Taru樽
A collaborative AI research partner that helps you cultivate a rich knowledge garden.
Origin Story
Over the past five years, I've consumed thousands of pieces of content—newsletters, podcasts, articles from thinkers I admire. Every idea I've written is some combination of what I've read and my own experiences. But I have zero tangible documentation of all that intellectual input. I can't trace the lineage of my own thinking.
The options available weren't working. Newsletters and podcasts give me agency—I choose exactly what I read and listen to—but the insights just evaporate. I'd read something brilliant on Monday, and by Friday it was gone. Social media has the same evaporation problem, plus the added issue that my attention is being algorithmically mediated for advertisers' benefit, not mine.
Existing knowledge management tools promised to solve this by automating everything: save it all, let AI summarize, search when you need it. But that felt like surrendering the very thing I valued—conscious choice about what enters my mind.
What I actually needed was a research partner. Someone who could do the pre-reading, help me identify what's worth going deeper on, and synthesize connections I couldn't see across 50 sources over months—but still let me make the calls. I built Taru to be that partner: an AI research assistant that works with me to tend my knowledge garden, not an autonomous agent that tends it for me.
Design Principles
Conscious Consumption That Compounds
Most tools force a tradeoff: agency (you choose what to consume) or accumulation (insights build over time). Taru gives you both. You control what enters the system. Those insights actually grow into something you can draw on.
Partnership > Automation
Taru functions as a research assistant who does the synthesis work humans can't do at scale. But you remain the gardener. The AI pre-reads, surfaces connections, identifies contradictions. You decide what stays.
Aligned AI > Black-Box
Social media algorithms choose what you see and never explain why. Taru only works with content you've consciously added. When it says 'this is relevant,' it tells you why—connecting to articles you previously read.
Earned Trust > Blind Delegation
You wouldn't hand a human research assistant full autonomy on day one. More feedback upfront, more autonomy over time as trust builds. The cultivation review isn't resistance to automation—it's onboarding.
What It Does
Taru processes the content sources you've chosen to follow — newsletters, podcasts, articles — and serves as your research partner in making sense of it all.
Triage
The system does a quick pre-read and helps you understand what's worth diving deeper into, so you're not wading through everything yourself. Quick decisions: read now, save for later, or skip entirely.
Deep Analysis
For content you select, Taru performs deeper research—extracting insights, identifying key themes, and connecting them to what's already in your knowledge garden.
Cultivation Review
You see what Taru suggests adding to your garden and how it wants to organize it. You approve, edit, or reject. The system learns your preferences over time, earning more autonomy as it demonstrates understanding of what matters to you.
Thought Partnership
Once your garden has depth, Taru becomes a conversation partner. Share a half-formed idea, and it surfaces connections and contradictions from your own reading. Eventually, it can steelman arguments against your thinking—not from generic internet sources, but from the sources you chose to follow.
The result: After a year, you have an actual intellectual asset. A second brain that knows what you've read, how it connects, and can help you sharpen your thinking when you need it.
Built with Python, Next.js, and Claude · Last updated: November 2024