Coordination (and coherence) are scarce and will be unless we specifically solve for it, and I don't think more intelligence (aka AGI) is going to solve for it.
@owocki What i'd add: coordination primitives still assume you know what the humans in the system actually want. That's what we're trying to unlock with @beaconinfra
Instead of giving my AI instructions every time, it already knows how I think across a range of situations, which produces higher quality responses and lets me delegate more and actually trust the advice I get back.
Over time, those choices build a judgment profile of how I make decisions. Dozens of signals are measured including how I reason under pressure, where I tend to overcorrect, and when I default to caution vs. conviction.
"Itโs infuriating to invest time into reading something only to realize the author didnโt invest the corresponding amount of time into writing it."
Yesssss @joladev
Read the full article ๐
Anyone in my network actively using OpenClaw or Hermes?
I'm looking to connect with a few people who are actively using agents and running into the limits of how well they actually know you.
We're building that with Sync on @beaconinfra - decision game where your digital twin learns to predict your choices. +60 games in, the AI told me I default to structured time-buying in team conflicts rather than being direct upfront. I think of myself as direct. Apparently not.
Zuckerberg is building what Meta calls a "Second Brain" - an AI that indexes company docs and acts as an "AI chief of staff."
That's not a second brain. That's a searchable filing cabinet.
A real second brain would know how you think, not just what you've written. ๐งต
Built an execution plan for a complex AI project. Context window crashed. The AI knew what to do but had no idea what had already been done. Work got redone.
Added a changelog - a running log of actual state. Next crash, two minutes to full recovery ๐ฅ
Training an AI to predict my decisions. 50 rounds in, it's still right only about half the time.
But the accuracy isn't the interesting part. After each round the AI explains its reasoning, and some of those explanations have been weirdly revealing about how I actually think.
To be clear, I don't think they're fully disappearing, but we definitely need to reimagine "landing pages" once a user logs in (I'm thinking more chatboxes with quicklinks to fill in the chatbox).
Visualization of data will still be used, but it will be on demand.
DASHBOARDS ARE DEAD.
Dashboards were always a workaround. They existed because querying data required technical skill most people didn't have. That's largely solved now.
Soon, advertising a dashboard is going to feel like a car company bragging about power windows.