IT Revolution

We love seeing the tools people are creating!

A Geração Z e Alpha não vão aprender a programar. E não deveriam mesmo! Semana passada, dev sênior me soltou essa: “Thiago, esse vibe coding é furada. Daqui 1 ano vão me ligar pra arrumar essa bagunça cheia de falha de segurança.” Errado. E vou te mostrar por quê. A objeção clássica: - “Ah, mas código gerado por IA tem falha!” - “Vai dar ruim e os devs de verdade vão ter que limpar!” - “Segurança? Esquece!” Cara, você tá criticando a ferramenta errada. O problema não é a IA. É quem não sabe usar. Quem tá faturando $20k/mês com IA sabe fazer isso: 1. Estruturar contexto em .md 2. Criar guardrails de segurança 3. Validar output automaticamente Enquanto você debuga linha por linha, eles escrevem specs em markdown que geram código production-ready com: - Testes automatizados - Security checks built-in - Logs estruturados - Error handling completo E muito mais (você precisa ver meu setup pra Vibe Coding 😅) A diferença brutal: Junior sem noção: Joga prompt genérico → Código bugado AI Orchestrator: Estrutura specs em markdown → Define guardrails → Valida output → Produção Um gera dívida técnica. Outro gera código melhor que 70% dos devs. Se você ainda acha que “escrever código limpo” é sua vantagem competitiva, você tá competindo com algo que faz isso instantaneamente atualmente (e tá melhorando absurdamente a cada dia). E faz com os guardrails que VOCÊ especificou. A skill valiosa não é digitar sintaxe. É estruturar problemas complexos em formato executável com segurança, observabilidade e testes por padrão. Gen Z já sacou. E tá faturando enquanto você reclama no X/LinkedIn. P.S.: “Mas vai dar ruim ainda!” Vai dar ruim pra quem não sabe estruturar specs. Pra quem sabe, é código melhor que você escreveria em 10x menos tempo. #Tech #AI #Programming #Security

One of the most significant advancements in the second edition of the Team Topologies book, is the elevated appreciation for cognitive load as a fundamental design principle in the operation of work systems. In an environment of relentless change, centering human mental load in organisational design is not merely practical but vital. When teams are overwhelmed, quality declines, innovation stalls, and risks escalate. By optimising cognitive load within well-defined team boundaries, organisations foster environments where teams can focus deeply on their specific domains, leading to deeper expertise, more innovative solutions, and ultimately, superior outcomes. Critically, the application of cognitive load principles has expanded far beyond IT. We've seen its central role in achieving success in diverse contexts such as pharmaceutical research and development, legal case management, HR, and emergency healthcare, all yielding safe, rapid, and compliant outcomes. The second edition introduces a scientific model for systematically assessing cognitive load in knowledge-intensive teams, developed with Dr. Laura Weis. This model identifies over twenty drivers across four clusters: team characteristics, work practices and processes, task characteristics, and work environments and tools, offering unprecedented clarity on factors impacting team effectiveness. Leaders must adopt a dynamic view of cognitive load, recognising that temporary increases may be necessary for clear objectives like increased service ownership or productivity. However, continuously piling on responsibilities without adequate support or allowing complexity to proliferate without robust platforms is detrimental. The Creditas case study, one of the new case studies featured in the second edition of the book, exemplifies mature usage of team cognitive load (TCL) assessments, showing how regular analysis of TCL drivers, combined with contextual team discussions, can accelerate delivery and boost team ownership and engagement. Read more https://lnkd.in/eCh5U3wv

Navigating the Tech Jerks of 2025: A Snapshot of Acceleration in Action We’ve always been fascinated by the concept of the “tech jerk”—that sudden, often uncomfortable change in the acceleration of a product or service that disrupts the user’s experience. Borrowed from physics, where jerk describes a change in acceleration (think of a lurching elevator or a car braking too hard), the term perfectly captures how poorly managed technological change can leave users feeling disoriented, frustrated, or even abandoned....

Performance testing can't be an afterthought when AI is generating code at scale. New functionality needs to maintain or improve performance profiles. Without systematic validation, regressions slip through and degrade user experience. The challenge is running meaningful load tests in pre-production environments that actually mirror production. The recommendation from Robbie Daitzman and Christina Yakomin: run moderate load tests before each deployment representing typical daily traffic patterns. Test in environments where all dependencies match production configurations or use service virtualization. Yes, ephemeral isolated test environments are technically challenging. But they provide the most accurate performance signals with minimal external interference. When code generation accelerates, performance validation becomes the new bottleneck. Better to invest in the infrastructure now than discover problems in production. https://itrev.io/4qrlPIp

Here's what's fascinating about the AI code generation challenge: the highest ROI improvements are often within direct control of technology executives. Not product strategy. Not market positioning. The PR-to-production pipeline—code review, testing, policy enforcement, performance validation, resilient rollout. These downstream processes determine whether AI generates value or just generates backlog. And unlike many enterprise transformation initiatives, these improvements don't require board approval or cross-divisional alignment. Robbie Daitzman and Christina Yakomin's research maps out exactly where to start: identify the biggest bottleneck, automate it, move to the next one. Build a sustainable delivery model that evolves with the technology landscape. The fire hose is already on. The only question is whether the drain can handle it. https://itrev.io/4qrlPIp

[YOU NEED TO SUPPORT #AI IMPLEMENTATIONS] In the 90s, we didn't just drop PCs on desks. We built IT Help Desks. Today, we are dropping AI on people and walking away, leaving thousands of #Copilot licenses underutilized and executives frustrated. The solution isn't more training. It's structural. You need to support AI rollouts in a meaningful, ongoing way. In my latest Substack article, I provide ways from the #HyperadaptiveModel to provide this support. If you want flow, you have to build the pipes. https://lnkd.in/gaisj3xb #AITransformation #AILeadership #AIsupport

I've seen it a lot: even with great people, work gets sluggish when teams don't know their goal or have too much to do. Team Topologies, 2nd Edition, looks at how to set up teams so they concentrate on what matters. Here are some patterns we point out: ⚡ Clear boundaries to reduce cognitive load ⚡ Teams empowered to make decisions quickly ⚡ Work organised to minimise handoffs and delays ⚡ Technology used to support humans, not replace them Organisations using these approaches report faster outcomes, more innovation, and teams that feel more engaged. Grab your copy here → https://lnkd.in/eu35PvqD

Learning a lot from Gene Kim and Steve Yegge's "Vibe Coding" book. It's about AI, but it's also about being a better, more responsible engineer—doing work in small increments, testing everything, building a keen sense of what can go wrong. A great entry point into working with these tools with the right mindset.