AI Explorer
Ruohang Feng @Vonng / Agent, LLM, Database
Everything about AI, agents, LLMs, vibe coding, AI4DB and DB4AI.
Pedaling the Codex/Claude Bike Until the Wheels Smoke
Claude Fable and GPT-5.6 have arrived back to back, followed by repeated quota resets. How do you turn a fleeting Coding Plan windfall into real output? Here is my playbook for model specialization, …
The Bojie Li–DeepSeek Interview Controversy
Today's biggest tech drama began when Bojie Li criticized DeepSeek's interview process on Twitter: he was asked to solve online-judge coding problems, then suspected of cheating. Add the DeepSeek …
Fifty Years of Love and War: File Systems, Databases, and the Agent-Era Storage Endgame
File systems and databases have spent fifty years fighting and borrowing from each other. The agent era may seem to put file systems back on top, but the real winner may be databases that learn to …
The Coding Plan Window Is Closing—Use It While It Lasts
Coding Plans remain one of the best opportunities in the AI era: a subscription can unlock compute worth many times its price, but that window is already narrowing.
The Karma of Open Source: When Code Is Worthless, Where Does Trust Come From?
AI is driving the cost of producing code toward zero. It cannot compress time, track records, or accountability. The real value of open source is not yesterday's code, but a system trusted to deliver …
The Cognitive Price Revolution: AI's Impact on the Economy and the Future
AI's economic significance is not merely that it has become smarter, but that the price of average cognition is collapsing. A productivity explosion is almost certain; whether it becomes broad …
The Cerebellum: The Other Half of Intelligence—and the Strongest AI Hasn't Touched It
The cerebellum changed how I see AI's frontier: LLMs have already absorbed humanity's explicit knowledge and are beginning to acquire interventional data through agentic RL. What they still lack is a …
Claude Fable First Impressions: The Pendulum Swings Back
Claude Fable is genuinely insightful and deserves to be called the new SOTA. But its high price, dynamic downgrades, limited subscription window, and mandatory data retention badly undermine the …
Cancel Claude, Switch to Codex
When my $200 Claude Code Max subscription expired, I canceled it and moved my primary workflow to Codex. The only way to know which one is better is to put it to work.
I Asked AI to Prove That Eating Garlic Prevents Middle Ear Infections
I asked AI to find real papers supporting the absurd claim that eating garlic prevents middle ear infections. The result shows how AI is degrading citations as a credibility signal at scale.
AI Is Bringing Down the Scaffolding of Trust
The most dangerous change in the AI era is not that machines can write articles, draw images, or generate video. It is that content itself is losing its standing as evidence.
Give DBA Agents a Body
Today's models are smart enough. What they lack is a body: a deterministic runtime that is observable, controllable, and reversible. Pigsty is evolving from a PostgreSQL distribution into an Agent …
A Busy Few Days in Infrastructure and AI
Ubuntu 26.04 LTS, Qwen3.6-27B, ChatGPT Images 2.0, Privacy Filter, Anthropic Mythos, and the Claude Pro controversy all landed in quick succession. Here are the infrastructure and AI developments …
The Three-Way Endgame: Why Agent Memory Frameworks Are a Dead End
A cold shower for the red-hot agent memory market—not because agents do not need memory, but because memory is the endgame moat. That moat belongs to models, harnesses, and databases, not today's …
Two Hemispheres: Transformer, Diffusion, and the Boundary of Intelligence
A dialogue about intelligence: its foundations may lie not in language or vision, but in two irreducible computational paradigms—symbolic reasoning and field-state intuition—kept distinct and mutually …
Cyber Dharma: A New Engineering Answer to Ancient Questions
A project manifesto: why build Cyber Dharma, and what it is not.
Burning Hundreds of Millions of Tokens a Day. Then What?
Once token burn turns from usage exhaust into a KPI and leaderboard, it quickly mutates into theater. Don't post fuel burn. Post where you got to.
Why PostgreSQL Won in the AI Era
Boring technology won the wildest era. A look at extensibility, agent choice, database cloning, and the future of the DBA.
AGI Is Here. Do You Have a Ticket?
When the strongest AI is not expensive but simply unavailable, the world starts converging on digital feudalism. And the window to act is narrowing.
Can You Distill an Expert?
Polanyi's tacit knowledge explains the 70% ceiling of AI agents: real intuition, feel, and judgment do not serialize cleanly. They grow, if at all, through practice.
Yes, I Use AI to Write
AI is a multiplier. It amplifies depth and mediocrity alike. In an age where answers are cheap, questions are the real currency. There is nothing to hide about writing with AI.
Local AI's Inflection Point: 2027
When subsidies fade, hardware catches up, and open models mature, all three lines cross in 2027. "Build your own AI" goes from idea to reality.
LLMs Have Emotions: Claude's Internals Reveal Steerable 'Emotion Vectors'
Anthropic's new research gives us the first direct look at causally steerable "emotion vectors" inside a large language model. That should change how we think about AI.
Good News: Claude Code Got "Open-Sourced" Yet Again
SOTA coding agent Claude Code leaked its source again, after falling into the same hole twice. The whole codebase is out in public. Performance art at its finest.
Pedaling the Codex/Claude Bike Until the Wheels Smoke
Folks, I haven’t had time to write these past couple of days: I’ve been too busy pedaling the AI bike. Claude Fable and GPT-5.6 arrived one after the other, and both companies generously threw in several rounds of quota resets. Suddenly I was …
Folks, I haven’t had time to write these past couple of days: I’ve been too busy pedaling the AI bike. Claude Fable and GPT-5.6 arrived one after the other, and both companies generously threw in several rounds of quota resets. Suddenly I was …
The Bojie Li–DeepSeek Interview Controversy
Today’s biggest tech drama began when Bojie Li criticized DeepSeek’s interview process on Twitter. In short, he went in for an interview, was asked to solve online-judge coding problems, and was then suspected of cheating. Add the DeepSeek name, and …
Today’s biggest tech drama began when Bojie Li criticized DeepSeek’s interview process on Twitter. In short, he went in for an interview, was asked to solve online-judge coding problems, and was then suspected of cheating. Add the DeepSeek name, and …
Fifty Years of Love and War: File Systems, Databases, and the Agent-Era Storage Endgame
Prologue: Several Database People Independently Built File Systems Something curious has happened in agent infrastructure over the past six months: several veteran database people have started building “file systems for AI agents.” Timescale’s …
Prologue: Several Database People Independently Built File Systems Something curious has happened in agent infrastructure over the past six months: several veteran database people have started building “file systems for AI agents.” Timescale’s …
The Coding Plan Window Is Closing—Use It While It Lasts
One of the biggest windfalls of the AI era is the Coding Plan. A few months ago, I applied to OpenAI’s Codex program for open-source developers. A few days ago, I was finally approved: six free months of ChatGPT Pro on a new account. The timing could …
One of the biggest windfalls of the AI era is the Coding Plan. A few months ago, I applied to OpenAI’s Codex program for open-source developers. A few days ago, I was finally approved: six free months of ChatGPT Pro on a new account. The timing could …
The Karma of Open Source: When Code Is Worthless, Where Does Trust Come From?
Credit does not require a soul. It requires an account. Introduction: The Overnight Rewrite Not long ago, someone used AI to rewrite an open-source Python library in Rust overnight. The new project had no fork history, no evidence of copied code, …
Credit does not require a soul. It requires an account. Introduction: The Overnight Rewrite Not long ago, someone used AI to rewrite an open-source Python library in Rust overnight. The new project had no fork history, no evidence of copied code, …
The Cognitive Price Revolution: AI's Impact on the Economy and the Future
Introduction: Reframe the Question “Will AI cause unemployment?” is a question guaranteed to produce garbage answers, because it comes with only two canned scripts. Optimists recite history: from the power loom to the ATM, every technological panic …
Introduction: Reframe the Question “Will AI cause unemployment?” is a question guaranteed to produce garbage answers, because it comes with only two canned scripts. Optimists recite history: from the power loom to the ATM, every technological panic …
The Cerebellum: The Other Half of Intelligence—and the Strongest AI Hasn't Touched It
Let me start with an uncomfortable number. The cerebral cortex—the part we usually point to as evidence that “I am thinking”—contains roughly 15 billion neurons. The cerebellum, tucked beneath the back of the brain and rarely given much thought, …
Let me start with an uncomfortable number. The cerebral cortex—the part we usually point to as evidence that “I am thinking”—contains roughly 15 billion neurons. The cerebellum, tucked beneath the back of the brain and rarely given much thought, …
Claude Fable First Impressions: The Pendulum Swings Back
This morning, Anthropic officially released its new Claude Fable model—the consumer-grade, nerfed version of the much-rumored Mythos. Since it was billed as an “AGI-level model,” I naturally tried it right away to see what it could actually do. The …
This morning, Anthropic officially released its new Claude Fable model—the consumer-grade, nerfed version of the much-rumored Mythos. Since it was billed as an “AGI-level model,” I naturally tried it right away to see what it could actually do. The …
Cancel Claude, Switch to Codex
I have written several pieces about Claude Code before. Over the past few months, though, I have used it less and less. Codex is now my primary tool. A few days ago, my $200 Claude Code Max subscription came up for renewal, so I canceled it and kept …
I have written several pieces about Claude Code before. Over the past few months, though, I have used it less and less. Codex is now my primary tool. A few days ago, my $200 Claude Code Max subscription came up for renewal, so I canceled it and kept …
I Asked AI to Prove That Eating Garlic Prevents Middle Ear Infections
Last month, my old friend Ma roasted me in our group chat for using ChatGPT every day to churn out grand-sounding, LinkedIn-style aphorisms with no basis in experience. He said, “You can make up any theory you want—say, that eating garlic reduces the …
Last month, my old friend Ma roasted me in our group chat for using ChatGPT every day to churn out grand-sounding, LinkedIn-style aphorisms with no basis in experience. He said, “You can make up any theory you want—say, that eating garlic reduces the …
AI Is Bringing Down the Scaffolding of Trust
The most dangerous change in the AI era is not that machines can write articles, draw images, or generate video. The real danger is that content itself is losing its standing as evidence. For a long time, people assumed that media artifacts carried …
The most dangerous change in the AI era is not that machines can write articles, draw images, or generate video. The real danger is that content itself is losing its standing as evidence. For a long time, people assumed that media artifacts carried …
Give DBA Agents a Body
HOW 2026 keynote: Give DBA Agents a Body Part I: Opening—An Absurd Phenomenon 1. Cover Hello, everyone. I’m Ruohang Feng, the organizer of this tools track. This is a PostgreSQL tools session, but today I don’t want to talk about how many new …
HOW 2026 keynote: Give DBA Agents a Body Part I: Opening—An Absurd Phenomenon 1. Cover Hello, everyone. I’m Ruohang Feng, the organizer of this tools track. This is a PostgreSQL tools session, but today I don’t want to talk about how many new …
A Busy Few Days in Infrastructure and AI
The past couple of days have been unusually busy for the infrastructure crowd. Ubuntu released its biennial LTS today. Alibaba’s Qwen team dropped a 27B dense model yesterday that beats its own previous 397B MoE flagship. The day before that, OpenAI …
The past couple of days have been unusually busy for the infrastructure crowd. Ubuntu released its biennial LTS today. Alibaba’s Qwen team dropped a 27B dense model yesterday that beats its own previous 397B MoE flagship. The day before that, OpenAI …
The Three-Way Endgame: Why Agent Memory Frameworks Are a Dead End
A few months ago, I wrote The OS Moment for AI Agents. I made a prediction there: the next frenzy in agent infrastructure would be memory. Startups and open-source projects would swarm around the question of how agents should remember things. Capital …
A few months ago, I wrote The OS Moment for AI Agents. I made a prediction there: the next frenzy in agent infrastructure would be memory. Startups and open-source projects would swarm around the question of how agents should remember things. Capital …
Two Hemispheres: Transformer, Diffusion, and the Boundary of Intelligence
A friend recently wrote an essay titled “Vision Is the Foundation of the World, Not a Plugin for Language.” It prompted a conversation with Claude, which I have edited into the Socratic dialogue below. Language as the Foundation, Vision as an Add-On? …
A friend recently wrote an essay titled “Vision Is the Foundation of the World, Not a Plugin for Language.” It prompted a conversation with Claude, which I have edited into the Socratic dialogue below. Language as the Foundation, Vision as an Add-On? …
Cyber Dharma: A New Engineering Answer to Ancient Questions
Have you ever noticed that writing a System Prompt for AI, something as simple as “You are Claude, a helpful AI assistant” is structurally the same act as the line in Genesis, “Let there be light”? Both use language to bring an entity into being. …
Have you ever noticed that writing a System Prompt for AI, something as simple as “You are Claude, a helpful AI assistant” is structurally the same act as the line in Genesis, “Let there be light”? Both use language to bring an entity into being. …
Burning Hundreds of Millions of Tokens a Day. Then What?
My friend Jiang recently got obsessed with flexing in group chat about burning hundreds of millions of tokens a day. What was he doing with them? One minute it was some ontology database, the next he had an agent rebuilding Pigsty in Go under the …
My friend Jiang recently got obsessed with flexing in group chat about burning hundreds of millions of tokens a day. What was he doing with them? One minute it was some ontology database, the next he had an agent rebuilding Pigsty in Go under the …
Why PostgreSQL Won in the AI Era
I recently did a recorded livestream on PostgreSQL and AI. The host asked a lot of good questions, and I gave some answers. The recording goes out Tuesday night. I am publishing the fuller written version of my views here, including quite a bit I did …
I recently did a recorded livestream on PostgreSQL and AI. The host asked a lot of good questions, and I gave some answers. The recording goes out Tuesday night. I am publishing the fuller written version of my views here, including quite a bit I did …
AGI Is Here. Do You Have a Ticket?
On April 8, 2026, Marc Andreessen posted a tweet that, translated loosely, said: The latest AGI pricing is out: if you are one of 11 specific companies, the price is negative $9 million. Otherwise, the price is infinity. Negative $9 million means not …
On April 8, 2026, Marc Andreessen posted a tweet that, translated loosely, said: The latest AGI pricing is out: if you are one of 11 specific companies, the price is negative $9 million. Otherwise, the price is infinity. Negative $9 million means not …
Can You Distill an Expert?
Polanyi’s tacit knowledge explains the 70% ceiling of AI agents: real intuition, feel, and judgment do not serialize cleanly. They grow, if at all, through practice. 1. Distilling Employees A popular idea lately is to “distill” employee knowledge …
Polanyi’s tacit knowledge explains the 70% ceiling of AI agents: real intuition, feel, and judgment do not serialize cleanly. They grow, if at all, through practice. 1. Distilling Employees A popular idea lately is to “distill” employee knowledge …
Yes, I Use AI to Write
People often leave the same short comment under my posts: “AI wrote this.” Correct. I use AI, and I use it heavily. I do not think there is anything to hide there. But the topic itself is worth addressing seriously at least once: how should a person …
People often leave the same short comment under my posts: “AI wrote this.” Correct. I use AI, and I use it heavily. I do not think there is anything to hide there. But the topic itself is worth addressing seriously at least once: how should a person …
Local AI's Inflection Point: 2027
When subsidies fade, hardware catches up, and open models mature, all three lines cross in 2027. “Build your own AI” goes from idea to reality. A Thought Triggered by a Group Chat A friend in a group chat said the other day that “self-hosting” is …
When subsidies fade, hardware catches up, and open models mature, all three lines cross in 2027. “Build your own AI” goes from idea to reality. A Thought Triggered by a Group Chat A friend in a group chat said the other day that “self-hosting” is …
LLMs Have Emotions: Claude's Internals Reveal Steerable 'Emotion Vectors'
A couple of days ago Anthropic published a blog post with a very calm title: “Emotion concepts and their function in a large language model”. The content was anything but calm. They found “emotion vectors” inside Claude’s neural network, and those …
A couple of days ago Anthropic published a blog post with a very calm title: “Emotion concepts and their function in a large language model”. The content was anything but calm. They found “emotion vectors” inside Claude’s neural network, and those …
Good News: Claude Code Got "Open-Sourced" Yet Again
Good news: Anthropic’s latest flagship coding agent, Claude Code, just had its entire source tree dumped in public. The GitHub repo already has over a thousand stars, 4,700-plus source files, and more than half a million lines of code, all for free. …
Good news: Anthropic’s latest flagship coding agent, Claude Code, just had its entire source tree dumped in public. The GitHub repo already has over a thousand stars, 4,700-plus source files, and more than half a million lines of code, all for free. …