Ai on Danilo Falcão da Silva

Karpathy at Anthropic, the Pope at the Vatican, and the Internet Losing Its Mind

Mon, 25 May 2026 18:05:00 -0300

Two things happened in the same week. Both of them involved Anthropic. One was a major hire. The other was a Vatican press conference. Neither was fake. But the internet managed to smash them together into something that was, and now people are asking me if the Pope works at Anthropic.

Let me untangle this.

What actually happened: the Karpathy hire

On May 19, 2026, Andrej Karpathy posted on X: “I’ve joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D.”

PostgreSQL Stopped Being 'Just SQL' a Long Time Ago

Mon, 25 May 2026 17:45:00 -0300

Every few years someone publishes a blog post titled something like “PostgreSQL: The Everything Database” and the comments fill with people saying “well, obviously.” The thing is, it wasn’t obvious. In 2010, if you told a room full of engineers that the correct database for their document store, their geospatial queries, their full-text search, and their vector similarity lookups was the same 30-year-old relational database, they would have politely suggested you hadn’t used MongoDB yet.

Anthropic's Third-Party Claude Crackdown: How It Hit OpenCode, Zed, and the Wider Tooling Ecosystem

Sun, 24 May 2026 22:00:00 -0300

On January 9, 2026, at roughly 02:20 UTC, Anthropic flipped a switch on their servers and broke thousands of developer workflows overnight. No blog post, no advance notice, no migration path. Third-party coding tools — OpenCode, Cline, RooCode, OpenClaw, and others — that had been using Claude subscription OAuth tokens suddenly got a single error: “This credential is only authorized for use with Claude Code and cannot be used for other API requests.”

GPT-5.3-Codex: A Late Review (And Why I'm Paying Attention Now)

Sun, 24 May 2026 21:00:00 -0300

This is a late review. GPT-5.3-Codex shipped on February 5, 2026, and here I am almost four months later writing about it. I have no excuse beyond the usual one — too many things to look at, not enough evenings. But having spent the past few weeks reading benchmarks, watching demos, and following what developers are actually saying about it, I want to be honest: it is a good model. A genuinely good one.

AI Bug Reports: The Real Vulnerability Is That We Weren't Looking Hard Enough

Sat, 23 May 2026 10:00:00 -0300

On May 18, 2026, Linus Torvalds called the Linux kernel security mailing list “almost entirely unmanageable.” The reason: a flood of AI-generated bug reports. The reaction was predictable — ban AI, blame researchers, declare the tools aren’t ready.

I wrote about the maintenance crisis last week and I think that framing misses the deeper story. The problem is not that AI is generating too many reports. The problem is that the code was more broken than we thought, and for twenty years nobody had the tools to look at it properly.

Hermes Agent vs OpenClaw: Why I Run Hermes Every Day

Wed, 20 May 2026 14:30:00 -0300

There are two open-source autonomous agents in 2026 worth a serious DevOps engineer’s time, and they have made opposite architectural bets. I tried both. I run Hermes Agent every day. This is the analysis of why — not a both-sides post, not a head-to-head benchmark, but a direct argument that one of these two architectures is right for infrastructure work and the other one isn’t.

The headline: agent-first beats gateway-first when the work rewards familiarity. Most infrastructure work does. The rest of this post is the why.

AI Bug Reports Are Drowning Open Source — And the Fix Isn't 'Stop Using AI'

Wed, 20 May 2026 10:00:00 -0300

On May 18, 2026, Linus Torvalds said the Linux kernel security mailing list had become “almost entirely unmanageable” because of duplicate AI-generated bug reports. Two months earlier, longtime stable maintainer Willy Tarreau had already shared the numbers: a list that received two to three reports per week in 2024 was getting five to ten reports per day by March 2026. In January, Daniel Stenberg shut down the curl bug bounty after the valid-report rate on HackerOne dropped from above 15% to below 5%, with twenty submissions in 21 days — seven of them in one 16-hour window — and zero real vulnerabilities among them.

Claude Code, opencode, Cursor: My Daily Driver and My Plan B

Tue, 19 May 2026 11:00:00 -0300

I’ve spent enough of the last six months working alongside an AI coding agent that I now have actual opinions, in the way you only get from shipping production code with a tool, not from reading benchmarks about it. There are three names that dominate the conversation in 2026 and they represent three genuinely different bets about how humans and language models should collaborate on code.

This is my honest read on Cursor, Claude Code, and opencode. The headline: Claude Code is my daily driver. opencode is my Plan B. Cursor is not what I reach for. Here’s why.

Zed 1.0: When 'Fast Editor' Finally Stops Being a Marketing Line

Mon, 18 May 2026 21:00:00 -0300

I have a complicated relationship with code editors. I used vim for fifteen years out of stubbornness, switched to VS Code because the ecosystem made it impossible not to, and have spent the last four years quietly resenting it every time the laptop fans spin up because I opened three monorepos and a markdown file.

Zed 1.0 shipped at the end of April 2026. I’ve been running it as my daily driver for the two weeks since, and I’m writing this post in it. Here’s where I’ve actually landed, including the things I think the typical Zed review under-sells.

Musk vs. OpenAI Ends on a Technicality — and the Questions Everyone Wanted Answered Are Still Open

Mon, 18 May 2026 19:30:00 -0300

A nine-person jury in the Northern District of California took less than two hours today to unanimously dismiss every claim in Elon Musk’s lawsuit against Sam Altman, Greg Brockman, OpenAI, and Microsoft. Musk had sought up to $134 billion in disgorgement, the removal of Altman and Brockman from OpenAI’s leadership, and the unwinding of OpenAI’s October 2025 conversion into an $852-billion public benefit corporation.

He got none of it. The technicality matters more than the headline.

Ollama in 2026: The Boring Choice for Open-Source LLMs (And Why That's the Whole Point)

Mon, 18 May 2026 17:30:00 -0300

There’s a particular kind of project I love: the one that takes something new and weird and turns it into something you barely have to think about. Docker did this for containers. Hugo did it for static sites. In 2026, Ollama is doing it for open-source language models.

You install one binary. You type ollama pull <model>. You type ollama run <model>. Your application talks to localhost:11434 over an OpenAI-compatible HTTP API. That’s the whole user manual. The fact that this is now boring — that nobody has to argue about it on Hacker News anymore — is the point of this post.