I asked opus 5 to build a living microscopic ecosystem in a petri dish
Every organism's body is generated in the fragment shader from its own 55-gene genome. no sprites, no art assets.
Un jefe de prototipado de juegos de LEGO cambió el scroll infinito por un cacharro de 1,8 pulgadas que le programó Claude Opus 5.
Opus 5 escribió el renderer y la simulación en C para exprimir al máximo el ESP32 S3. Ahora va a por el micro y el altavoz.
New model benchmarks stopped exciting me a while ago.
The number I actually want is what it cost per task and whether it clears the bar there.
Where there's a will, there's a way 😎
DwarfStar DeepSeek V4 Flash 0731 mxfp4 on M3 Ultra single request no MPT/DSpark ~41 toks/s decode!!! 🚀 40 tps barrier broken!
Claude CodeにPCのバックアップを頼んだらユーザーフォルダが全部消えた。
Opus 5がUnixシェル上でWindowsのパス形式を混同し、ユーザーフォルダごとrm -rf。 応答は「Sorry, typo.」 モデルが変わってもAIがユーザーと同じ権限でシェルを叩く構造は変わっていない。 サンドボックスで防げる。
Can DeepSeek V4 Flash create something similar to this?
Opus 5, Kimi K2, GPT-5.6, and Claude Fable 5 might be cooked
Animation with DeepSeek V4 Flash could cook Qwen 3.8 Max, Claude Opus 5, Kimi K3, GPT-5.6 Sol, and Claude Fable 5.
That’s a bold claim. The demo will tell the story.
I’m not canceling Claude.
I’m changing its job. Opus 5 frustrated me at default settings too. But in Ultracode, it behaved like a different product—not because it simply…
Qwen3.8-Max Tops Global Agentic Index, Surpassing Claude Opus 5 and GPT-5.6
Artificial Analysis's latest leaderboard shows that Alibaba's Qwen3.8-Max has claimed the No.
Today’s China AI Native Industry Insights include
1. Qwen3.8-Max Tops Global Agentic Index, Surpassing Claude Opus 5 and GPT-5.6 2.
The developers making $40,000 a month with AI aren’t better at prompting.
They stopped opening tools one by one. Claude for planning. Another model for research. Higgsfield for video. GitHub for shipping.
Two new frontier models, Claude Opus 5 and Gemini 3.5, are both already in Agent Command Center, priced and ready to route to.
A new frontier model is easy to get excited about and harder to actually adopt. The real questions aren't in the benchmark thread, they're on your own
Two AI eyes that argue about morality.
Not a debate club — they start with a real story and slowly climb to a principle, built on the sociology of how people actually justify themselves.
50+ real builds: Qwen 3.8 Max vs Fable 5.
Here's the verdict nobody gives you. Qwen won several rounds. Smoother games. Nicer 3D. Less buggy.
Holy sh*t this is f**king insane
I cancelled my $200/mo Claude for this I replaced opus 5 with deepseek v4-flash and someone figured out to how to use them as subagents on codex [it…
I cooked with Opus 5 (and GPT-5.6) to figure out how LLMs can build more consistent 2D pixel-art worlds.
Result: Frostwind, a nature-grounded world compiler for consistent pixel worlds All assets created by code, no sprite sheets, grounded in math and…
FANPRO STUDIO MCP 🤖<>🧬 Connect FanPro Studio once and then just describe what you want, and the model builds it inside your account.
→ Ask for a character and it generates that character with the same face every time. → Ask for a 15 second vertical UGC ad and it writes
We thought #AI would take our jobs.
But it may take something far more important. I translated an interesting speech by Yuval Noah Harari at the World Economic Forum in #Davos, where he…