William Oliveira

16 Jan

Eu ando MUITO afastado das redes sociais. Por conta disso, estou sentindo saudade da galerinha que acompanho por aqui. Vocês estão em algum grupo de WhatsApp ou similar? Me diga aqui ou manda no privado ❤️

github.blog/ai-and-ml/ge...

18 Sept

github.blogMigrating the GitHub Copilot runtime to Rust, using CopilotA rewrite this size wasn't affordable before agents. Here's what porting the Copilot agent runtime to 800,000 lines of production Rust actually took.

alltechishuman.org/all-tech-is-...

17 Sept

alltechishuman.orgTwo Views of the Same Transformation: AI Adoption, Invisible Work, and the Human Side of Organizational Change — All Tech Is HumanDownload the new whitepaper by All Tech Is Human and Notre Dame-IBM Technology Ethics Lab, Two Views of the Same Transformation: AI Adoption, Invisible Work, and the Human Side of Organizational Chan...

woliveiras.comHow Early Is Too Early to Warn About a Bubble? Looking at AI in 2026Historical warnings arrived months or years before market peaks. This article compares those lead times with the AI investment boom and asks which signals matter more than the word bubble.

16 Sept

How Early Is Too Early to Warn About a Bubble? Looking at AI in 2026 woliveiras.com/posts/how-ea...

arxiv.orgThroughput per Megabyte: A Pilot Benchmark of Language-Stack Efficiency for Self-Hosted HTTP Services on a Raspberry Pi 5Cloud-centric web benchmarks miss constraints that matter for self-hosted services on ARM64 single-board computers, especially idle RAM footprint and energy per request. We ran a pilot benchmark on on...

14 Sept

Throughput per Megabyte: A Pilot Benchmark of Language-Stack Efficiency for Self-Hosted HTTP Services on a Raspberry Pi 5 arxiv.org/abs/2609.11932

youtu.beCOMO a IA resolveu NAVIER-STOKES (Problema do Milênio)YouTube video by Tem Ciência

13 Sept

youtu.be/KGTe0QAELVU?...

brennan.day/normalized-f...

11 Sept

brennan.dayNormalized Fascism in Open Source: $12 Million Given to DHHOpen source has become a safe harbor for far-right politics, from David Heinemeier Hansson's white-nationalist rhetoric and the Omarchy Foundation's $12 mi…

William Oliveira reposted

Steam

@steampowered.com

10 Sept

Do you have a strong opinion on tabs versus spaces? If yes, Steam Programming Fest might be for you - from now until September 14, check out all sorts of games that have you building for loops, cursing at syntax, and solving all sorts of logic puzzles. store.steampowered.com/category/pro...

store.steampowered.comProgrammingSteam Programming Fest 2026

vitest.devAnnouncing Vitest 5.0Vitest 5.0 Release Announcement

9 Sept

vitest.dev/blog/vitest-5

www.oneusefulthing.org/p/agency-and...

31 Aug

oneusefulthing.orgAgency and AgentsFrom the Hugging Face Incident to Twilight Factories

woliveiras.com/posts/automa...

28 Aug

woliveiras.comAutomating Weekly Sentry Triage with GitHub Actions and CopilotA practical architecture for selecting actionable Sentry issues every week, asking humans for approval, and handing one bounded task to GitHub Copilot.

youtu.beHarness Engineering is not Enough: Why Software Factories Fail — Dex Horthy, HumanLayerYouTube video by AI Engineer

20 Aug

youtu.be/Ib5GBkD555M?...

woliveiras.comBaselineBaseline gives coding agents a proportional path from a software request to a verified change, while keeping risk, evidence, safety, and authority explicit.

19 Aug

I just launched Baseline: open-source Agent Skills for disciplined, proportional software engineering with coding agents. Risk before ritual. Evidence before confidence. Safety before speed. Human authority stays explicit. woliveiras.com/baseline/

William Oliveira reposted

Matt Kane

@mk.gg

19 Aug

Great post from Lovable about how and why they migrated from Next.js on Vercel to TanStack Start on Cloudflare.

lovable.devHow we migrated lovable.dev away from Next.js and turned it into another Lovable app | Lovable42M monthly visitors, 400 routes and 850K lines of code moved from Next.js on Vercel to TanStack Start on Lovable—route by route, mostly by agents, with one 11-minute incident along the way.

www.youtube.com/watch?v=87Dy...

19 Aug

youtube.comBlack Hat USA 2026 | The 'Breaking' News: The OpenAI–Hugging Face IncidentYouTube video by Black Hat

woliveiras.comWhy Git Worktrees Matter More in the Age of Coding AgentsGit worktrees predate coding agents. Learn what they share, what they isolate, and why parallel agents make this old Git feature feel like infrastructure.

15 Aug

After studying how coding agents interact with our projects and how we can build an efficient harness to prevent errors from these actors, I realized that worktrees are an important strategy for isolating the work of different agents working simultaneously woliveiras.com/posts/why-gi...

github.comGitHub - woliveiras/storehouse: A curated set of skills for Phaser, Godot, Android Native, TypeScript, Python, Golang, Rust, Terraform, Data Modeling, Cloud Infrastructure and AI EngineeringA curated set of skills for Phaser, Godot, Android Native, TypeScript, Python, Golang, Rust, Terraform, Data Modeling, Cloud Infrastructure and AI Engineering - woliveiras/storehouse

A curated set of skills for product design and performance, Phaser, Godot, Android Native, TypeScript, Python, Golang, Rust, Terraform, data modeling, cloud infrastructure, research, writing, software engineering, and AI engineering github.com/woliveiras/s...

Corsarr v1.2.2 is now available! 🚀 This release makes the onboarding experience smoother and more reliable Feedback and contributions are always welcome! #OpenSource #SelfHosted #Docker #MediaServer #GoLang #Go

🌍 Support for English, Spanish, Portuguese, and Italian 🎬 Guided quality profile setup for Radarr and Sonarr using Recyclarr 🐳 Improved Docker startup detection and automatic status recovery 🪼 More reliable Jellyfin configuration

github.comRelease v1.2.2 · woliveiras/corsarrDesktop support Corsarr Desktop for macOS is the supported non-technical onboarding release. Windows and Linux Desktop archives are experimental previews: they launch on their target systems, but a...

Corsarr Desktop for macOS is the supported release, while Windows and Linux builds remain available as experimental previews. Download it here: github.com/woliveiras/c...

engineeringblog.yelp.com/2026/08/migr...

12 Aug

engineeringblog.yelp.comMigrating a Large Flow Monorepo to TypeScriptMigrating a Large Flow Monorepo to TypeScript Shawn Walton, Software Engineer Aug 4, 2026 In early 2017, Webcore selected Flow as Yelp’s next-generation typechecker over TypeScript. At the time there....

12 Aug

Parece que os LLMs só conhecem Supabase como ferramenta para criar um produto digital 👀👀👀

corsarr.woliveiras.comCorsarr — Your media server, made simpleInstall, connect, and care for your media-server apps in one place.

11 Aug

Set up a complete experience to organize, download subtitles, and watch your movies and shows without complicated configuration files, creating VMs, or understanding networking corsarr.woliveiras.com/en/

github.comSupport PNPM v11 · Issue #14794 · dependabot/dependabot-coreIs there an existing issue for this? I have searched the existing issues Feature description PNPM vesrion 11 is in release candidate stage, and I would like to use it in my project. However, I see ...

11 Aug

O ano em que eu mais havia produzido código na vida foi em 2016 com ~1900 commits. Agora, em pouco mais de meio ano de 2026 já foram ~3600 commits e eu não me vejo parando, pois ainda tenho coisa pra caramba pra terminar Os agentes de código me animaram muito

William Oliveira reposted

pnpm

@pnpm.io

11 Aug

Dependabot still doesn't support pnpm v11. Upvote this issue please: github.com/dependabot/d...

10 Aug

Código não importa, o que importa é código!

This post is unavailable.

10 Aug

Criando um cliente desktop para o Corsarr para ficar mais fácil pra galera não dev a criar um servidor de mídia 👀

William Oliveira reposted

Nick Miller

@nickdmiller.bsky.social

abc.net.auHow a simple request for AI to book a gym class exposed a major threatWhen Andrew asked his AI personal assistant to book him a spot in a gym class, he had no idea he would accidentally initiate an autonomous cyber attack.

Great story from ABC this morning about a chap in Melbourne who asked an AI agent (OpenClaw) to book a gym class, and it ended up hacking the gym's booking system to kick people off the waiting list. www.abc.net.au/news/2026-08...

William Oliveira reposted

Liz Fong-Jones (方禮真)

@lizthegrey.com

10 Aug

let me guess, another gpt-4o case

Mother Jones@motherjones.com

Women just hate me. Idk what to do about it…What’s the point in living." The accused FSU mass shooter fantasized about violent sex—a scenario ChatGPT scripted for him in graphic detail. AI chatbots fueling violent thinking and planning among troubled people is an urgent problem, experts warn.

motherjones.comInside a mass shooter’s harrowing history with ChatGPTA year of chats reveals glaring red flags—and disturbing ChatGPT replies—from long before a rampage at Florida State University.

Acabei de publicar! Are companies really replacing GitHub Copilot, Cursor, Claude Code, or Codex with local AI models? The answer is more nuanced than “just run an open model on a GPU.”

The evidence suggests that control-plane ownership usually comes before full inference replacement. In many cases, companies still rely heavily on external frontier models, while moving high-volume or sensitive workloads to specialized internal systems.

woliveiras.comHow Companies Are Replacing AI Coding Subscriptions with Internal Developer PlatformsCompanies are unbundling coding assistants into internal agents, gateways, private inference, tools, and governance. Real deployments show why control often moves in-house before frontier inference do...

I analyzed these deployments, their economics, architectural patterns, and limitations in the full article: **How Companies Are Replacing AI Coding Subscriptions with Internal Developer Platforms** Read the complete post: woliveiras.com/posts/how-co...

Acabei de publicar! Are companies really replacing GitHub Copilot, Cursor, Claude Code, or Codex with local AI models? The answer is more nuanced than “just run an open model on a GPU.”

Companies such as Cloudflare, Samsung, Meta, Walmart, Zup Innovation (Brazilian like me ❤️), and others are taking different parts of the stack in-house:

- Internal coding agents and model-neutral clients - SSO, access control, centralized policy - AI gateways and model routing - Private or self-hosted inference - Repository, documentation, CI, and MCP integrations - Sandboxed execution and observability - Frontier-model fallbacks for difficult tasks

Acabei de publicar! Are companies really replacing GitHub Copilot, Cursor, Claude Code, or Codex with local AI models? The answer is more nuanced than “just run an open model on a GPU.”

The most interesting shift is not from one model to another. It is from renting an inseparable coding product to operating a platform where agents, models, tools, and providers can be replaced independently.

github.comGitHub - woliveiras/tuxedo: Spec Driven and Test Driven software engineering skillsSpec Driven and Test Driven software engineering skills - woliveiras/tuxedo

7 Aug

Revisei e atualizei meus skills de engenharia de software (trabalho com Spec Driven e Test Driven Development) Agora, além da validação prática, adicionei os testes com promptfoo pra garantir que os skills, router e tudo mais estejam funcionando corretamente github.com/woliveiras/t...

5 Aug

Escreve código que não sabe se funciona, escreve teste que não sabe se testa, mas o agente disse que você está absolutamente certo 👍

github.comGitHub - woliveiras/web-graphics-skills: Production-oriented Agent Skills for building, reviewing, debugging, testing, and profiling browser graphics codeProduction-oriented Agent Skills for building, reviewing, debugging, testing, and profiling browser graphics code - woliveiras/web-graphics-skills

3 Aug

Galera do bem, eu trabalho bastante com #webGPU e #webGL (MapLibre, MapGL). Nenhum skill estava me atendendo bem (a maior parte do tempo eu passava tendo retrabalho), então hoje eu decidi criar meu próprio set de skills para trabalhar com isso Mais detalhes aqui: github.com/woliveiras/w...

2 Aug

Está no ar a atualização do Rusted com tradução para pt-br, novas configurações, mais feedback visual, melhorias de campanha e correção de bugs rusted.woliveiras.com

youtu.beA VERDADE SOBRE OS CAMPEONATOS DE FARMAR AURAYouTube video by República Coisa de Nerd

2 Aug

youtu.be/e24HYt7604I?...

William Oliveira reposted

Scott Tolinski

@tolin.ski

forms.gleAI Coding & Mental HealthAll answers are anonymous.

I’m exploring the mental health effects of AI coding. This is a super quick (blazing fast), anonymous survey. Please fill out and share so that the responses will be well rounded. forms.gle/dpD8Tk1QnXAz...

aws.amazon.comSecure code execution for AI agents with AWS Lambda MicroVMs | Amazon Web ServicesDevelopment teams building serverless applications with AI coding agents face the question of how to let those agents generate and execute code without losing control over governance. Agent-generated ...

1 Aug

Secure code execution for AI agents with AWS Lambda MicroVMs aws.amazon.com/blogs/comput...

etsy.comEtsy Engineering | Building Etsy Buyer Profiles with LLMsEvery day, shoppers from Etsy's community of nearly 90M buyers visit our marketplace to search for unique, handmade, and vintage...

1 Aug

Look at this awesome post from Etsy's Code and Craft! www.etsy.com/codeascraft/...

rusted.woliveiras.comRustedRusted is a casual browser roguelike autobattler.

Tá em produção! Final de semana eu vou trabalhar nas traduções e melhorias de UX rusted.woliveiras.com

woliveiras.com.brAbandonando ou reduzindo o uso das redes sociaisSobre o porquê estou abandonando as redes sociais. Pode ser um adeus ou um até logo, mas até agora tem sido muito bom.

Desde 2020 reduzindo o uso de redes sociais Hoje estava pensando na quantidade de gente que parou de me responder nas DMs por conta da minha ausência digital. É triste woliveiras.com.br/posts/abando...