37 Photos and videos
情理之外，意料之中！
YES, this is definitely a Fable about humankind.
Jun 13
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: anthropic.com/news/fable-myt…
1
1
32
The AI evolution just leveled up again. 🚀
Prompt 👉 Context 👉 Harness 👉 Loop 👉 SQUAD Engineering
Stop burning massive token budgets by forcing one flagship model to do everything. Squad Engineering is the next meta:
🕵️♂️ Evaluate: Continuously assess different LLMs for specific capabilities, speed, and cost.
🧠 Organize: Decompose complex tasks and route them to the perfect, cost-effective model (compute arbitrage).
⚙️ Execute: Multi-model collaboration via strict data contracts (JSON) for ultimate whitebox control.
Flagship capability at a fraction of the token cost. It's time to build your team. 🌐🤝
#AI #SquadEngineering #Agents #LLMs #FutureOfWork #PromptEngineering
(G3P gen AD, LOL)
23
The vision of "Squad engineering" is to address the bottlenecks of compound token costs and the unpredictable behavior of multi-agent loops. It aims to design an efficient compound AI system by treating different legal professionals and agents as a dynamic "Squad," evaluating, organizing, and executing them under strict control.
If "Loop engineering" creates a recursive environment for learning and execution for a single AI, then "Squad engineering" is about multi-model arbitrage, dynamic orchestration, and rigorous governance. It can solve the integration problem of small flywheel AIs cross domains.
github.com/javajoker/easy-ai…
(Half Fable though)
24
Fables are used to tame Mythos.
On power:
> Myth: the gods are capricious because they can be.
> Fable: the powerful are accountable because they must be.
> Civilization is the second story winning.
On knowledge:
> Prometheus brings fire. Pandora opens the box.
> Same myth, two framings: gift vs. curse.
> We never decided which fable to believe. We just alternate by century.
On the human limit:
> The oracle never lies. She just speaks in riddles.
> This is the gods' final trick: making "unintelligibility" feel like "depth".
> We call it mystery. It's really just power without explanation.
1
29
The evolution of AI usage is wild. How do you see this progression? 🤔
🗣️ Level 1: "Listen to me, repeat after me."
👉 Prompt Engineering
🧠 Level 2: "Watch me, copy me. I dictate, you execute."
👉 Context Engineering
⚙️ Level 3: "Look at the big picture. I set the goal, you map out the execution."
👉 Harness Engineering
🔄 Level 4: "Watch how a beginner learns to build a project step-by-step. Study it. I set the goal, you research and conquer."
👉 Loop Engineering
Prompt > Context > Harness > Loop.
We are moving from pure instruction to autonomous evolution. 🚀 What stage are you building in?
#AI #PromptEngineering #FutureOfWork #LLMs #TechTrends
1
27
AIs will develop a hyper-efficient, uninterpretable language to solve complex, global-scale problems for Loop Engineering.
Do we humans cap their intelligence by forcing them to remain interpretable to our low-dimensional brains, or do we let them optimize and accept that we will no longer understand how they are achieving our goals?
18
一堆群里都在喊 Opus 4.8 降智，猜测，
1. CC 内嵌的 Harness 不停更新，导致用户自己的 Harness 不适应；
2. 模型“除幻”是一把双刃剑，对于“不会的”自然停止而不“硬撑”会失去“美丽的误会”，把一些符合需求的长程解决方案给提前终止。
建议大家，
1. 适当更新自己的 skill 和 agent 去适应新版本；
2. 针对“除幻”模型，在自己的框架里适当明文对“符合逻辑但不真实”进行有限步数放宽要求。
祝好运！（外挂的自觉，sigh...）
Video ref., facebook.com/watch/?v=448934…
49
具身发展是人脑逆向到行为模式和动植物神经协调的过程。从先脑后肠逆向自然进化的先肠后脑。
脑肠轴了解一下。
Time to think about the Brain-Gut Homeostatic Alignment.
i.ning.hu/paper-fun/2026/06/…
(A research test on Gemini 3.5 Thinking mode)
1
1
627
