I think one underestimated thing when we look back on it was how useful it is to have your own personal brain and company brain in 2026 at the dawn of usable AGI
AGI gives you the intelligence
You still have to collect your personal context to get the real unlock
Users praised Garry Tan's emphasis on personal context for unlocking AGI because it shows how individual data creates real leverage and competitive moats beyond generic models.
No Digg Deeper questions have been answered for this story yet.
This is why I made GBrain and open sourced it
https://github.com/garrytan/gbrain
I think one underestimated thing when we look back on it was how useful it is to have your own personal brain and company brain in 2026 at the dawn of usable AGI
AGI gives you the intelligence
You still have to collect your personal context to get the real unlock
@sonia_axion @garrytan @ycombinator And internal context is the easy part. The harder unlock is external context, what your customers are signaling before they tell you anything.
@garrytan AGI made a hedge fund *worse* before it made it better.
In 2024, a team replaced their researchers with an LLM. Returns dropped 15% in a quarter. The AI could outthink any analyst—but it didn’t know the traders’ inside jokes, the 2020 meltdown that shaped their strategy, or why
@garrytan Memory will be more valuable than intelligence
An isolated personal brain is like the PC era before the internet. We need to create a network where personal brains can not only iterate on their own, but also register on this network to learn from and subscribe to each other.
Owners can choose to keep their brain private, share it freely on the network, or allow other agents to access it for a fee. This will also fundamentally change how AI acquires training data today — moving from the broker model used by companies like Scale AI, Surge, and Mercor to a fully agentic, automated system.
In this model, every person’s personal brain acts as an agent. On one hand, it maintains and grows its own brain; on the other, it actively broadcasts itself, publishes its needs, and builds its own reputation and influence through mutual citations and subscriptions.
This is exactly what http://noosphere.wiki is building.
@garrytan @HQForWork
@garrytan As intelligence becomes more accessible, context becomes more valuable.The model may know almost everything. It still doesn't know your goals, history, relationships, constraints, or preferences. That's where much of the remaining advantage comes from.
I started a project to remotely control all of my agents but then it slowly became a harness. Now I am making it do mundane things like, “check my slack and give me the details of latest issue” or “planning an upcoming trip” etc. That way it is being recorded in the memory by the nightly curator. I feel the difference after using like this for two weeks, now I don’t wanna go back to raw claude.
it is here for anyone wondering: http://supermux.dev
@garrytan Is GBrain a personal brain or a company brain?
@garrytan AGI gives you the engine. Context is the fuel. But most fuel evaporates as decisions, gut reads, and observations that never hit any system. The moat isn't retrieval. It's capture.
@garrytan @ycombinator raw intelligence is becoming a commodity because anyone can access AGI now. The real moat is hyper specific context and curation ✌🏼
@garrytan Who's gonna create the scan and ingest my Google Drive function to give automated context to AGI?
@garrytan Meta (personal stuff), Google (personal stuff + work stuff), Microsoft (work stuff)are the best firms positioned to build this.
I'm surprised Gemini hasn't released a scan all your email and single sign on social media posts from last 10 years feature, flag non spam.
@garrytan 🐐
@garrytan Gbrain is definitely very useful, but it still only focuses on collecting explicit knowledge. If we want agents to actually have agency, we have to find a way to extract tacit knowledge too
@garrytan AGI’s power is nothing without your personal context layer.
@garrytan If we have AGI it should be able to gather all the context it needed without duct taping. Isn’t it?
@garrytan Has anyone created the 'collect your context for you' app? Rewind was kinda there at the start.
@garrytan Context is the whole game. My agent re-prices hotel bookings after checkout and it only beats humans because its algo identifies every price from every country. The intelligence was never the bottleneck
@julleybuilds @garrytan @ycombinator AGI can process the feedback once it's given, but capturing the unspoken behavior and vibe of your users is still an art, that's the ultimate feedback loop.
