最近 Qwen3.8-27B 挺火，出于两个动机我也想测试一下目前本地模型的能力。 第一，我一直在想一种 Local × Cloud 的 24/7 AI workflow： 日常、重复、隐私敏感的任务，不一定需要最强模型。一个「表现不错、成本接近于零、可以永远在线」的本地模型，或许可以承担大量 background work；真正困难的 任务，再交给云端的 frontier model。 第二，我最近也在做 AI hardware。 在产品制作过程中，我开始关心一个问题：一个足够小、能够真正部署在本地甚至端侧的模型，如果可以完成感知、理解、简单决策和绝大多数日常交互，产品体验可能会上一个台阶。 不过目前 Qwen3.8 官方开放的权重还是 2.4T-A95B，27B 并没有正式开放；Qwen3.7 也没有对应的小参数 open-weight model。 于是我退一步，测试目前可以直接部署的 Qwen3.6-27B。 俺的设备： MacBook Pro M1 Max / 64GB Unified Memory / 24-core GPU Ollama 同一个 prompt下的表现对比： Qwen3.5-9B（图1） Prompt eval: 127.52 tok/s Generation: 34.21 tok/s 2,011 tokens Total: 59.3s Qwen3.6-27B（图2） Prompt eval: 40.13 tok/s Generation: 9.98 tok/s 6,103 tokens Total: 10m 13s 测试的结果不支持最初的理想情况。 27B 确实可以非常轻松地装进 64GB unified memory，但 generation speed 只有 9B 的约 29%。 而且这次 27B 默认生成了大量 reasoning tokens。 于是「更大的模型 + 更慢的 tok/s + 更长的 reasoning」叠加起来，一个相同的任务，从约 1 分钟变成了 10 分钟。 期待未来的 1B–7B，能做到今天 20B、30B 模型的大部分事情。 但目前差距还有点远。

节假日我是不会出去玩的 原因很简单——花 3 倍的钱享受 1/2 的体验 举个例子： 吉域合是阿勒泰最好的酒店，这两天去哪儿的价格是 1242 然而，假期一过就会降到 397 btw，我之前淡季去住的时候 350 一晚，人少给升房 整个酒店就两三个人去滑雪，所以免费接送，公务车 + 2 人服务

$ETH exit queue has hit 773,447, the highest level in 2026. Investors are now rushing to unstake their Ethereum.

其实我觉得还是有一定必要的，很多事情想清楚了，会减少自己很多焦虑。如果你没有这些焦虑，那你确实不需要向内求

Considering what ChatGPT is doing... Just shipped Cockpit v0.4 ✨ Claude Code joins Codex and Hermes in a local, read-only dashboard. Browse saved conversations, project instructions, memory, and files—all in one place. It's for Developer preview. Claude Code users: I’d love your feedback! github.com/swen-chan/cockpit…

Hi, Tomorrow we are re-opening the Pro $200 subscriptions to new subscribers, but together with it we are also changing how we calculate the usage for it. In effect, if you do the math, it will net out at half the dollar in API spend compared to the old Pro $200 plan. Now that it's said, let me explain why this is happening and why you will still get more work done than if you were on the Pro $200 subscription one month ago. (a) We didn't want to compromise in other ways and are committing to not reintroducing the 5h limit, so that you can fully use the weekly usage when you want. (b) On the subscription, we guarantee that over time you always get more work done and with an increasing level of quality. This means that you will continue to get more value per dollar spent as a result of models getting more efficient and us passing down the improvements in the form of API price reductions. (c) We don't want to put an incentive on ourselves to artificially inflate the API list prices to make it look like you are getting a lot (and workaround it through discounts, etc). Instead we want to continue to both rapidly reduce prices and increase capabilities of models on the API. This week we introduced GPT-6 Sol and GPT-6 Luna at 50% of their previous price. Over time, we see prices go low enough that it makes sense for most to buy usage as needed without there being a significant gap between what you get in a subscription and what you get in the API for a dollar spent. (d) Tomorrow, we are adding more things to the subscription that won't draw on the usage, I won't reveal what that is yet. I wanted to be transparent before all the big announcements tomorrow. Lots of new exciting things are coming to the subscriptions that will make it super compelling, but I wanted to make sure to share this change ahead of time so you can all understand it before we shower you with good news. Codexingly, Tibo

一个蛮有意思的量化因子，叫做量价结构变化反转因子（Price-Volume Structure Shift Reversal Factor）（下面会放测试图） 逻辑也很直觉：当价格与成交量之间的关系（例如高价与成交量的相关性）在短时间内出现明显变化时，往往代表市场内部的资金行为正在发生转折。原本稳定的量价结构一旦被打破，背后通常意味着主导市场的力量正在切换 同时，我们再引入价格本身的波动性（例如过去一段时间收盘价的标准差）进行排序，去衡量当前市场的不稳定程度。当价格波动较高、且量价结构正在快速改变时，市场更容易出现情绪主导的错误定价 因子公式如下： (-1*(DELTA(CORR(HIGH,VOLUME,5),5)*RANK(STD(CLOSE,20)))) 本质上是在捕捉市场在「量价结构突变」与「波动放大」共振下的非理性行为，以及由此带来的短期反转机会 两行源代码跑完整回测，大家可以去测试看看 github.com/eliasswu/Alphapur…