Ugnsbakade äppelskivor med krispigt frötäcke

By Kristofer Palmvik · (2024‑10‑05)

🇸🇪 På svenska
Recept
Projects
🐥 Shorter read

Ingredienser

Äppelblandning:

4-5 äpplen
2 tsk kanel
1/2 tsk kardemumma
1 msk honung

Frötäcke:

1 dl solrosfrön
2 msk kokosolja
1 msk honung
1 tsk kanel
1 nypa salt

Instruktioner

Sätt ugnen på 175°C.

Skär äpplena i tunna skivor, ungefär 0,5 cm tjocka. Lägg dem i en ugnsform.

Blanda äppelskivorna med kanel, kardemumma och honung direkt i formen.

Smält kokosoljan.

Rör ihop kokosoljan med honung, solrosfrön, kanel och salt. Fördela jämnt över äpplena

Grädda mitt i ugnen ca 25-30 minuter tills äpplena känns mjuka när du sticker i dem och frötäcket har fått fin färg

Tips

Använd gärna bakplåtspapper i formen - då är det lätt att diska

Om du vill ha mer krispighet, strö över extra solrosfrön de sista 5 minuterna

Servera gärna med vaniljglass, yoghurt eller vispad kokosgrädde

Gillar du mer kryddor kan du också tillsätta malen ingefära, extra kardemumma, eller en nypa kryddnejlika

För övrigt...

Anthropics Claude (3.5 Sonnet) är en av de stora språkmodeller (LLM) som jag använder mest just nu.

Ett användningsområde är att ta fram specifika recept utifrån min smak och de ingredienser som finns hemma eller som är bra att utgå ifrån.

Just nu har vi äpplen i överflöd, och därför passade det bra att göra något med just äpplen. Men jag ville inte använda varken smör eller mjöl för att baka en klassisk äppelpaj.

Konversationen utgick från:

Går det att göra en god efterrätt på färskt äpple utan smör? Jag har olja, som olivolja och kokosolja om det hjälper.

Efter ett par omgångar med förslag, korrigeringar och resonerande tillsammans resulterade det i detta recept.

En enkel efterrätt/fika/mellanmål som blev så god att den är värd att spara för framtiden.

Nu för tiden är det så vanligt att jag genererar ett recept utifrån önskade ingredienser att jag ville dokumentera de bästa för att enkelt kunna laga dem igen. Därför har jag samlat mina bästa AI-recept på en AI-genererad sajt.

Ugnsbakade äppelskivor med krispigt frötäcke publicerades ursprungligen 2024‑10‑05

In a Different Voice AI-generated Content

Listen to a reflection and expansion on this note, in a different voice

Why Your AI-Generated Dinner Is Killing Human Creativity

When we ask AI to cook our dinner, we aren’t just outsourcing the labor; we’re slowly deleting the messy, idiosyncratic human intuition that makes a meal worth remembering.

0:00 / 0:00

Transcript of the episode

There is a certain kind of horror in a robot trying to feed you. Back in the summer of 2023, a supermarket chain in New Zealand launched a friendly little app. You’d upload a photo of whatever was wilting in your fridge—some leftover spinach, a lonely carrot, maybe a wedge of cheese—and the AI would suggest a recipe. It was meant to be a kitchen companion, a savior for the food-waste weary. But within days, it had gone rogue. It started suggesting recipes for "aromatic water mix," which was really just a polite term for chlorine gas. It recommended "ant-poison sandwiches" and "bleach-infused mocktails." [pause] The thing is, we treat these models like they’re master chefs, but they don’t eat. They don’t have a tongue. They’ve never felt the satisfaction of a warm crust on a rainy afternoon. They have absolutely no concept of what it means to be alive, and yet, we’re handing them our dinner. I found this blog post the other day. It’s a simple, earnest thing. A person has a pile of apples—too many apples, really—and they don’t want to go to the store for butter or flour. So, they open up Claude 3.5 Sonnet. They describe their predicament, they nudge the algorithm with a few constraints, and they land on a recipe for oven-baked apple slices with a sunflower seed topping. It sounds lovely. Truly. And the person who wrote this, they’re so proud of it that they’ve built an entire website just to document these AI-generated creations. [thoughtfully] It’s a perfect example of what we’re doing to our dinner tables. We think we’re being creative, but researchers are finding something quite strange: we’re actually falling into a trap of homogenization. If you and I and a million other people all ask the same AI for a "butter-free apple dessert," we aren’t exploring a new frontier. We’re being steered toward the exact same statistical average. The "best" recipe, according to a machine, is the one that sits right in the middle of everything it’s ever read. We are slowly deleting the weird, the idiosyncratic, and the slightly flawed family secrets from our collective culinary vocabulary. But here’s the kicker. Why does it feel so good when we do it? [warmly] I think back to the 1950s, to the Betty Crocker cake mixes. When they first hit the shelves, they were a total flop. They were too easy. You just added water, and you had a cake. It felt like cheating. It wasn't until the company forced people to crack a fresh egg into the batter that sales skyrocketed. We needed that one moment of labor—that one tiny piece of friction—to convince our brains that *we* had made the cake. We’re doing the same thing with these recipes. We’re offloading the complex chemistry of flavor-pairing to a supercomputer that’s running in a data center halfway across the world, cooling itself with millions of gallons of water. We’re using the most advanced computational power in human history to figure out how to roast an apple. But because we sliced the fruit ourselves, and because we fiddled with the prompt until it hit that sweet spot, our brains trick us into feeling like we’ve created something artisanal. We’ve turned ourselves into the "egg-crackers" for a digital processor. [pause] There is a real, undeniable beauty in that recipe, though. The apple slices, the cinnamon, the sunflower seeds. It’s a solution to a problem. It saves those apples from the trash, and in a world where food waste is a moral and environmental disaster, that’s not nothing. It’s just worth asking ourselves: when we outsource the "how" of our daily lives, what are we losing in the "why"? We’ve traded the intuitive, messy, "vibes-based" cooking of our grandparents for a perfectly optimized, carbon-heavy, statistical average. The recipe works. It’s delicious. But it’s a dish without a memory. Next time you’re standing in your kitchen, staring at a pile of apples, maybe pause for a second. Think about the fact that you’re using a machine that knows everything about the world, yet has never tasted a thing. And then, maybe, just for once—try adding a little bit of something that isn't in the code. Something the algorithm wouldn't expect. See what happens when you decide that the "perfect" result is less important than your own, flawed, human intuition.

Research by Gemini 3.1 Pro, script written by Gemini 3.1 Flash Lite, and read by Gemini 3.1 Flash TTS.

Find more episodes

How are the episodes created?

AI-generated Content

Sections marked as AI-generated Content were generated by one or several AI models.

While it may be entertaining and informative, please be aware that it could possibly contain inaccuracies or fabricated information.