What went well
AI coding tools went from "cute toy" to "I genuinely cannot imagine working without this." Claude Code became my pair programmer. I can describe a complex migration across 42 NestJS modules and it does 80% of the work correctly, which is a lot more useful than the fizzbuzz demos people clap for. The remaining 20% is where my brain still matters, and honestly, that's the fun part.
TOPO Contabil grew from an ambitious project to a platform with 42 modules and 252 use cases. Clients use it to manage their money. The Brazilian tax system is still insane, but our software makes it less insane, and that feels meaningful.
Yooga had a strong year. The POS system keeps evolving, restaurants keep signing up, and our payment processing keeps improving. Being a Tech Lead at a company that's actually growing and profitable is a rare and wonderful thing.
The M4 Max MacBook changed my relationship with my tools. It sounds dramatic but when your machine stops being a constraint, you work differently. I wait less, so I can experiment and iterate faster. It's the first computer I've had where I never think about the hardware.
Sigil Protocol took shape. AI agent identity is going to matter more and more as agents become autonomous. We're early, but the foundation is solid.
FlowForm proved that LLM-powered conversational forms can work better than traditional ones. Users complete them at higher rates because talking is more natural than filling boxes.
What did not
AI hype reached truly unbearable levels. Every startup became an "AI company." Every feature became "AI-powered." Your company added a ChatGPT wrapper and called it innovation. The VC money flowed like water toward anything with "AI" in the pitch deck, regardless of whether the product made any sense.
The job market stayed weird. Companies laying off engineers while claiming they can't find talent. Senior roles demanding 15 technologies. Junior roles requiring 3 years of experience. The whole thing is broken and nobody is fixing it because it's cheaper to complain on LinkedIn than to actually hire and train people.
Open source sustainability is still a joke. The most critical infrastructure on the internet is maintained by exhausted volunteers while billion-dollar companies use it for free and send stickers as thanks. We have not fixed this, and I do not think we are going to.
What people made up
LinkedIn became truly unhinged. Every post is either "I was FIRED and here's why that's the BEST thing that ever happened to me" or "I gave my employees UNLIMITED vacation and our revenue TRIPLED." Nobody tells the truth on LinkedIn. It's a platform of performed success.
AI-generated content flooded every platform. Blog posts that say nothing in 2000 words. Twitter threads that are clearly GPT-4 with a personality prompt. Comment sections full of "Great insight! I particularly agree with your point about..." which is 100% generated because no human talks like that.
"AI will replace developers" articles were written every single week by people who have never written a line of code. AI replaces tasks, leaving developers who learn to use the tools in a position to replace those who do not. That's been true of every tool since the invention of the compiler.
My year
I shipped more this year than any year before. The combination of Claude Code, the M4 Max, and accumulated architectural knowledge meant I could move faster without cutting corners. Both quality and speed went up, which is rare.
I got better at turning down features that do not matter and refactors that feel good without helping users. I also said no to meetings that could be messages. This is maybe the most important skill a tech lead can develop.
I kept writing on this blog. I became the dev who builds a blog and actually uses it, fulfilling the prophecy.
2025 was good. Not perfect, but good. The work was meaningful, the tools were better, and the code shipped.
Let's see what 2026 breaks.