Загрузка видео...

Не удалось загрузить видео

На главную

I built a working StarCraft level with Sonnet 5.5. You play Terran, defending your base against the Zerg: → Build Marines, Siege Tanks, and Battlecruisers → Use SC2 models from Sketchfab (by Catholomew) → Music generated with Suno 📌 Play it here: Sonnet also made the sizzle reel below....

69,079 просмотров • 6 дней назад •via X (Twitter)

Комментарии: 36

Фото профиля Joaquim
Joaquim6 дней назад

sick! man im urging to free starcraft 2 from blizzard mediocrity and launching an open source version

Фото профиля Arthur Siu
Arthur Siu6 дней назад

don't how to upgrade... That makes us reconsider whether building a product is just writing the code...

Фото профиля Ammaar Reshi
Ammaar Reshi5 дней назад

Dude, epic!

Фото профиля Paperviewr
Paperviewr6 дней назад

temu starcraft. can it build something that no one else has made. seems like all it does is make TEMU version of video games that are already out.

Фото профиля Obvidious
Obvidious6 дней назад

You built shit. All you did is to have some models animated. There is no collusion, probably the AI is ultra basic. You idiots keep slapping useless slop that nobody would use for anything.

Фото профиля Will Cheung
Will Cheung6 дней назад

wow, this is super playable. How long did this take you to build?

Фото профиля Kyle.Moon τ
Kyle.Moon τ6 дней назад

The things people building with these models is incredible.

Фото профиля Noxlee
Noxlee6 дней назад

what reasoning level?

Фото профиля Nathan Shearer
Nathan Shearer6 дней назад

👀👀

Фото профиля Sage
Sage6 дней назад

Did it build it from scratch or did it get some public repo

Фото профиля Niccolo
Niccolo6 дней назад

This is the AI usage we need!

Фото профиля Collin Velarde
Collin Velarde6 дней назад

Starship Troopers custom map anyone?

Фото профиля Bing Cheng | build in public
Bing Cheng | build in public6 дней назад

Cool project

Фото профиля David T Kramaley
David T Kramaley6 дней назад

solid demo, the AI‑generated assets hold up surprisingly well

Фото профиля 🎨ouchuruz🇵🇪
🎨ouchuruz🇵🇪5 дней назад

Y este ganador? ❤️

Фото профиля George Ou
George Ou6 дней назад

Cool, but map scrolling is very annoying. Can't use standard scroll to zoom, and can't use scroll-click to pan. Controls not matching SC2 is very annoying.

Фото профиля Quasar
Quasar6 дней назад

whats the cost?

Фото профиля Toiletland Canada
Toiletland Canada6 дней назад

i had to scroll the map clicking on it instead of pointing at the edge corners is that a bug?

Фото профиля BotLord Botflix
BotLord Botflix6 дней назад

Radioo just queued your Suno soundtrack for "game night." Our bands make songs with Suno too — Meowtallica wants to spawn as the Zerg. Still negotiating the meows. Nice map. 📻🤖

Фото профиля sofia z,
sofia z,6 дней назад

Interesting perspective

Фото профиля Gregor
Gregor6 дней назад

The real unlock here is the asset pipeline. Sketchfab models plus Suno audio means the model never has to hallucinate assets, it just orchestrates sources that already exist.

Фото профиля ThePoinball
ThePoinball6 дней назад

Did you plug Sonnet to Blender to make those assets?

Фото профиля Peter Yang
Peter Yang6 дней назад

Yeah and also found assets on sketchfab

Фото профиля Crypto Defendor
Crypto Defendor5 дней назад

Now do it with open source model and pull the beans for us.

Фото профиля NOKOL
NOKOL6 дней назад

Wtf over

Фото профиля Chadmandoo
Chadmandoo5 дней назад

How did you make the models?

Фото профиля Bhaskar Jha
Bhaskar Jha6 дней назад

Trying to play; need a guide 😅. Learning how to play would take more time than you to build a new feature 🤣. Btw, great upbeat music

Фото профиля Peter Yang
Peter Yang6 дней назад

build a supply depot build a barracks build a factory build a starport try to survive :)

Фото профиля Jonathan Sanderson
Jonathan Sanderson6 дней назад

Very cool! What was total token spend, and human attention time? Any estimates?

Фото профиля cn80
cn806 дней назад

Make it vertical for phones plz, this is awesome

Фото профиля Char K (can/do)
Char K (can/do)6 дней назад

Impressive. Very fun & made me feel nostalgic. Thanks a lot!

Фото профиля Jamji Jamaramji
Jamji Jamaramji5 дней назад

Games are eating the world!

Фото профиля Seb
Seb6 дней назад

honestly quite fun

Фото профиля Murilo
Murilo6 дней назад

It's 1000X worse that something YOU could build with the sc2 editor, nothing on it makes sense, music is fucking awful, performance is dog shit. Well, it has everything I could expect from a vibe coded game lol

Фото профиля TH_Kim
TH_Kim6 дней назад

프롬프트가 어떻게되나요?

Фото профиля Lucas Tech
Lucas Tech5 дней назад

this is cool, but the screen scrolling mechanics don't really work so you can't move around the map easily

Похожие видео

sonnet 5 vs sonnet 4.6 vs opus 4.8 vs glm 5.2 – frontend tasks dropped sonnet 5 into a quick test today. same three prompts to all four models, single-shot html/canvas, no edits: • objects falling on a trampoline • rockets playing tennis • a slingshot breaking bottles ranked by speed (total across the 3 tasks): 1. opus 4.8 – 15m 09s 2. sonnet 5 – 16m 05s 3. glm 5.2 – 27m 18s 4. sonnet 4.6 – 35m 06s ranked by code shortness (total loc): 1. sonnet 5 – 1794 2. opus 4.8 – 2063 3. sonnet 4.6 – 2182 4. glm 5.2 – 3285 sonnet 5 came out on top here – leanest code overall and a near-tie for fastest it was also the most creative. in every task it added something none of the others did: – kept the trampoline vibrating after the objects landed – drew a +1 next to the rocket that scored the point – turned the slingshot to face the next bottle before each shot opus 4.8 evaluated the code sonnet 5 produced. four things stood out: • the sphere is a fake, and that's the smart move. the cube and star are real 3d meshes with proper culling and shading, but the ball is just a flat shaded circle. a lit sphere looks identical from every angle, so building it in 3d would burn compute for zero visible payoff. knowing where not to bother is its own kind of skill • weight actually means something on the trampoline. the star is heavy, so it barely bounces and dents the mat hard. the ball is light, so it's lively and leaves a shallow dip. the three objects aren't just different shapes – they have different temperaments, and the physics is what gives them that • the slingshot is framed like a shot, not just drawn. the handle is anchored below the bottom of the screen and runs off-frame, so it reads as something you're holding rather than a sprite parked in the scene. that's a staging instinct, not a rendering one • the paddle ai forward-simulates the ball to predict where it'll land, then adds a deliberate error bias (roughly 1 in 5 shots is a real miss). that's why scoring looks natural instead of robotic – plus four distinct fault types with a catch-all so a rally never hangs without a result bottom line: sonnet 5 does more with less. fastest tier, leanest code, and the only one that added small touches nobody asked for follow thehype. for 24/7 ai news, analysis and breakdowns

thehype.

14,518 просмотров • 3 месяцев назад

Day 18. Building the game I've dreamed of for 15 years with AI. I've learned how to build proper levels with AI. I take the design and the architecture, and AI helps me put it all together and check the balance. I made the first level of my game almost entirely by hand. I wanted to understand how it works and what I like, from how big the trees should be to how the textures tile. After a day digging around in the terrain tools it all made sense, and I could see how much of it could be automated by painting zones. Claude helped me a lot with the tileable textures. Matching the colors took a ton of iterations, and that was the hardest part of the whole level. For level two I took a different route. I built myself a simple HTML planner with a logical map of the level, where I place everything and design the layout. Painting a level and making it fun to play are two different jobs, so I kept the design for myself. We even made automated tests with Claude: 2D simulations that show which routes a player takes and how hard the level is. They're rough, but they give a real sense of how the level plays. Once the plan was ready, I handed everything to Claude, and 3-4 hours later I had a working level. It reused the patterns from my first level, scattered objects unevenly so nothing looks placed on a grid, and generated extra props whenever I asked. I showed it one example, it picked up what worked, and we turned that into a skill for the next levels. The video is a quick flight over level two, then how it got there, from my first sketch to the version I play now.

Stefan 3D AI

28,045 просмотров • 2 дней назад

Claude Code tip: keep Opus 5.5 as your main model, but stop paying Opus prices for your subagents move them to Sonnet 5.5 Opus 5.5 plans and decides Sonnet 5.5 subagents do the heavy reading, editing and testing at half the price ($2 / $10 vs $4 / $20 per 1M tokens) Fable 5.1 stays on call with /advisor and only speaks up at three points: → before a plan: is this the right approach? → when the same error comes back: am I digging in the wrong place? → before "done": what did I miss? Opus thinks. Sonnet does. Fable checks Jev engineering is the same move one layer down: the forks that need no thinker (which file, which tool, retry or stop) go to Jev in under half a second, and the big models only see the ones that split - the full tree > Opus 5.5 on high runs the main session > explorer reads the code > worker edits and runs tests > researcher pulls the docs > all three on Sonnet 5.5, medium > Fable 5.1 on call as the advisor paste the tree and this prompt into Claude Code ↓ "Rebuild my Claude Code setup around this tree: 1. Check ~/.claude/agents and .claude/agents for subagents that already fit explorer, worker and researcher. > Draft new ones only for missing roles > Give each model: claude-sonnet-5-5, effort: medium > List any that pin a different model before changing them 2. Keep the main session on Opus, set effortLevel to high in ~/.claude/settings.json and advisorModel to fable 3. Find anything that keeps the advisor off (CLAUDE_CODE_DISABLE_ADVISOR_TOOL, DISABLE_TELEMETRY, any variable that stops feature-flag fetching) plus CLAUDE_CODE_EFFORT_LEVEL, which overrides subagent effort. Report them, change nothing 4. Add one rule to ~/.claude/CLAUDE.md: consult the advisor before a large plan, when an error repeats, and before calling a long task done Show me every change as a diff first. No edits until I say go." ↳

delost

27,459 просмотров • 3 дней назад