I have a few projects in flight this week.
My fixation right now is ballotine. That's when you debone, stuff, truss, and roast a whole bird.
It seems like a great Thanksgiving recipe: there aren't any bones to complicate table-side carving or turn the meat red! Plus it accomodates stuffing and looks whole enough to be Thanksgivingy. It doesn't solve the undercooked breasts problem, but I guess that means the overcooked breasts have a nice veneer of old-school French provenance.
But it's also a terrible Thanksgiving recipe: it's pretty technical, and Thanksgiving is an important meal to not ruin! It's a bad time to whip out an advanced technique that you aren't any good at.
This year, though, I don't have anyone coming for Thanksgiving, so it's ok to ruin dinner! So my mission is to get good at ballotine. I did a chicken on Friday, a capon on Sunday, and I'm working towards something for a dinner party on Wednesday.
I'm getting better at the deboning. On the chicken, I did the gradually-scrape-meat-from-bone technique, and it was awful. It worked in the end, but it took forever and I constantly felt like I was doing it wrong. On the capon, I used Jacques Pepin's technique where you peel the flesh from the cage with your hands. I still felt like I was doing it wrong, but it only took 5 or 6 minutes, so that's a win.
This was my first capon. Apparently if you castrate a rooster, they lose interest in sex and exercise and gain a passion for lounging and drinking cream. This gives them more intramuscular fat. Capons are harvested at all sizes, but the conventional ones at my local supermarket are about 10 pounds, which is a few pounds larger than the chickens. Verdict: definitely juicier, but not a big flavor difference. I think swapping a conventional for a heritage bird, or a young for an old one has more of an impact. I wouldn't buy a larger bird than I wanted just to get a capon, but if there were both capons and chickens in my desired size, I'd probably go for the capon.
I like the presentation, and I definitely want to keep practicing deboning, but I'm not thrilled with the overcooked breasts. I see that some folks get around this by doing a ballotine-like skin-wrapped-log of -just- the breasts, then perhaps incorporate the (differently cooked) legs into the stuffing. Separately, I wonder if I could carefully remove the meat from the skin, sous vide the breasts and legs separately, recombine them into the log, put the skin back on (meat glue?), and crisp with hot oil. That might be next.
I have a program for organizing my movie library and displaying a Netflix-like
browsing interface. The program has accreted functionality over time: first, I
just dumped an ls of my movies folder into sqlite, because that is Always A
Good Thing To Do. A while later, I wrote a thing to try to fetch metadadta from
TMDB. The UI came next. Etc.
This week, I added two integrations I'm pretty happy with. First, a thing to pull ratings from Metacritic, enabling a sort-by-rating. Second, a thing to pull my reviews from Letterboxd, enabling a show-only-unwatched and a sort-by-my-personal-review-score.
As a bonus, I added a thingy to show my last n reviews on my homepage. This
is a bit of a hack. All of my little personal services are compiled into the
same binary, and I run it with different configuration on various computers,
including on fly.io for public-facing servcies. The movie library service runs
on my home server, so its data lives in an sqlite file at home. But my homepage
runs on a rented machine from fly.io, so how does it get the movie data?
Answer: a clunky expect script that copies the db file over and restarts the
fly box. I am very excited for the day that Hermit, my syncing embedded
database, is featureful enough to support use cases like this.
I picked work back up on an incremental-search feature for my little task runner. I started this a few months ago, but realized it'd require writing my own line-aware wrapping display thingy, and kind of lost interest.
My personal services binary (see above) uses the task runner, and I noticed that if I leave it running for a long time, the log display gets gunked up and stops working. Profiling says that this is because of wrapping in the bubbletea-default viewer component I'm using: each frame rewraps the whole log, so it's O(log length) rather than O(screen size).
So now I have to rewrite the log viewer anyway, so I may as well take this opportunity to add incremental search. It works, kinda! WIP in a branch here.