Saw a concern on social media today. Where is the demand for AI? It’s both right and wrong. Demand hasn’t magically appeared out of thin air, but suppressed demand is being unleashed. I’ve had a realization lately: AI creation—coding, image generation, video generation—is like the new liberal arts of our era. Technique Is No Longer the Barrier Donald Knuth’s famous The Art…
NBot took first place on Product Hunt’s daily board. It’s a product we spent two months building and launched over the Christmas holidays. The timing was partly because of delays and partly because competition is lighter during the break. We’re happy with the result, but the whole process gave me a much clearer view of growth for AI products. Product Hunt Isn’t as Easy as It Looks Posting on PH is…
Life has been busy and I barely had energy to watch AI updates. I wrote half of last week’s (12.05) notes and never finished, so here comes a catch-up after two weeks away. Last week saw a bunch of new models or reports: DeepSeek, Mistral, QwenVL. I skimmed DeepSeek-3.2’s report and nothing really grabbed me; Mistral feels forced, with the only bragging point being a programming win…
It’s the first day of Thanksgiving break—thanks for reading. Last week’s vibe coding interview piece became my most-read article ever, thanks to Ruanyifeng for featuring it in his tech weekly. This week I want to talk about model selection for AI apps at the end of 2025. It’s critical: it drives both quality and cost. Don’t shrug off the difference between $0.20 and $0.25 per million tokens—that’s…
Cover created with Nano Banana Our team’s interview format just changed. Leetcode puzzles are out; building a full feature is in. If you don’t code with AI—and your typing speed isn’t absurd—you won’t finish. It sounds like a tooling upgrade, but it’s actually reshaping what we mean by an “excellent engineer.” The Gap Gets Wider In classic algorithm interviews, the difference between failing and…
On my commute I listened to a podcast episode featuring Songlin Yang—nicknamed the mother of linear attention—where she patiently explained the mechanics behind the architecture. The structure is red-hot right now, especially because several frontline Chinese model labs are doubling down on it. I figured it was worth retelling the story in my own words. How Is a Token Generated? To keep things…
This week didn’t revolve around a single theme. It wasn’t a neat follow-up to last week’s “product fundamentals,” nor a deep dive into “triple-speed coding” like 10.19. It felt more like a handful of scattered observations and experiments, each sparking new thoughts. If there’s a shared thread, it’s probably this: in an era where AI rewires productivity, what actually matters? Beautiful view…
This week I listened to LatePost Chat’s episode “Agents Are Opportunities—and So Are Agent-Building Tools | Starting From OpenAI DevDay.” Henry, the guest, mapped out six major inflection points in the evolution of agent tooling: from ChatGPT’s launch at the end of 2022 to the recent wave of “computer use.” Every leap in model capability exposes a new “body part” that needs strengthening, and each…
This week a teammate and I ran a deep experiment: we handed almost the entire coding workload of a new project to AI. The outcome was stunning—we wrapped up in a single week what normally takes us three. Yet behind that unprecedented velocity were a few counterintuitive truths that quickly surfaced. The Cost of Efficiency and Cognition on Overdrive When coding hits three times its usual speed, the…
I tried a different workflow this week: let the AI interview me first, then spin my answers into a story. Two big moments stood out—catching the vibe coding wave with Claude Code, and feeling the AI heat in San Francisco during Tech Week. Vibe Coding Sends Productivity into Orbit Trying vibe coding wasn’t a spontaneous whim. I’ve long used Copilot mainly for inline completion, occasionally…
Went to the Silicon Valley 101 Alignment conference today. After being a listener for several years, I finally saw Jane and Qian in person. Did not expect the scale to be this big—absolutely packed, and not just Chinese attendees; plenty of international friends too. As Xiaodi Hou said: Silicon Valley really is the promised land. The collective hunger for new tech and new chances is just awesome.…
ACL 2025年5月的ARR cycle已经出分了,估计这会儿学术圈又忙起来了,作者在忙着写rebuttal,AC在忙着找紧急审稿。而我这个审稿人却感觉格外的轻松,因为这次看的论文分都太低了,估计作者都不会来rebuttal,自然我也不用再回复作者。 这一次给我分的都是RAG和Hallucination方向的论文,我一共审了5篇,打分情况是3篇2分,2篇2.5分。说实话,总体上我已经抬一手了,全凭良心打的话,还要更低一点。 给不熟悉ARR评分标准的读者稍微补充一下,ARR是5分制,分以下9档: 5 = Consider for Award: I think this paper could be considered for an outstanding paper award at an *ACL conference (up to top 2.5% papers). 4.5 =…
The scores for ACL 2025’s May ARR cycle have been released, and I imagine the academic community is busy again—authors scrambling to write rebuttals, area chairs frantically searching for emergency reviewers. Yet as a reviewer, I feel exceptionally relaxed because the papers I reviewed scored so low that I doubt the authors will even bother with rebuttals, which means I won’t need to…
The recent “Dawn of the East” hype around Manus.im sparked significant attention. However, its evolution over the past few days has been dramatic, with numerous open-source projects attempting to replicate it. OpenManus, one of the earliest clones, claims to have replicated it in just 3 hours, though its parent team MetaGPT has been working on agents for much longer. I believe both…
On March 1st, 2025, the sixth day of DeepSeek’s open-source week, they released something quite special - an introduction to the V3/R1 model’s inference system. Unlike the code repositories from previous days, this time it was a blog post. I haven’t read the article in detail, and I might not fully understand it anyway, but the Twitter post alone has already caused quite a stir.…
每个人的世界无非是ta看到的,听到的,闻到的,尝到的,摸到的加上想到的。 今天Sora的问世让我感觉到离计算机能够模拟这个感官世界已经不远了。虽然不是精确的,但肯定可以是精彩的,令人满意的。在享受面前,谁又会在意精不精确呢。 2022年ChatGPT出来的时候有模糊的感觉:预测下一个token的任务竟然能在效果上模拟推理、逻辑、甚至扮演角色,这不就说明文字世界的一切其实是被一个概率刻画的规律主宰的吗。放在玄学语境里,就是所谓的因果,它真的存在。 这次OpenAI更近一步,“世界模型”的概念已经再清楚不过地表达他们已经接近找到用概率模型刻画世界的方法了。 Sora is able to generate complex scenes with multiple characters, specific types of motion, and accurate details of the…
作为一名Prompt Engineer,每天都在跟GPT打交道,时常被他惊艳,也看过很多模型失效的案例。在不精调的情况下,prompt基本上是影响效果的唯一因素了,虽然网上有很多Prompt编写指南,但我认为OpenAI出品的这份,你一定要看一下。 这篇文章就给大家划一下重点。 ChatGPT基操 主要包含在How to work with large language models这个文档里,同时适合网页和API用户。首先,介绍了向ChatGPT提问的三种主要范式,一种是直接给指令,例如 Extract the name of the author from the quotation below. “Some humans theorize that intelligent species go extinct before they can expand into outer…
Logistic Regression is a statistical method used to analyze the relationship between a categorical dependent variable and one or more independent variables. It is widely used in machine learning and predictive modeling for binary classification problems. In this article, we will discuss the basics of logistic regression and its mathematical formulation. Binary Classification Binary classification…
Linear regression is a widely-used statistical method for modeling and predicting the relationship between two variables. In essence, it is a technique for finding the best-fitting line through a set of data points. This article provides a beginner-friendly introduction to linear regression and its underlying concepts. What is Linear Regression? Linear regression is a statistical method that…
可能是网站slogan:AI算命,朋克养生的关系,最近我看google search console后台总是有很多"Chatgpt算命"这个查询导致的网站展现。 既然大家有这个需求,我就实践SEO里面的方法,上网搜索了一下如何使用chatgpt算命,总结在这个文章里。 好多chatgpt算命的查询 {: .align-caption style=“text-align:center;font-size:smaller”} 基本上,这要用到chatgpt的**“角色扮演”**功能,让它站在算命者的角度回答你的问题。 先从英文开始,英文里算命先生叫做fortune teller,那么我们可以这么问chatgpt Act like a coffee fortune teller, I can see a cat shape in the…