Productivity Hacking: From Second Brain to AI Coworker

Why 2026 Changes Everything About Personal Knowledge Management

2025 was the year we collectively figured out vibe coding, the ROI of AI, and which tools to use for what. The paradigm for software engineering now feels relatively settled, and AI-assisted coding is the norm. But what about everything else?

In 2026, I’m trying to figure out how to optimize my day-to-day work. These are some lessons learned so far. If you have productivity tips or AI-driven micro-optimizations, I’d like to hear examples of how you operate (or point me to where I can learn more).

Situation

At work, when I take notes just for myself (vs collaboratively), I use Google Docs, then rely on an AI tool called Kiro (with MCP) to ask questions like: “what was the thing we said we were going to do for the ML pipeline project?” and it pulls the answer from scattered docs, Gemini meeting notes, and Slack. It works, but my notes still live everywhere, and I still need to remember roughly when and where I wrote something to find it efficiently.

In my personal life, I’ve tried Notion, Asana, spreadsheets, Google Docs, Apple Notes, Obsidian, Keep – basically anything to help me create that “second brain.” Each worked for a while, each had tradeoffs. What consistently happens is that my notes end up fragmented because I take them where the work happens: Slack, email, notebooks, agenda docs, etc. and I get overwhelmed by the maintenance. I no longer have time to design or maintain “perfect” systems and I use what’s available in the moment.

Constraints and preferences

I use Kiro at work because Amazon is bullish on it (and I work at Twitch, an Amazon subsidiary). For my use case, it’s actually not bad, and I prefer it over some of the other internal options. At home, I use Claude Code since I’m already paying for a subscription. (I also subscribe to ChatGPT. Each tool has strengths, and I like having options.)

I’m focused on optimizing my work knowledge management system because that’s where I context-switch the most, collaborate with hundreds of people across various projects, and constantly process/translate information for stakeholders at different levels.

At minimum, the system needs extremely low latency and friction. I can’t wait more than ~300ms for an app to load or fight through login screens in the middle of a meeting. I also need to be able to search later without perfectly organizing everything up front. Using AI with MCP capabilities largely solves the retrieval side for me.

What I’m trying differently

I switched back to Obsidian recently, paired with Raycast for fast keyboard shortcuts to specific notes and commands, and set up a couple of Kiro hooks:

• One extracts tasks assigned to me from Gemini meeting notes during the day (and writes them into Obsidian).

• Another generates a daily priority list by looking at my calendar and recent notes across different sources to infer what I should focus on that day: follow-ups, docs to read or write, dependencies/risks to investigate.

I also defined a Kiro agent steering file with north star goals for the year and TPM role guidelines so my daily work ties back to larger company objectives.

Historically, I organized notes by subject (one doc per project). I’m now trying “daily notes” instead: chronological brain dumps with AI handling organization later. It feels chaotic, so we’ll see how it goes. The read side is mostly solved (ask AI for things), so I thought the harder problem was the write side. But the challenge is actually in creating a personalized system.

The keys to making this successful

What I ultimately want is to click buttons at the end of the day and have AI summarize what happened – but only the things I’d actually care about. The breakthrough would be teaching the system what “important” means to me specifically. Right now, everything is surfaced with equal weight, which means I’m still doing the cognitive work of filtering signal from noise.

I need a personalized system: one that understands my communication style, my approach to coordination, and what needs my attention. This is all shaped by context and years of experience, so it really boils down to context engineering and rather than telling the AI what’s important to me, I gave it examples of my actual work: my daily standup notes where i naturally highlighted what mattered, my Slack messages where i escalated or summarized for leadership, and program updates i’ve written as examples of what i prioritize communicating. I put these into a separate agent steering file so that when AI synthesizes summaries and tasks, it will hopefully be less generic and more focused on what I find to be important.

Different AI tools have similar concepts like Kiro’s agent steering files and hooks, so this can be implemented within the tool of your choice.

What’s your personal KMS look like?

If these tools had existed when I was in college, or when I had more leisure time in my life, they could have enabled a different level of productivity. For busy professionals now, I think we’re at a point where we need to evolve how we read, write, and process information. The baseline expectation may soon be that everyone has a personalized “second brain” assisting how they operate at work. Executives will continue to have human executive assistants but for everyone else, we now have the capability to build our own AI assistants.

If you’ve built a personal knowledge management system that actually sticks, especially one using AI for retrieval or synthesis, I’d love to hear examples of how it works and what your setup looks like. I tried finding examples, but most content skews toward tutorials rather than real-world systems in use. I know everyone’s brain works differently, but I think seeing how others operate is still useful.

I’ll follow up mid-year if this system is still working for me!