Latest Update
Claude Opus 4.8 Is Here, Work AI Race Begins
2026.05.29Hi, this is Team Carat. 🤗
Today, we’re covering Claude Opus 4.8 and the new race for work-focused AI, Perplexity moving into office apps, image AI that changes only the outfit, and music AI that can switch genres mid-song.
Today, we’re covering Claude Opus 4.8 and the new race for work-focused AI, Perplexity moving into office apps, image AI that changes only the outfit, and music AI that can switch genres mid-song.
Claude Opus 4.8 Is Here, Work AI Race Begins
Anthropic released a new AI model, Claude Opus 4.8. The key point is not just better answer quality. Claude is moving toward breaking complex work into smaller steps, running multiple tasks in parallel, and carrying the work through to the end. Anthropic also says coding work is stronger, and Fast mode lets people use the same model faster and at a lower cost.
In particular, Dynamic Workflows lets Claude plan the order of work, run multiple smaller tasks at the same time, and review the results before moving forward. It is closer to assigning work to an AI than simply chatting with one.
What matters now is not only whether AI gives good answers, but whether it can finish complex work reliably. Claude Opus 4.8 is a signal that this race is starting in earnest.
📌 3 AI updates today
1️⃣ AI search moves into office documents
Perplexity introduced Perplexity for Microsoft Office, which works inside Word, Excel, PowerPoint, and Outlook. It helps write documents, analyze spreadsheets, create presentations, and handle emails from a side panel inside the apps.
AI search is moving from giving answers in the browser to sitting inside the documents and inboxes where people already work. For workplace AI, the important question may become less “where do I ask?” and more “can it help inside the workspace I already have open?”
2️⃣ Image AI that changes only the outfit
Black Forest Labs introduced FLUX Virtual Try-On. You can provide a person photo and a clothing image, and it creates a try-on image while preserving the face and body shape, as well as details like logos, stitching, and prints.
For fashion stores, this could make it faster to create many try-on images without shooting a new model photo every time. For shoppers, trying clothes on their own photos before buying may start to feel more natural.
👉 You can also try virtual try-on on Carat
3️⃣ Music AI that changes genres mid-song
ElevenLabs released Music v2, its music generation model. It improves vocal and instrumental expression, and highlights features like changing genres within a single song, such as moving from opera to heavy metal, or regenerating only the section you want to fix.
AI music is moving from “generate one full song at once” to “edit the intro, verse, and chorus separately until the track feels finished.” For creators, the key is not just the final output, but how precisely they can refine the mood they want.
👉 You can also try music generation on Carat
🧪 Today’s prompt tip
Make a Geoje Yaho-style gyaru vlog meme
A meme combining K-pop idol RESCENE Minami’s gyaru concept with the “Geoje Yaho” trend has been popular on social media. Today, we prepared a prompt that turns one portrait photo into a vertical card with two gyaru-style vlog screenshot frames.
Here’s a result made on Carat.
Use the ChatGPT Images 2 Medium model and create an image with the prompt below.
Use image 1 only for the adult woman's identity and facial structure. Create a vertical 4:5 Korean pop-culture Instagram carousel card with two stacked realistic smartphone vlog screenshots. The same woman appears in both screenshots, transformed into a clearly recognizable modern Japanese/Korean gyaru style.
Mandatory gyaru styling: voluminous glossy warm-blonde hair with soft big waves and face-framing pieces; dramatic doll-eye gyaru makeup with extended black winged eyeliner, thick upper and lower false lashes, visible lower-lash definition, glittery aegyo-sal highlight, bright filtered complexion, pink blush, glossy overlined pink lips. Add maximalist Y2K/kawaii accessories: layered silver and pearl necklaces, heart charms, phone strap charms, multiple rings, rhinestone bracelets, tiny plush keychain, decorative hair clips. Outfit must be hyper-feminine and kitsch: pale pink cat-ear cap or fuzzy animal-ear hood, fitted turquoise or baby-blue top, lace camisole layer, ruffled mini skirt or layered lace trim, pink/teal color accents.
Top screenshot: inside a small Korean tour bus, close selfie angle, she does a gyaru peace sign near her cheek with a playful pout. Bottom screenshot: sunny Geoje seaside harbor with boats and low mountains, she crouches in a cute exaggerated pose, showing the layered outfit and accessories. Add a rounded white Korean caption sticker near the lower left that clearly reads exactly: ‘거제도 갸루 브이로그♡’. Fan-edit collage layout, realistic smartphone video-still texture, high-key daylight, slightly overexposed social-media filter, cute chaotic Y2K vibe. Avoid natural minimal makeup, avoid plain casual traveler styling, avoid celebrity likeness, no brand logos.
The key is not to write only “gyaru style.” Specify the hair color, lashes, aegyo-sal highlight, accessories, outfit colors, and shooting location so the result clearly reads like a meme.
Choose ChatGPT Images 2.0 Medium on Carat, upload your photo, and paste the prompt above. By changing only the location name and caption text, you can easily adapt it into a Busan, Jeju, or Seongsu-style meme card.
That’s it for today. We’ll keep making it easy to follow how AI is changing work and creative culture. 🙌
Was this helpful? Share with your friends
Share
Get daily AI trend updates
