AI
Claude、Anthropic、编程智能体
English interviews about AI itself, in Chinese — AI-dubbed in each speaker's own voice, faithful translation only. Models, agents, and the people building with them.
Unofficial fan channel · not affiliated with Hyperliquid · AI-dubbed, faithful translation.
Curated. Original-voice Chinese dubs.
26 episodes
Peter Steinberger: What Happens When 4.7 Million People Let It Cook
Last November, Peter Steinberger was annoyed that there was no good way to talk to his coding agents from his phone, so he built one himself. A few months later, OpenClaw had exploded into one of the biggest open source AI projects in the world, with nearly 3,000 contributors and a peak of 4.7 million weekly downloads.
Sam Altman - How to Start a Startup
Sam Alman, co-founder of OpenAI, discusses how the landscape for startups has dramatically shifted over the past decade, with AI's rapid advancements and reduced barriers to entry making it easier for small teams to achieve what once took larger companies much longer. He emphasizes the importance of embracing exponential growth and being prepared to tackle increasingly complex challenges as part of building successful startups in today’s rapidly changing environment.
大模型参数竞赛背后的产业逻辑
文章分析了Kimi、Qwen等大模型在参数规模上的竞争及OpenAI、Anthropic的战略分化，指出最强模型的护城河正在贬值，核心壁垒转向业务闭环与生态系统。同时，强调开发者不应依赖单一基座模型，而应关注实际工作流数据的价值。
Opening Keynote, Michael Truell | Compile 26
Cursor, an AI company founded in January 2022 by four young programmers, started as a project to build a development environment tool that the team would genuinely find useful, leading to the creation of their open-source code editor Cursor. Since its initial release early 2023, Cursor has evolved significantly, focusing on becoming simpler and more powerful, particularly in working with agents, reflecting a fundamental shift in its development approach over the past year and a half.
rody - Anthropic's Claude Code engineer Sid Bidasaria: "Stop babysitting y...
This interview discusses strategies to improve efficiency in managing AI agents like Claude, focusing on creating high-quality ClaudeMD files, connecting tools to enhance context, and setting up a remote environment for better reliability. The talk aims to help users transition from babysitting their agents to leveraging them more effectively by rethinking traditional software tooling designed for human use.
Codez - Anthropic Managed Agents team: "Fable 5 is our best model for runnin...
Welcome to Code with Claude, where Jess from product and Michael from engineering share insights on the exponential growth in AI capabilities and how it's transforming developer workflows. They discuss the challenges developers face as models become more advanced, such as managing context, scaling infrastructure securely, and ensuring observability, leading to the introduction of Cloud Managed Agents designed to handle these complexities out-of-the-box.
You are using Claude Fable 5 wrong
This interview transcript discusses the practical applications of Anthropic’s Fable 5 AI model for making money and increasing productivity, covering examples from video editing to content creation, and emphasizes the importance of leveraging the AI's capabilities with strategic inputs and workflows. The host aims to provide viewers with actionable insights on how to use Fable 5 effectively for various business and personal projects.
Running 128 Coding Agents at Once
Sam Whitmore from Curser's cloud agents team discusses his experience running multiple AI agents for tasks such as KB cache compaction and how he delegates work among them using messaging scripts. Harry from Base 10 shares the company’s journey from open-source model research to specializing in post-training inference, emphasizing the importance of coupling real-world feedback with training to enhance performance.
How Anthropic uses Claude in GTM Engineering
This interview transcript showcases how Jared, a startup account executive at Anthropic, leveraged Claude, Anthropic’s AI assistant, to develop Class—a tool for generating customer emails and managing inboxes more efficiently, significantly reducing his workload and enabling him to focus on higher-level product design. By integrating Claude into his workflow, Jared was able to save several hours daily and transition into a go-to-market architect role, transforming his career trajectory.
The CEO Must Be the Chief AI Officer
Pedro Franchesci, co-founder and CEO of Brex, discusses how his company has integrated AI deeply into its operations, drawing parallels between the early days of electricity and the current landscape of AI. Pedro shares his journey from initial experimentation with GPT-3 during the pandemic to fully embracing agentic loops with tools, emphasizing that good AI products are essentially about empowering human agents rather than replacing them.
Training Composer 2
Sasha, a researcher at Cursor, introduces Composer 2, their in-house coding model designed for agentic coding. With performance comparable to Opus 4.6 and GPT 5.4, Composer 2 offers strong coding speed and affordability, making it efficient for various coding tasks.
Boris Cherny: Claude Code & the Future of Engineering | Acquired Unplugged presented by WorkOS
Claude code, a coding agent developed by Anthropic, emerged from an internal prototyping team tasked with exploring future AI product possibilities. Initially struggling with basic coding tasks, the project leveraged Anthropic’s focus on AI safety research and the clear, constrained nature of coding to iteratively improve its capabilities.
Your CLAUDE.md Is a Vault With No Lock
Connor Bronden, host of Chain of Thought, interviews Tyler Aikido, CTO of Red Panda, about the challenges in managing AI agents in enterprise environments. Aikido argues that current governance layers are inadequate for handling the unpredictable and technically advanced nature of AI agents, highlighting issues such as hallucination, prompt injection, and rapid execution of destructive tasks, necessitating new approaches to HR and management in the age of autonomous digital employees.
Venice vs. OpenAI and Anthropic: The Fight for Private AI
John and Jesse from Venice, the head of strategy and CTO respectively, explain that Venice aims to become a mass-market consumer app where privacy is a core feature, attracting users who may not be crypto enthusiasts. They discuss the importance of private AI to protect personal data from potential misuse by large tech companies, emphasizing that Venice offers an alternative for individual privacy in the age of advanced AI models.
The FDE Playbook for AI Startups with Bob McGrew
Bob McGrue, an early engineer at PayPal and recent chief research officer at OpenAI, discusses the rise of the Forward Deployed Engineer (FDE) model in AI startups. This model involves sending technical engineers to customer sites to tailor products to specific needs, driving product discovery and value creation, rather than scaling a one-size-fits-all solution. McGrue explains how this approach emerged from early efforts at Palantee, emphasizing its relevance in the current landscape of AI agent companies.
How To Build A Company With AI From The Ground Up
In this interview, YC partner Diana discusses how AI is transforming startup operations by enabling entirely new capabilities that can replace entire teams or previously impossible feats. She argues that founders should view AI not just as a productivity tool but as the operating system for their company, advocating for closed-loop systems where every process is captured and improved by intelligent agents.
How are coding agents changing software engineering?
This interview transcript discusses Cursor's latest developer habits report, which highlights trends such as accelerated coding speed, larger pull requests, and the increasing use of AI-generated code. The report also examines how top power users are more productive, with context becoming more important in interactions, and models improving in their ability to call tools, reflecting overall advancements in model intelligence.
How to Build a Self-Improving Company with AI
This interview transcript explores how companies can be reimagined using artificial intelligence (AI) as a core component rather than an add-on tool. The speaker discusses the concept of transforming traditional hierarchical organizational structures into AI-powered systems that self-improve continuously, highlighting examples such as intelligent agents and automated loops for product optimization and customer service improvements.
Andrej Karpathy: From Vibe Coding to Agentic Engineering
Andrej Karpathy, co-founder of OpenAI and former head of AI at Tesla, shares his insights on the rapid evolution of language models like LLMs, describing them as a new computing paradigm that shifts programming from explicit rules (software 1.0) to prompting context for agentic responses (software 3.0). He discusses how this shift is transforming development practices, citing examples such as automated OpenClaw installation and image-based menu descriptions generated by Gemini, highlighting the profound impact on traditional software engineering approaches.
How to Build an Internal AI Agent That Evolves Itself
Ayush, founder of Answer This, details how his two-person company achieved $2M ARR by leveraging a self-extending internal AI operations agent. This agent automates high-volume daily tasks like emails and customer support, and critically, can autonomously code new tools for itself and update its own behavioral instructions based on feedback, effectively evolving its capabilities by learning from the company's database and codebase.
Simon Eskildsen on scaling Shopify, building turbopuffer, and the future of databases
The interview discusses the scaling challenges faced by Shopify from 2010 to 2020, particularly with MySQL databases. One humorous incident involved an unexpected PHP cron job running every hour as part of Percona utilities, causing a soft lock-up due to its dependency on PHP, which was continuously tracking open files and sessions. The team spent considerable time optimizing systems like Memcached, Redis, and MySQL by reducing connection counts and employing various tricks since there were no open-source proxies available at the time. The infrastructure team evolved from a traditional operations-focused structure into one capable of handling multiple severe incidents (SEVs) during flash sales. Flash sales occurred when high-profile influencers promoted new products on platforms like Instagram or Facebook, leading to sudden surges in traffic and inventory contention that could overwhelm systems. Shopify had to manage these by ensuring robust inventory reservation processes and preparing for the influx of demand from millions of users trying to purchase limited stock simultaneously.
Introducing Cursor 3
The interview highlights advancements in coding that are narrowing the gap between imagination and reality through the improvement of coding agents' capabilities, making them smarter and faster. This is being celebrated with the announcement of Cursor 3, which serves as a platform for these agents, combining professional development environments, language servers, files, remote SSH, and the ability to run multiple agents locally or in the cloud. The new system enables users to interact with agents directly while building software, allowing them to test changes and potentially automate parts of the process by moving projects onto cloud-based agents that can work independently.
Head of Gemini: You're Using 5% of What Gemini Can Actually Do | Josh Woodward
In the interview, Josh Woodward discusses the launch of Gemini Spark, an AI agent created by Google that operates 24/7 in the background to automate tasks for users, freeing up time for personal activities. This marks a significant shift towards an "agentic era" in which products increasingly handle tasks autonomously. The Gemini Omni model represents another major change as it transforms how input is processed and output generated. Woodward also mentions that Google IO will feature additional science-based products under development. A standout update highlighted by the interviewer was the voice-to-email capability, which gathers necessary information from various sources like Google Drive and Gmail to create a comprehensive email with tables. The Gemini Spark tool, still in testing phases, is expected to roll out soon to selected users, offering seamless integration within the Google ecosystem for tasks such as document creation, video making, image generation, and song composition.
Anthropic's Boris Cherny: Why Coding Is Solved, and What Comes Next
Boris Cherny, creator of "Claude code," discussed its accidental origins in late 2024 at Anthropic Labs, aiming to transcend basic code completion by enabling an agent to write entire codebases. Though initially struggling, Claude code experienced exponential growth starting with Opus 4, and is now credited with reinventing modern software development.
Tips for building AI agents
Anthropic’s Barry Zhang (Applied AI), Erik Schluntz (Research), and Alex Albert (Claude Relations) discuss the potential of AI agents, common pitfalls to avoid, and how to prepare for the evolving landscape. Read more advice on building agents: 00:00 Introduction 00:26 Defining AI agents and workflows 02:55 Anatomy of an agent prompt 04:29 Behind the scenes stories 07:29 Why write about agents now 08:53 Overhyped and underhyped aspects of agents 09:57 Identifying useful applications of agents ...
A conversation on Claude Code
A detailed conversation about Claude Code, exploring its capabilities and applications in software development.
Request any English interview in Chinese — $1/episode launch price (≤60 min), live within 24h, requester credited. Listening is free for everyone.
how it works →