Grok 벤치마크·Alibaba 통합 메모리·vLLM 모임·엔지니어링 계층이 이번 주 트렌드였다

Elon Musk

@elonmusk

And Grok 4.6 comes out in a week

Your workflows just got a brain boost! @n8n_io now officially supports Alibaba Cloud & Qwen. With Bring Your Own API Key (BYOAK), you can now plug in Qwen’s power directly into your automation canvas. Total control over your tokens, total flexibility for your AI Agents. · Secure enterprise data connection · Complex Agent orchestration · Automated SaaS decision-making Try the new node: https:// docs.n8n.io/integrations/b uiltin/cluster-nodes/sub-nodes/n8n-nodes-langchain.lmchatalibabacloud … Try QwenCloud: https:// click.qwencloud.com/m/20000000978/ #AlibabaCloud #n8n #Qwen #AI #WorkflowAutomation #AIAgent #SaaS #AI #Productivity

One PolarDB, one MemOS—persistent, never-blackout memory for AI. PolarDB & MemTensor launch a one-stop AI memory solution: relational, vector (PGVector) & graph (PolarAGE) retrieval in a single PolarDB-PG instance, cutting P99 latency up to 89.2%. #AlibabaCloud #ApsaraDB #PolarDB

vLLM is coming to Taipei! The first vLLM Taipei Meetup is filling up fast. Meet our speakers: Tun Jian Tan, Embedded LLM Tiezhen Wang, Inferact: vLLM Kimi K3 Support and Q3 Roadmap Kuan Fu Liu, Embedded LLM: Maximizing LLM performance on limited GPUs Ray Wu, WEKA: Extending vLLM's KV cache beyond GPU memory Dr. Hen-Hsen Huang, Academia Sinica: Practical paths to trustworthy local LLMs Wen-Bin Lin, Twinkle AI: High-throughput, reproducible LLM evaluation with vLLM Details: Sat, Aug 1 · 2–5 PM Taipei New Horizon Building Request to join:

From prompt → context → harness → loop → graph engineering: The list keeps growing, and every new term gets treated as a replacement for the last one. In reality, however, each layer wraps the one before it, and the cleanest way to tell them apart is to ask what a single unit of work looks like. > Prompt engineering is the message: The model remembers nothing before this call, so the prompt has to carry the full universe of what it needs. a role, the background, the instructions, a few examples, and a format. When the output falls short, the skill is working out which ingredient lets you down, not rewriting the instructions every time. The unit of work is one input. > Context engineering is the memory: Across many steps, the window is finite, and the available information is not, which forces a curation step. A curator keeps what matters, compresses what is useful but bulky, and drops the rest. Good curation is mostly about knowing what to throw away, not packing more in. The unit of work is what stays in the window. > Harness engineering is the machine: On its own, a model just generates text. The harness gathers what it needs, runs it, calls tools or sub-agents, and verifies the result with tests or a judge. That verify step is the entire difference between calling an api and running an agent. The unit of work is one pass through the machine. > Loop engineering is the run: One pass rarely finishes the job, so something has to decide whether to run the machine again. That decision needs a goal defined upfront, brakes like max iterations and budget caps, and a completion check that is automated rather than felt. An agent that stops asking for tools has ended its turn, which is not the same as finishing the task. The unit of work is the whole run. > Graph engineering is the coordination: Once several loops have to work together, you need to say what runs when, what runs in parallel, and who checks whom. Nodes do the work, edges decide what runs next, and shared state flows between them. A single loop is just a one-node graph with an edge pointing back at itself, which is why graphs govern loops instead of replacing them. The unit of work is the whole job. Here is the part that ties it together. Prompt and context both live inside the harness gather step. The harness is one pass, the loop decides whether to run that pass again, and the graph decides which loops run at all. Zoom out, and the unit of work gets bigger. Zoom in, and you are back at the prompt. That also tells you where to debug. Find the layer whose unit of work broke, then fix that layer. The prompt is the easiest layer to edit, which is why it keeps taking the blame for failures that live three layers up. My co-founder published a deep dive on graph engineering, covering the core idea, how to get started, shared state, routing you can trust, and when a graph is genuinely overkill. Read it below.

Quick reminder of what's ok vs not ok with harnesses used for playing ARC-AGI-3: 1. Not okay: harnesses that were custom-made to solve the benchmark or that contain knowledge about the benchmark format / contents. 2. Fine: general-purpose API settings that were not developed for ARC-AGI-3 and that are available to all API users. In the past, we've had a lot of back and forth with OpenAI about how to best test their models, especially with regard to compaction. I'm glad they're starting to figure out the answer. Of course, if each provider uses different settings when getting their model tested, it creates a potential parity issue. My take is that this is fine as long as the settings and the cost are clearly reported.