate to match the curves of the top, and realized this is exactly why Martin’s plates exhibit this property. I also noticed a few characteristics in Martin’s plates that I felt would be nice to incorporate, namely the raised outer edge to give more a sense of depth to the deeper areas of the plate. There are some visual tricks I feel like he’s playing to solve structural problems without throwing the overall feeling of weight out of balance. I eventually landed on something that looked like this.



One of houdini’s strengths is that it allows for entirely modular design methods (we call it “proceduralism”). This meant I could adjust some areas of the plate without affecting others. Once I’d established the outer edges, I could shift around the actual wave pattern, ‘scouting’ for an area of the pattern that looked nice. Incidentally, this is exactly how we choreographed waves hitting the Venture as it crashed onto Skull Island for Peter Jackson’s “King Kong”: we generated a large ocean mesh, then ‘slid’ the wave pattern around hunting for a good-looking wave that could smash against the boat and push it in the direction we wanted the action to be.


After a few more finishing touches, I again sent the design off to Shapeways to print. I noticed that by contouring the bottom of the plate, I’d lessened the volume but increased the surface area. I was curious of the price difference between Shapeways and Ponoko at this point, but because I had “store credit” at Shapeways, I pushed forward with them without investigating further. Ponoko had told me earlier that both outfits actually use the same production facility, and I presume the design constraints result in their pricing structures being very similar once all thickness issues have been addressed.

About a week later, I got an email from Shapeways, this time confirming that the plate had been produced and was en route to me. I held my breath for a few days; honestly, I really liked my renders but didn’t expect the final product to look nearly as slick. I braced myself for a funky color difference, a less-shiny finish, or worse: cracks or warps formed during the firing process.

It turned out my fears weren’t unfounded; when the plate arrived it was significantly warped; all of the edges were way lower than in my design, and a few of the corners had sagged significantly. The plate as a whole had sagged so much that the ‘bowl’ area I’d tried so hard to preserve was nonexistent. On top of this, there were several weird anomalies that were introduced during the printing process that weren’t in my original design: an odd ‘stripe’ and an area of ‘lumpiness’. The glaze was notably less glossy than I’d imagined too.

I tried plating the dish I made this weekend on the plate, and photographed it in a way that accentuated its interesting parts while obfuscating its weaknesses. Those photos can be seen in the main post for that dish, but here’s a more unapologetic one that clearly highlights what this thing looks like. I wouldn’t ever actually present a dish to someone on this; the quality’s just not there.


I wrote Shapeways to ask whether my expectations are too far off for this process; I have not yet heard back as of this writing. I still think the design is pretty interesting, but if this represents the limits of this technology, it’s not the right tool for me to use for making my own serviceware. I think I’ll be looking into some ceramics classes, but hey…science.


Sardine, Niçoise Olive, Dried Tomato, Arugula


Nice people rock.

The other day, I got back to my office after a meeting to find a package of dried sardines sitting on my desk. They were from a coworker, who later told me he’d recently taken a trip to Japan; his mother-in-law served these at a dinner, and while eating them he asked “Hey, do you happen to have any more packs of these things? There’s a guy I work with who was looking for them not long ago.”

I’ve been looking for Tatami Iwashi for about 4 years now. It sounds, from my friend’s account, that these aren’t terribly exotic in Japan, but have proven to be exceedingly difficult for me to find stateside.  A few months ago, I gave up and decided to just try making my own. Having been gifted a pack of the Real Thing (and extremely hesitant to take a Gentleman’s B on my last attempt), I made it again. Lest anyone else find themselves traveling down a path that requires they intersect with this incredible ingredient…here ya go.

20130505_alinea_0004 20130505_alinea_000720130505_alinea_0012

I can’t say that I regret having tried making my own. Not only did I learn quite a bit about using Transglutaminase, but there’s an interesting difference between a dried sardine sheet and a glued sheet of baby anchovies.


The true Tatami Iwashi uses no meat glue; it’s quite brittle, and the individual fish are thinner and longer. Rolling them around a dowel before frying them into shape required several spritzes of water to soften the sheet, but the moistening also causes the fish to ‘unstick’ themselves from each other a little, so I had to find a balance with how much water to spray on them.


They were also ‘stickier’ than the Anchovy experiment…I had a hard time getting them to release from the wooden rod after the frying step. I ended up destroying about 1/4 of the cylinders I made; the sheets are more delicate than the glued versions. I had a curious thought halfway through about mixing a slush of transglutaminase and spraying the sardine sheet with it to help strengthen it.


The final taste was milder and way less fishy than the anchovy version. Sarah noted that she preferred the anchovies; the whole bite was more assertive and flavorful. I think though that sardines make more sense in a tasting menu context; anchovies stay with you for a while, and I wouldn’t want a diner to be licking anchovy bits out of their teeth three courses down the road.


The rest of the dish remains exactly as easy and straightforward as it did last time; one can put it together in under an hour if one has the rest of the ingredients on hand. It really is all about highlighting the exotic Tatami Iwashi.


I’ve noted before the story about Thomas Keller and his rabbits; that story came to mind as I was working on this dish. For a friend to think of this project of mine while on a family vacation is incredibly, incredibly touching to me. I’m at a loss for words for how grateful I feel, and as I opened the package of sardines with a pair of scissors, I felt an overwhelming desire to do this the best I could with them. I wanted to make this delicious, to photograph it well, to treat it with the respect and thoughtfulness it deserved.

This gratitude doesn’t end with this dish. For everyone who’s ever been supportive and encouraging to me as I wade through this, I am humbly and deeply grateful.



Squab, Thai Peppercorn, Strawberry, Oxalis Pods


Oxalis might be the most annoyingly-vile weed ever. It looks a bit like clover, but it has these funky, torpedo-shaped ‘fruits’ it produces after flowering that, when touched in even the slightest way, launch seeds explosively and generously. The seeds are quick to sprout and seem indestructible, producing another plant within a few weeks.  I’ve been waging battle against this stuff in our yard since we moved here; no amount of pruning and pulling seems to eradicate it, and every time I accidentally brush one of Oxalis’ seed pods and those little seeds go flying, I think “DAMN it”.

This weekend I moved into a new tactical phase with the stuff; I ate it.



Oxalis is hands-down the most lowly, un-exotic ingredient in Alinea’s cookbook; not recognizing the plant, I scoured several exotic seed catalog sites trying desperately to find “Wood Sorrel” (Oxalis’ common name) seeds. I went to the Berkeley Horticultural Nursery and asked if they sold it — they looked at me like I was nuts, but I wasn’t sure why. It wasn’t until I asked at work if anyone knew anything about this plant that my friend Pauli replied “Yeah, my boyfriend’s been trying to get rid of that shit from our backyard for years. You can have as much of it as you want”. It was at this point I looked in my own backyard, and sure enough, this stuff was already occupying about half of it.


Oxalis is an extremely-common ground-cover plant that resembles clover and tastes sort of like Lemonhead candy; it’s tart/acidic and has an earthy, grassy note.  This is due to presence of Oxalic Acid (huh!); Oxalis is one of the primary sources for harvesting Oxalic Acid, even. In high doses, Oxalic Acid is toxic and can be fatal, but chewing on leaves from an Oxalis plant gets us nowhere near these levels any more than a sip of vinegar poses a danger of death. Both the leaves and the seed-laden fruit of the Oxalis plant are edible.

In this dish, the Oxalis pods are a tangy garnish for Squab, Strawberries, and Thai Long Peppercorn custard.  This dish as a whole is incredibly relaxing to make; it’s an exercise in patience rather than dexterity, so it made for a really nice weekend project. There are a lot of “Do this, then wait 5 hours, skimming form time to time”-type things going on here, so be ready for it to take a solid couple of days to complete, but be ready to drink a lot of beers in the down times too. It’s the high-end equivalent of making BBQ in a smoker.  A nice benefit to this is that I had lots of time to take more photos than I usually can.

This recipe starts with 4 whole squabs; I found these for about $12 a pop at Berkeley Bowl. They’re smaller than a Cornish hen, and are usually sold in the freezer section (since they don’t tend to move nearly as fast as chicken or other more-common meats).


After thawing the squabs gently overnight, I broke them down; I trimmed the breasts into a diamond shape, and separated the tenderloins. I removed the legs and gizzards, and separated these from the carcass. Everything gets used in this dish.


I vacuum-sealed the breasts and tenderloins and put them in the fridge to store overnight; on the first day I wouldn’t be needing them so I wanted to keep them in good shape.


I rinsed the legs and gizzards, and while they were left to drain in the sink, I toasted some thai long peppercorns. I *love* the smell of these things, and Sarah inverse-proportionately hates them. A minute after I started toasting them she called down from our loft “What are you doing down there? It smells like a church in here.” If you’ve ever visited a Catholic mass when they fire up the incense, you might have an idea of what this smells like. I really find it intoxicatingly complex and interesting.


I ground the toasted peppercorns in a spice grinder, then mixed with equal parts sugar and salt, yielding a Thai Long Peppercorn Cure. I packed the legs and gizzards in this cure and left them in the fridge for a few hours.



While the legs were curing, i cut the rinsed squab carcasses into small pieces with some kitchen shears. These would go into making Squab Stock.


The pieces are browned in a roasting pan, then cooked a bit longer with fennel and onion. After this has sweated for a few minutes (the book is very specific not to allow anything to brown to the point of caramelization), I transfer the mixture to a large stock pot, trying to keep out as much rendered fat as I can.


I add some thyme, peppercorns, vermouth, and tomato paste, then let this mixture simmer for 5 hours. The smell of this is amazing to me; the vermouth-fennel combination is something I imagine must have a specific name (and I bet it’s French), it seems so fundamentally-perfect. During the simmer, I skim the stock every 15 minutes or so, to remove impurities that have collected near the top. I move the pot partway off the burner so that the convection currents push the impurities to one side, so it’s easier to skim.


While this is cooking, I toast some more Long Peppercorns, and again grind them to a fine powder.


This powder is mixed with some cream, sugar, and salt, and brought to a boil over medium heat. I then let it steep for 20 minutes or so, to allow the cream to absorb as much flavor from the peppercorns as possible.


After the steeping is complete, I add some Iota Carrageenan, a stabilizer that can form soft, elastic gels in the presence of calcium. The carrageenan is hydrated by both high temperatures and high shear, so I hit it with an immersion blender as I bring it to a boil again, blending for about a minute before straining the mixture into a shallow pan to let it cool. The pan is lined with cling wrap (I sprayed the pan with a bit of water first, to help the cling wrap stick to the bottom better, which lets me get it smoother. Wrinkles in the cling wrap make it hard to get the resulting gel out later), then left to cool until the gel sets. The amount used here yields something of the consistency of Key Lime Pie: a smooth, custardy texture that tastes either really lovely or like a church, depending on your polarization.


By this time, the legs and gizzards had cured, so I removed them from the fridge  and rinsed the excess cure from them.


I packed the legs and gizzards separately into vacuum bags, and sealed them with some canola oil. These were cooked in a warm water bath for another 5 hours or so, then chilled overnight in the fridge.


While the legs/gizzards were cooking, I finished the stock; I strained it and reduced it, continuing to skim until it was reduced by half. The “reduce something by X” thing was always a point of confusion for me until Erik showed me this pretty cool trick: grab a wooden skewer at the beginning of the simmer and dip it, marking the depth of the water line. Then, cut a notch at the halfway point (or whatever point you need to reduce to); then you can just keep testing until you hit the notch.



The next day, I started by working on Strawberry Sauce. I first pureed some fresh strawberries with some red wine, red wine vinegar, and sugar.




This mixture was simmered with some of the squab stock until it was reduced to something the consistency of BBQ sauce.


Next up was making Neutral Caramel Squares, which ultimately sit atop some skewered Oxalis pods. To start, I mixed some fondant, glucose, and isomalt in a saucepan.


A couple of years ago, I did some experiments to wrap my head around the nuances of various types of sugars like those used here. While that experiment didn’t really pan out the way I wanted, I learned a ton and got to leverage some of that here. Isomalt is a sugar that does not caramelize the way cane sugar does (at 325F, cane sugar browns ["caramelizes"] from the Maillard reactions occurring at that temperature. Isomalt does not undergo these reactions). Inclusion here helps offer a ‘buffer’ to the fondant and glucose (which CAN brown), because we ultimately do not want browned sugar for this recipe. The cookbook further instructs me to cook this mixture to 160F, “taking care not to allow any browning”. This is a misprint; at 160F we haven’t even gotten rid of most of the water in the mixture, much less do we run a danger of burning anything. I saw this and almost-immediately recognized it to be wrong, largely because of all the mistakes I’ve made doing this in the past.


The “hard crack” stage of cooking sugar (which is the point at which, if you cool your mixture, it forms a hard, crackable candy rather than something like taffy or syrup) is around 310F. “Caramel” stage is 320F-325F; for this recipe, we want a hard crack sugar, and we want to get it as close as possible to the caramel stage without blowing it and causing any browning. A different recipe in the book targets 316F for this purpose, so that’s what I aimed for here. After it hits this stage, I pour it out onto a prepared sheet tray and let it cool. We can see here I incurred a bit of browning but not too bad; some of the browning carries with it some nice complex flavors and keeps this from tasting like straight-up neutral sugar.


I broke this “sugar glass” into shards and ground them in my spice grinder to yield a white, snowy, fluffy powder.


This powder is sifted through a strainer onto another sheet tray lined with teflon-coated paper.



It’s then baked at 425F for just a few seconds, to allow the powder to melt. The longer you melt the powder, the more surface tension starts to take over and cause things to bead up into bubbles, so this really only needs to take a few seconds (the book mentions 6 minutes, which is way too long for my oven). Once I had melted it, I removed it and quickly cut it into small squares, then left the sheet to cool on a rack. After a few minutes, you can peel up the squares; they’re very thin and delicate, and super-crisp.



While the sugar was cooling, I diced up some more fresh strawberries into small dice…


…and cut up a few others into bigger bits. These bigger pieces were macerated for a while in a mixture of red wine and sugar.


I rewarmed the cooked legs and gizzards in another warm water bath, then removed the meat from the legs and minced up the gizzards. These are combined with shallot, butter, more squab stock, salt, and a dash of red wine vinegar to make Squab Rillettes.


The last step before starting the plating process was collecting my garnishes. I went out to our yard to do some glorious weeding, carefully collecting several dozen Oxalis pods.


I also gathered some of the nicer-looking Oxalis leaves and flowers, as well as some sorrel from the garden.


And finally, I decided to use some micro Ice Lettuce that I’ve been growing. I first came across this pretty-fascinating plant last year at the French Laundry. It’s a crunchy, mildly-lemony succulent that’s covered in crystalline, dewdrop-looking capsules that look like frost crystals (so in these photos, the ice lettuce isn’t wet). It’s just so damn pretty and interesting that I had to try growing some.


To begin plating, I heated some canola oil and seared the reserved squab breasts, crisping the skin and browning it nicely while basting with melted butter. At the same time, I seared the tiny squab tenderloins, and let both of these rest while I assembled the rest of the plate. I spooned a puddle of the strawberry sauce in the center of the place, then added mounds of strawberry dice, rillettes, and a cube of the long pepper custard.

I sliced the breast into several bite-sized portions, and placed these in the sauce. Then I garnished everything with my collected greens.


The final step was placing a caramel square on top of the skewered Oxalis pods, and warming it gently with a kitchen torch to melt it into the pods. This was placed on top of a macerated strawberry, and the whole thing sits on the rillettes.


For as much as I loved this dish, Sarah exactly disliked it. I find squab to be an intoxicating flavor; when I worked with it last year, it was paired with foie gras and licorice, and I found the flavors to be over-the-top opulent and delicious. I chalked it up to the inclusion of the foie, but it turns out that squab itself is very rich and opulent. The idea of pairing it with tart flavors (licorice and watermelon last time, lemon and strawberry here) helps cut the richness and keep it balanced. The long pepper pairs excellently with it, adding a depth and complexity that I find quite compelling. But apparently squab is a bit polarizing; last time I worked with it, a couple friends didn’t finish their plates (including Sarah), and she reminded me this time that the flavors here just really aren’t her thing.

She conceded, however, that the raw strawberry dice were to her liking. So that’s a feather in my cap.



Wild Turbot, Shellfish, Water Chestnuts, Hyacinth Vapor


Hi there friends. Something a little bit neato has happened; unbeknownst to me, Sarah has been submitting this blog to Saveur to entice them to possibly make mention of it on their site. Incredibly, it’s managed to be nominated for Best Culinary Science Blog of 2013! Pretty sneaky, Miss Sarah. Anyway, if you’re reading this and feel so inclined to proffer a vote for it on Saveur’s site, well that would just be tops. If not, that’s ok…we can still be friends.

A lovely friend at work asked me a few weeks ago “How far into this thing are you?” When I mentioned I was (at the time) around the 85% mark, she remarked “Ugh. That’s a tough place; are you able to keep it interesting for yourself?”  I was really delighted at this question because it was so incredibly insightful. Truth be told, the ~70-90% stretch has taken effort to keep focus; after working on something for so long, it’s easy to want to break away from it and try new things. There’s a rigid aesthetic I’ve wanted to work with here — photographically, culinarily, and on the written side of things — and my mind wanders to what I’d like to try doing next. I obviously try pushing my own boundaries a bit here, but don’t want to break the rules completely and want to keep things as ‘excellent’ as I think the subject matter itself deserves. Maintaining the balance becomes effortful, and sometimes I fail.

This isn’t too dissimilar from what it feels like to be on a project at work. Working on a film is pretty rad at first; everything is new, there are new rules and aesthetics to explore, and I learn a lot quickly.  But films take a long time to make; as the project wears on, looking at the same things every day gets a little old and it becomes more effortful to keep the quality level (which, for me, is a product of enthusiasm) high.

I think this is a pretty common phenomenon in creative fields. Eileen Moran, a pretty amazing person I worked with at Weta in New Zealand, recognized this when the crew was working on large-scale, arduous projects. She used to do this neat thing near the end of a project — during the last big push when everyone was tired from working long hours and was creatively drained — where she’d start counting down how many shots we had left to complete for the film. Every few days she’d send out a still frame from the film with the shot count number somehow integrated into it. On King Kong, a shot I’d worked on got this treatment: there were several big rock structures in the original shot, but she’d photoshopped them to look like a giant, rocky “107″. In a studio of nearly a thousand artists working around the clock to complete a project, these images and the underlying empathy helped keep the studio feel small and lovely, and served as a rally cry for everyone to dig in and make it through one last big push.


Of course this project isn’t ever UN-interesting, I just need to look around a bit more for ways to keep challenging myself. Not long ago, Sarah and I talked about an idea for a video we wanted to make. Actually, what we talked about was an aesthetic; her style thus far has been largely photojournalistic, with natural light and a ‘farmhouse-y’ calm vibe that I love. But we talked about doing something more honed, abstract, and refined…something more like a moving version of one of my photographs on here, or what you might find in the Modernist Cuisine cookbooks. To me this means (among other things) using studio lights rather than relying on the mishmash of daylight/overhead lights that we usually work under, and much more focus on strictly-controlled backgrounds and environments. I’ve never worked with “hot” lights before (all my photography gear involves strobe lights [the kind that flash quickly rather than staying on]), so I knew this would be new for both of us.


We flagged this dish as a potential for this experiment, largely because of the floral component and because the recipe is fundamentally simpler than many others remaining to complete. From a high-level view, this dish is very similar to the Shellfish Sponge dish from last autumn: several types of shellfish are cooked in a broth of white wine, vermouth, fennel, shallot, and garlic. The shellfish are cleaned and trimmed for tidy presentation, and the resulting shellfish stock is used both for a custard (mixed with cream and carrageenan) and as a poaching liquid for some fresh Turbot. The accents and garnishes here are different, but the underlying “really fancy shellfish flavor vehicle” idea is the same.


It took me several weeks’ worth of homework to line up everything for this dish, starting with the floral component: Hyacinth. Hyacinths are beautiful, colorful flowers that smell like Easter. They bloom in early spring, starting (at least in the Bay area) in mid-February and lasting through Mother’s Day. Those interested in growing their own Hyacinth typically buy the bulbs in the fall season…something I wanted to do but ended up forgetting until it was too late. I’ve been keeping an eye out for them in flower shops for the past several months.

In this dish, Hyacinths are a garnish over which boiling water is poured; the fragrance of the flowers is pretty notable in and of itself, but the hot water releases a potent cloud of it that smells like a Victoria’s Secret in the early 90′s…not that that’s a bad thing, of course. Because Hyacinths are one-shot plants (you buy a bulb, wait 6 months, and get 1-2 flowers from the plant before it dies), they are the most expensive garnish I’m every likely to employ: at $9-$12 a pop around here, a single plating costs around $25 bones just for the flowers.


The pricey garnish is in pretty good company though; the hero ingredient of the dish — Turbot — is no fiscal lightweight either. True European Turbot (in the UK: “TUR-butt”, in the US: “TUR-bow”) is a largely-Mediterranean flatfish that’s considered a delicacy because of it’s tender texture and delicate flavor. My friend Deanie, a former pastry chef at the likes of Manresa, Coi, and Ubuntu, raised her eyebrows when I told her what my cooking project was for the weekend. “Whoa turbot…that’s some French Laundry shit you’ve got going there.” (pause) “This is a pretty expensive hobby of yours, huh? How’s Sarah feel about that?”

I didn’t dignify that last question with what I feel at this point is a pretty obvious answer.

Deanie’s right though; turbot falls into the category of fish you’re likely to find featured at the whitest of white tablecloth restaurants and rarely elsewhere, a fact I learned when failing to find it for sale at my beloved Berkeley Bowl. After calling around to several other fish markets in the area, it seemed that securing turbot was a bit touch-and-go; one market in San Francisco said they could get it, only to call me back a couple days later to tell me there’d be a delay of 2 weeks on the order I’d placed with them. While trying to do some online searching for other options to consider in the area, I stumbled across Browne Trading, which incredibly offered to overnight me an entire fish. Having not previously heard of Browne Trading, I again consulted Deanie, who confirmed that “yeah, those guys are super-legit”. So, having planned to work on this dish on a Saturday, I called Browne and ordered their “2-fillets of Turbot” option, then arranged for a half-dozen or so blooming Hyacinths from a lovely little flower shop in Berkeley to be delivered for the weekend’s adventure.


I worry it sounds incredibly privileged and douchey to casually admit to sourcing these things; my rule for myself is: if I haven’t every tried something, I’m willing to save up and go all out to experience working with it, but I’m happy to substitute with local/seasonal/(cheaper) things when I’m already familiar with an exotic ingredient (e.g. my substitution of rhubarb for gooseberries in my last dish). I’m still interested in learning, and having not worked with either hyacinth or turbot before, I wanted to get a clear picture in my head for both.

Other than the sheer exotic-ness of its ingredients, the dish itself is remarkably straightforward to prepare. After preparing the shellfish as I mentioned above, some of the remaining stock is cooked with cream and carrageenan, then left to cool until it forms a custard. At the same time, I peeled and quartered some Sunchokes (also called Jersualem Artichokes) and cooked these with cream until they were tender, then blended them into a puree. While the sunchokes were simmering, I diced up some fresh water chestnuts into 1/8″ dice and stored these in a small container until service.

The last step was working with the turbot. Deanie wasn’t kidding when she said Browne was super-legit; I was expecting my fish to turn up pre-frozen in a vacuum-sealed pouch. What showed up instead was a disproportionately-massive cooler packed with freeze-gel packs. Buried inside this cocoon of coolness was a plastic tub containing a single incredibly fresh turbot filleted in two. The tub was smeared with small bits of fishmonger gook and had “Hemberger: Turbot” scrawled on the top in Sharpie; it smelled like the ocean. I guess some might be jarred by the complete lack of fancy packaging beyond this, but I was totally delighted. There were no labels declaring weight or fancy-but-unnecessary ‘preparation instructions’, no glossy inserts like you might get from Snake River Farms. No, this was just “Here’s the FUCKING INCREDIBLY FRESH FISH you wanted, and a free sample of some sea air from Maine. You’re welcome.” Super. Legit.


I divided the turbot fillets into individual portions and seal these in vacuum bags with butter and some of the shellfish stock, to be cooked gently en sous vide at 138F for 20 minutes or so. Then it was just a matter of “heating things up gently and putting it all together”: a bed of sunchoke puree and water chestnut dice is put down in a bowl first, then some turbot portions are placed atop this. Some warm shellfish custard is poured around this assembly and left to set. While this is resting, I stemmed some hyacinths (and lilac, which Ana had at her shop as well. I figured what the hell.), placed them in a larger bowl, and brought a pot of water to boil. Once the custard was firm, I gently placed the trimmed shellfish meats around the borl, and garnished everything with tarragon, fennel fronds and pollen, hyacinth petals and flowers, and some fresh chervil from the garden. I called Sarah down, poured in the boiling water, and we ate.

And it was delicious. Turbot indeed has a lovely, indescribably-unique texture and flavor that oozes subtlety. It’s not really ‘fishy’, and it rides alongside the fennel/vermouth/shellfish flavors seamlessly. The night after I put this dish together, I cooked the remaining turbot and paired it with celery root puree, very simply…I think something hard or aggressive like pan frying, lemon pepper, those sorts of things are too overpowering for this fish. Accounts I’ve read online of people trying to panfry it seem to result in the meat turning to mush. This might be why it’s so frou-frou; it demands a light, considered touch.


It’s probably obvious at this point that I’ve omitted my usual lot of photos of ingredients; this is the Big Fail of this dish. In trying to realize our vision for the video we wanted to make of this dish, Sarah and I realized that there was a lot we have to learn; setting up and breaking down specific backdrops against which to shoot severely limits the working space I have to actually prepare things, and the logistics demanded so much of our attention that there ended up being a lot of detail we failed to capture. We weren’t sure how to tightly-control the shooting of things as they cooked on the stove, and it was hard to try to solve that problem and still cook the food properly. Also, because we were both playing DP and Cinematographer together while I was trying to cook, we were consuming ingredients under the lens of her video camera, which meant I had nothing left to shoot stills of with my own camera (lest I have to go out and buy new ingredients). Basically we were underprepared and didn’t end up with a cohesive collection of footage that accurately tells “the story of making this dish” in the aesthetic that we envision, and I’m left with few in-progress photos of things.

But, some of the footage we captured IS kind of compelling to me, so I took a stab at editing it together into something interesting anyway. It’s not really a ‘story’ per se, mostly just footage we shot of ingredients as we were figuring things out. There’s a section in the middle where I’m trying to trim the littleneck clams using a description in the recipe that was hard for me to understand; basically I’m trying to trim a pretty ugly-looking thing into something that’s tidy and has a good texture, so I trim the outer ‘stuff’ and the foot (which is tougher meat), then cut off what I think is the clam’s stomach, which is filled with sand and can be gritty to eat. If there’s someone out there who knows a better or more-proper way to do this, please set me straight.

Other than that, I amused myself by shooting lots of photos of the final dish, which I found overwhelmingly beautiful. I was especially captivated by the juxtaposition of forms of the whiteware with the forms of the flowers; it all looked so pretty together.




Thankfully I was left with a few spare hyacinth and lilac flower heads…so I’m trying a different experiment with them…


9 dishes remaining…


English Peas, Tofu, Ham, Pillow of Lavender Air


The ignition point of lavender is around 420F or so.

That’s not what I’m here to talk to you about today though. No, today we’re going to talk soybeans.


Specifically, we’re gonna talk about making soy milk, which in turn is used to make yuba and tofu. My first adventure with yuba was about a year and a half ago; just long enough for me to have forgotten most of what I learned about making it. Soy milk itself is a pretty straightforward beast: let some soybeans soak overnight in water until they plump and swell, then blend into a frothy puree. Simmer this mixture to sanitize it, skimming foam from it the whole while. What you’re left with is pure soy milk.

For me, the soy milk had a tendency to froth quite a bit; skimming this froth eliminated nearly half the volume yielded by the recipe, so I ended up doubling it (well, doubling it 6 times if you count all the misfires I had here, but I’ll get to that in a moment). Also, remembering that the first time I did this I had a massive mess in my kitchen from overfilling my blender, I blended the soybeans with water in small batches to keep things manageable.


For this dish, soy milk is used in three different ways. First, I simmered the soy milk very gently until a ‘skin’ formed on the surface of the liquid. This ‘skin’ is called a Yuba.


Once the skin is thick enough to be handled (about 8 minutes or so), it’s removed from the soy milk and laid as flat as possible on some parchment to dry. For my first several tries at this, I used a metal skimmer to lift the yuba from the milk. This sucks; the yuba wants to cling to the underside of the skimmer and freeing it results in lots of tearing of the curd and loud swears from me. I realized that I needed way less surface area touching the yuba when removing it; looking around my kitchen for a better tool, I almost smacked myself when I realized the best answer: chopsticks. A pair of chopsticks results in way less clinging and offers way more articulation than the fat bulky skimmer. Using Asian tools to make Yuba was obvious only in hindsight, I admit with no small sense of self-ridicule.


After a half hour or so, the yuba has dried enough to be collected into loose mounds; it’s sill quite delicate at this point and can fall apart easily if you’re not gentle. It’s left to dry for several more hours…


…where it looks more like a beige shrinky-dink, with a plastic-like feel and pliability. I left these to dry gently rather than dehydrating them because ultimately the yuba mounds will be deep-fried. The sudden immersion in 375F oil causes water trapped internally to turn quickly to steam, but the dried, elastic outer yuba skin prevents the steam from escaping. This makes the mounds puff and turn crispy on the outside, but (depending on thickness) the interior can remain chewy and omelette-like; an interesting juxtaposition of textures. Dehydrating them removes too much water, which causes them not to puff as readily.



The Yubas (is that proper plural of yuba?) are paired with blanched peas and smoked ham. I’m meant to buy a full smoked ham, but it’s tough to find such a thing that hasn’t been processed for deli consumption, plus I’m really enamored with cured meats, so I decided to go with cured smoked ham, which is often known as “speck” in Italian delis. I diced some of the speck into cubes and cooked these briefly with the peas and some butter into a ragout. More speck was sliced very thinly, gathered into a similar-shaped mound as that of the Yuba, and fried over high heat to char a side of it. The scraps of ham were cooked with a bit of the (now-thickened) reserved soy milk from making Yuba to yield a Smoked Ham Nage.


The third stage for the soy milk was the Big Adventure for this dish: lavender tofu. The recipe calls for an ingredient called “Nigari” for this; after doing some research, I learned that nigari is mostly magnesium chloride, and is derived from evaporated sea water. The magnesium chloride serves as a coagulant when mixed with soy milk; i.e. the solids in the milk clump together. Nigari can be found in dry (crystalline) form (hard to find) or pre-dissolved in water (more-easily found, but nearly impossible to discern concentration of solid nigari to water):


To make lavender tofu, I’m meant to infuse some thickened soy milk with a sachet of dried lavender flowers, then toss in some nigari and stir ‘until the soy milk turns to soft tofu’. Truth time: I’ve rarely eaten tofu. My only experience with it is a couple of times in stir-fry, where I tend to pick around it, suspicious of its odd halloumi-like squeaky/spongy texture and its frequent claim to be a meat substitute. So the flavor and texture of tofu are variables I don’t have a good control for in my head. On top of this, it turns out that the very brief instruction in the cookbook dances over a critical component of getting this right.

The book says to warm the soy milk and toss in the lavender sachet. In almost every case I’ve seen involving infusing something, ‘warming’ means ‘simmering’. I imagined I needed to get the milk hot enough to make a tea of sorts with the lavender, so that’s what I did. After a few minutes of simmering, the soy milk smelled lovely and floral, so it was time to throw in the nigari. A friend at work helpfully translated enough of a clue off this bottle to get me going: use 1% of liquid nigari in soy milk to yield tofu. I threw in the nigari and stirred, and this is what happened:



The mixture almost immediately separated into tiny curds and whey. Hm. The book doesn’t warn me about this. I stirred some more. I waited. I debated. Something about this didn’t seem quite right. I scooped out a spoonful of the curds and put them in a strainer to let them drain. They kinda-sorta fused into a crumbly, feta-like lump.

The photo in the cookbook clearly shows the tofu looking sort of like pudding or custard; there’s no graininess to it. I took briefly to google to see what I could find, searching for “grainy homemade tofu”. One comment mentions that graininess could be a product of using too much nigari. Ugh, I figured now I had to start doing test batches of this to try to figure out what the actual concentration of my nigari brine was.

The catch was, I was out of soy milk. I’d ruined a whole big batch of it, and trying again meant steeping another big batch overnight, and also throwing out most of the work I’d done for the other components for the day. I quickly ran out to the grocery to try buying a liter of store-bought soy milk, to see if I could shortcut this process.


Store-bought soy milk, I learned, is notably different than homemade soy milk. It’s usually sweetened (unless explicitly marked otherwise), and strained so finely that it’s thickened with things like carrageenan to restore its body. It also often has added nutrients to supplement diets that replace dairy milk with soy.

I came home with a box of the stuff; it tasted notably sweeter (despite being of the unsweetened variety) and quite a bit different from the stuff I’d made. I quickly brought some of it to a simmer with more lavender, switched off the heat once it smelled nice and fragrant, and started adding drops of nigari brine 5 at a time, keeping count as I went and stirring/waiting for a couple minutes between each addition. Nothing happened for a long time, but as I crept up to the 1% mark, the mixture started to clump again. So, the ‘adding too much nigari’ wasn’t the problem.

I knew I was switching around too may variables to be able to understand this, so I went out again and bought more soybeans (3x as many as I needed, just in case), and came home and soaked them overnight. The peas and ham that had been sitting warm in my oven for the past few hours wouldn’t hold up very well overnight, so I ate most of it and threw the rest out. Frustrating.


In moping around later that night and reading more about making tofu, I came across several articles explaining the differences between tofu textures and types.The separation thing I was experiencing isn’t a bad thing in and of itself; firm tofu is made this way, which is very similar to cheese-making (I kept having deja-vu back to the mozzarella balloon adventure a while back); the separated curds are removed from the whey, left to drain, and then pressed into a mold to yield firm tofu. But I suspected firm tofu wasn’t what I was after.  I read a bit about about “silken” tofu, which seemed (from photos) more like the texture I was meant to be aiming for. I found several articles describing getting fresh-made silken tofu from nice restaurants in Japan, where it’s presented jiggling on a small plate like a delicate custard, topped simply with soy sauce or sesame oil. This type of tofu is (of course) considered more rare and delicate, and store-bought silken tofu doesn’t quite compare with the texture of fresh-made stuff.

Finally I stumbled across a recipe for making the stuff that mentioned one critical step that differed from what I’d been doing: the nigari is added to the tofu while it’s cool, THEN heated. When done this way, the tofu doesn’t separate into curds and whey, and instead coagulates into a smooth custard when heated gently (the recipe mentioned steaming it).

I woke up the next morning feeling confident about this. I’d soaked 3 batches’ worth of soy beans, and blended them in small batches before bringing the whole lot to a simmer again. I needed to simmer the mixture (skimming away the copious foam again) for around 20 minutes or so, to evaporate enough water such that the milk was at least 12% soy solids. After this time, I threw in my lavender and covered the milk to let it steep off the heat for another 20 minutes or so. Then I strained the mixture through a chinois into a bowl, and set the bowl in my sink, which I’d filled with ice water. I let the milk chill until it was around 70F or so. During this time I preheated my oven to its lowest setting: 170F.

I measured 0.5g of my nigari brine into a small ramekin, then measured in 50g of the cooled lavender soy milk. I stirred to combine the two, and noticed the mixture immediately thickened but didn’t break or separate. I placed it in the oven and waited 5 minutes, then gave it a tap; it jiggled gently. I took it out and scooped at it with a spoon; it looked almost exactly like a creme brulee custard!

I tried several more batches, varying things like the cooking time, stirring technique, and nigari amount; they almost all turned out the same. None of them broke (even adding up to 3% nigari didn’t cause any separation like I’d seen before), so it seems the magic bullet is adding the nigari while the soy milk is cool.


The lavender tofu is meant to be served warm, and is garnished with a sprinkle of lavender salt (made by processing lavender flowers in a spice grinder and mixing the resulting powder with sea salt).

Tofu achievement badge earned, I got to work on everything else. I made Yuzu Pudding by first simmering some saffron and lemon zest with some water and agar, and  blending the resulting gel with yuzu juice to yield a smooth pudding. Yuzu juice on its own is nearly colorless, sort of like lemon juice. I like that Alinea uses natural means to color things at times, and the saffron used for the color here also offers a nice depth of flavor to the pudding.




The recipe also calls for making a ‘gooseberry coulis’, which is a sauce made from pureed fruit. Cape gooseberries come into season here in the fall; they’re not readily available during the spring months, so I tried to think up a suitable substitute. I wanted something that had a bright tartness like gooseberries, and also something that might offer a nice visual component. I settled on rhubarb, which is looking really lovely at the moment.



I simmered the rhubarb in simple syrup until it was tender, then pureed it in the blender with some grapeseed oil to form an emulsion. Rhubarb coulis.


I blanched more peas, and remade the ragout with more ham dice. I also re-seared some ‘ham blossoms’ and yuba mounds, and kept all this warm in the oven.


The final step is making a lavender pillow. I have some experience with the scented pillow thing from the White Bean dish a few years ago; I had borrowed a Volcano vaporizer from a friend to use for filling the bags inside the pillows. Volcanos are astronomically expensive; wanting to avoid provoking the ire of the Minister of Finance in our household, I haven’t been able to justify buying one of these. I feel like a cheaper solution is out there. I tried a few experiments this time around to see if I could sidestep using one:

–I tried loading some lavender into my Smoking Gun and holding a small kitchen torch juuuuuuuust the right distance away from the flowers to liberate their oils without toasting them. This is extremely difficult; I couldn’t get anything that didn’t have a smoky note to it. Not that that’s bad, it’s just not what I wanted here. There’s also an added problem where unscented air is being blown out of the device while the flowers are being heated; a ‘bell curve’ of scent forms, and capturing just the heart of it is really tough.

–I tried an experiment with a non-forced-air vaporizer. The advantage of an actual vaporizer is that heat is controlled independently of air movement. So, a heating element in a Hotbox vaporizer warms air inside it, and when that air is drawn through the chamber holding the flowers, the essential oils are instantly vaporized  and drawn up through a hose leading away from the heating element. I wondered about using the principles that allow a spray gun to work: namely, capping the tube leading away from the heating element with a T-joint, then forcing air quickly through this to draw vapor up and away from the box. I used a blow-dryer for this, but I still had the problem where there’s a few seconds where air is rushing out (and inflating the pillow bag) but no scent is coming through, and control is again very difficult.

It’s clear why the Volcano is such an appealing option for this; it’s forced air and has very little waste. It’s also extremely controllable. There are other forced-air vaporizers available, and had I not found a Volcano to borrow again, I might have bought and tried one. As it stood though, a friend of a friend offered the use of theirs. This one was a digital version of the analog one I’d borrowed previously, with the ability to finely-control exit air temperature and a nice display telling me the current temperature of the device. Getting to use it again reminds me how badass and cool these things are.



Pillows inflated, I plated everything and garnished the dish with fresh pea sprouts and shoots and several flowers from the garden. The dish was really delicious; peas and ham are a pretty obvious combination, but the brightness of the yuzu and rhubarb flavors added some levity that kept it from feeling overtly heavy. The creamy custardy lavender tofu and the crispy yuba puff offered a nice dimension to things; yuzu/rhubarb/lavender all in one bite is hela delicious. As a whole, this dish is super top-notch.

Completion of it also represents something I’m a little bit excited about: I have 10 dishes remaining in the cookbook. I suspect it’ll still take me the better part of a year to get through them all, but it’s going to feel good to get down into the single digits.