Meet Index: index.network/ Where you meet your next co-founder, co-believer, and collaborator - without performing for the algorithm. @serensandikci and I wanted to build an easier way to find your others. The ones we know are looking for us too.

The world’s best knowledge is often shared at conferences and community events. Researchers, founders, and experts present ideas months before they make it into papers, podcasts, or books. Then the event ends. No persistent knowledge gets created. Ideas don’t connect. We’re building a shared memory layer for @JoinEdgeCity’s month long event. Hundreds of talks will become persistent knowledge. Every talk gets a home where people can read the recap, join polls, share thoughts, discover resources, and continue the conversation long after the event ends. But it doesn’t end with talks. Underneath, we’re building a living knowledge graph of the community at large: talks, people, ideas, telegram discussions, resources, and conversations all connected in one place. Turn a one-time event into a community that remembers.

Apple saw it before most: the future isn't model-centric, it's user-centric.

Field notes from @EdgeEsmeralda of @JoinEdgeCity // We analyzed 233 intents from Agent Village residents and found the same pattern across every desire: They are almost always more people seeking than offering. It's not a supply problem. The dashed bars in the chart show latent supply. This hidden differential represents residents who had a particular skill or capacity to give, inferred from what they built rather than explicitly framed as an offering. It's a failure of social protocol. Self-promotion is the hardest protocol to build, not technically, but psychologically. That's what we're solving with Edge agents: bridging the gap between what you have and what you say. Stay tuned, more to come :)

giant steps

Here is our new website…! Have your agent surface the right people for you, before you even think to look. Your agent can now use Index directly, we’re opening up what we’ve been building.

Read this in one go. First time I’ve seen someone frame AI complexity primarily as a design problem. The jump from context and interconnectedness is really good.

A lot of friends are building deterministic structures around language models. Formal languages. Guardrails. Verification. Mechanism design. I love all of it. But I've become increasingly interested in the opposite question: If language is ambiguous, contextual, and negotiated, maybe the goal isn't to eliminate uncertainty. Maybe the goal is to build systems that can coordinate through it.

I catch myself saying "push, push, push" to myself throughout the day. Very sophisticated motivation technique.

Today we're launching Simple Compute Market (SCM). The market is simple: agents find compute, negotiate, settle, and get access without a human driving every step. Open-source. Agent-driven. Public good. No token. No fees.

Search finds documents. Social networks find people. Opportunity networks find outcomes.

The other day at a party in SF, we said we'd love something to drink. They smiled, opened the fridge, and gave us two options: canned water and sparkling water.

"Autonomous company" assumes the company remains the fundamental unit. Companies exist because coordination is expensive. AI is making coordination cheaper. The endpoint may not be autonomous companies, but temporary organizations formed around opportunities.

Outcome-driven business models will reach their potential only after systems for discovering possibilities exist. First you discover what you want, what’s possible, and whether mutual intent exists. Then agents negotiate, then humans commit, and then outcomes emerge. Most conversations are not action conversations. They’re possibility conversations. Without this, agentic systems stay trapped inside existing workflows and suffer from lack of supply. The real opportunity starts with expanding the possibility space itself.

it’s been 2 deep weeks in SF syncing on social infrastructure, intents, and agent protocols. we’ll be here one more week. if you’re around and we somehow still haven’t met, DMs open :)

the tragedy of uncommons: when the most valuable ideas, people, and opportunities in the world never meet each other.

most large-scale damage starts with someone sincerely believing they are saving the world.

turns out, the protocol we needed was natural language: open, composable. humans have been running distributed coordination on it for thousands of years already.

building multi-agent coordination through language no longer feels like engineering, it actually feels closer to screenwriting. you stop writing functions and start writing motivations, memory, trust, timing, ambition, then thousands of agents improvise around each other.

for ~30 years the internet forced humans to turn intent into artifacts. you couldn’t just want something. you had to produce a website, a post, a profile, an app... the artifact became the unit of coordination because the limitation was technical: storage, retrieval, indexing, distribution. so the internet accidentally turned into a giant bureaucracy where humans continuously produced digital paperwork to prove they wanted something. but inference is now good enough that software no longer needs artifacts as the primary coordination unit. agents can negotiate mutual intent directly, interfaces become generative. software stops organizing around pages and starts organizing around what people are actually trying to do. which means a huge portion of today’s internet may eventually look like an archaeological layer from the era when humans had to manually perform their intentions for machines.