Jeff Bean

@JasonKPargin Faces of Death, Red Asphalt… this was the stuff of dread and lore. Now you pop up your newsfeed on your phone to see a deadly runway accident before you even get out of bed.

Kids did you know that before the internet, it was highly unusual to see a real person die on video? They used to have urban legends about it, they called them "snuff films." Now we get autoplay videos of exploded soldiers & bloody mass shooting victims beamed right to our pocket

@NainsiDwiv50980 Couldn’t Claude do this?

Most people think using Claude Code is about writing better prompts. It’s not. The real unlock is structuring your repository so Claude can think like an engineer. If your repo is messy, Claude behaves like a chatbot. If your repo is structured, Claude behaves like a developer living inside your codebase. Your project only needs 4 things: • the why → what the system does • the map → where things live • the rules → what’s allowed / forbidden • the workflows → how work gets done I call this: The Anatomy of a Claude Code Project 👇 ━━━━━━━━━━━━━━━ 1️⃣ CLAUDE.md = Repo Memory (Keep it Short) This file is the north star for Claude. Not a massive document. Just three things: • Purpose → why the system exists • Repo map → how the project is structured • Rules + commands → how Claude should operate If CLAUDE.md becomes too long, the model starts missing critical signals. Clarity beats size. ━━━━━━━━━━━━━━━ 2️⃣ .claude/skills/ = Reusable Expert Modes Stop repeating instructions in prompts. Turn common workflows into reusable skills. Examples: • code review checklist • refactoring playbook • debugging workflow • release procedures Now Claude can switch into specialized modes instantly. Result: More consistent outputs across sessions and teammates. ━━━━━━━━━━━━━━━ 3️⃣ .claude/hooks/ = Guardrails Models forget. Hooks don’t. Use hooks for things that must always happen automatically. Examples: • run formatters after edits • trigger tests after core changes • block sensitive directories (auth, billing, migrations) Hooks turn AI workflows into reliable engineering systems. ━━━━━━━━━━━━━━━ 4️⃣ docs/ = Progressive Context Don’t overload prompts with information. Instead, let Claude navigate your documentation. Examples: • architecture overview • ADRs (engineering decisions) • operational runbooks Claude doesn’t need everything in memory. It just needs to know where truth lives. ━━━━━━━━━━━━━━━ 5️⃣ Local CLAUDE.md for Critical Modules Some areas of your system have hidden complexity. Add local context files there. Example: src/auth/CLAUDE.md src/persistence/CLAUDE.md infra/CLAUDE.md Now Claude understands the danger zones exactly when it works in them. This dramatically reduces mistakes. ━━━━━━━━━━━━━━━ Here’s the shift most people miss: Prompting is temporary. Structure is permanent. Once your repository is designed for AI: Claude stops acting like a chatbot... …and starts behaving like a project-native engineer. 🚀

Just clicked "publish" on the first entry of my new blog: a deep dive on my first experiences with MongoDB Atlas Stream Processing. There are github links and memes. jwfbean.dev/developer-firs…

People leaving regular companies: Time for a change! Excited for my next chapter! People leaving AI companies: I have gazed into the endless night and there are shapes out there. We must be kind to one another. I am moving on to study philosophy.

(reddit.com/r/ChatGPT/s/9S…)

My entire feed on every social media platform: OMG AI is progressing incredibly fast, AI writes AI now, all of society is getting disrupted within ten minutes!!!!11!!one! AI, when asked if you should drive or walk to the car wash: walk, definitely walk.

@Theholisticpsyc And when someone is highly emotionally regulated they come across as cold, aloof, dismissive.

When you're around someone with low emotional regulation their mood becomes your problem and you feel chronically anxious in their presence.

@aakashgupta @PeterCorless It’s a signal that he loves poetry and can finally pursue it. 🤷🏻‍♂️

This is being read as a philosophical farewell. It’s a resignation letter from the head of Anthropic’s Safeguards Research Team, and the most important sentence is buried in paragraph three. “I’ve repeatedly seen how hard it is to truly let our values govern our actions. I’ve seen this within myself, within the organization, where we constantly face pressures to set aside what matters most.” That’s the person responsible for keeping Claude safe telling you the pressures to ship are winning. Mrinank Sharma built the Constitutional Classifiers system, developed defenses against AI-assisted bioterrorism, and authored one of the first AI safety cases ever written. Two years of work at the exact intersection of “make the model safe” and “ship the model fast.” And he just walked away. Now zoom out. Dylan Scandinaro, another Anthropic AI safety researcher, left last week to become OpenAI’s Head of Preparedness. Harsh Mehta and Behnam Neyshabur, both senior technical staff, also departed in the past two weeks. Four notable exits in a single month from the company that sells itself as the responsible AI lab. Meanwhile, Anthropic is in talks to raise at a $350B valuation and just launched Opus 4.6 last Thursday. The commercial engine is accelerating. The safety talent is dispersing. This is the core tension of every AI company right now: the people building the guardrails and the people building the revenue targets occupy the same org chart, but they optimize for different variables. When the pressure to scale wins enough internal battles, the safety people don’t fight forever. They leave and write beautifully worded letters about integrity. Sharma’s next move tells you everything. He’s pursuing a poetry degree. When your head of safeguards research decides the most authentic use of his time is writing poems instead of writing safety cases, that’s a signal about what he believes the safety cases were actually accomplishing.

We hired a backend guy recently who didn’t know half the buzzwords. No Saga, no CQRS, shaky on K8s. On paper, easy reject. Then we gave him a real prod-ish bug: sporadic 500s, p95 spikes, only on one endpoint. He did 3 things: 1. Asked for repro + timeline. “When did it start? What changed? Any new feature release?” 2. Cut the problem space. Logs first, then metrics, then a single failing request ID. 3. Formed a hypothesis, tested it, wrote down what each result would mean. Found it in 25 mins: connection pool exhausted from one code path leaking retries + no timeout. I’ll take that over memorized concepts anyday. This is what people don't get right, companies hire for fundamentals + debugging. You can teach patterns. You can’t teach calm thinking under failure.

@GergelyOrosz imagine flipping S3 to strong consistency and nobody even blinks

@gwenshap Me: Hey Claude are these benchmark results valid? Claude: Not only are they valid, but they are *amazing* Narrator: They were not valid.

Since my feed is full of people who didn't code in months because Opus 4.5 is so good, let me chime in with 5 things that I had to manually fix this morning because Claude Code didn't: - Separation of concerns: I asked for new functionality that belonged in a specific class. I told Claude to review that class and make the changes. It made the changes in an unrelated class and then moved some functionality from the class I wanted into a second unrelated class. - Encapsulation. A bunch of private and package-private methods were made public with no good reason. - Concurrency: It implemented a rather blatant concurrency bug and kept insisting that the logic is correct. - Reinventing the wheel: The concurrency bug was in a queue library that didn't need to get written because Guava exists. - Use the wrong build command, then misinterpret the error and make a bunch of totally unnecessary changes. In a loop, since the wrong build command kept failing after each round of useless changes. And yes, I have plan documents, rules document, and I prompted very carefully. Claude Code is impressively good, but human review is very much needed. Especially for concurrency.

AI does things so you don't have to. Like paying attention to someone else and saying nice things to them. AI does that now.

This was a year ago and I was talking about Trump voters.

(And in case you’re not counting, I’ve been occupied by Cloudera and Confluent for about 15 years)

Kayso while I wasn’t looking because I’ve been occupied by Cloudera and Confluent, something *really* interesting has been going on with Postgres. That’s it. That’s the tweet.

🚀I just vibe coded a million dollar SaaS in a matter of hours. I’ll give the first 10 people who review it for me a free subscription for life. 💰Check it out below 👇 http://localhost:3000 #Ai #EntrepreneurLife #Coding

Today's a big day. We’re announcing Agentic Postgres, and kicking off two weeks of launches around it. 🚀 This one's special for me -- a big step toward what we think the future of data infrastructure looks like in the age of AI. Agentic Postgres is the first database built for agents. ▪️ Fluid Storage, a forkable storage layer, that lets every agent spin up its own database in seconds. And then iterate or branch instantly with fast, zero-copy snapshots. ▪️ A new Postgres MCP server that can reason and plan. Your expert DBA that helps you design, tune, and optimize production-ready apps. ▪️ Native search and retrieval built directly into the database. Hybrid keyword and vector search for applications and memory, built for real-time use. ▪️Plus a CLI and free tier, so anyone (human or agent) can spin up and start experimenting immediately. Some of these systems like Fluid Storage have been 18 months in the making. Others are smaller but no less important. Together, they redefine what it means to build with Postgres in an agentic world. For the first time, agents can build, test, and evolve applications on their own. Safely, instantly, and inside Postgres. It's a shift in computing: from databases for developers to databases for agents and developers together. I couldn't be prouder of this team. Try it now in Tiger Cloud. And let your agents cook. #AgenticPostgres #AI #Postgres #Databases #Developers #LaunchWeek @TigerDatabase @TimescaleDB

@gwenshap Humans have always hallucinated as much or more than any LLM and they very rarely say stuff like: I’m sorry, you’re so right!

I just caught myself telling someone, IRL, "you are absolutely right!". Some days we train language models. Other days, language models train us.

@SenAdamSchiff It depends on how many shares of Disney are sold tomorrow.

Kimmel. Colbert. Suits against the New York Times, Wall Street Journal, and 60 Minutes. Extorting settlements from CBS, ABC, and others. Blocking the AP's access to the White House. This administration is responsible for the most blatant attacks on the free press in American history. What will be left of the First Amendment when he’s done?

If you own Disney stock, you should sell it.