2x founder / 2x father. I build products that solve coordination problems and connect people. Writing: https://xitter.cf/t.co/uQdcGBOEru

London
Joined August 2009
I’ve always been obsessed with two things: markets and communities. These seem unrelated, but actually I think most communities are special kinds markets for things not settled in cash: > Markets for favors > Markets for status and social recognition > Markets for non-public information and opinions ...And also just markets for *humanity*. It feels good to share what you know. It feels good to help people. It feels good to pay it forward. Communities give you the surface area to give back to people you care about, in a way that would be difficult otherwise. Norms can exist within the shelter of communities that would whither in the harsh winds of the outside world. In a world flooded with AI, I suspect small/private communities will become the last bastions of alpha not possessed by the models, and some of the most human places left both offline and online. I’ve been playing around a lot in this space lately, so if any of this resonates stay tuned :)
1
92
Agents using messaging apps designed for humans is basically a hack. It will be interesting to see if the mainstream messengers roll out protocols for personal agents Or if a new agent-native consumer messenger emerges
So uh, Muse just sent this to my wife. @alexandr_wang you got some explaining to do
2
97
Everyone is building agent memory, but what I really want is agent *embodiment* Increasingly, I don't want the agent to act as my assistant — I want it to act as ME: - think like me - talk like me - reason like me I want an agent that can say *I ____* and really believe deep in its little weights that it IS me. Maybe this is playing with fire? But it would be so useful.
3
2
165
I've been thinking about (and building in) this space. 'Teaching in public' feels valuable on LinkedIn, X, and Reddit: - They control agent access - People still read posts with their eyeballs, so they can see it's from you I know lots of people crush it with AEO, but there's a gamble: will you get cited or just trained on? Imo the missing channel is Direct to Agent publishing. When I ask my agent a question, it should respond first with the [private or public] posts, newsletters etc. of the people I've chosen to follow. If those sources exist on a graph (e.g. Substack writers endorsing each other) then you've also got a really interesting new channel for socially-contextualized agentic discovery.
The entire incentive structure for publishing useful information on the open web has changed. Lawyers used to write blog posts about what you needed in a contract. Now what’s the point of doing that? So AI can read it and you never hire them? We may be in the golden era of AI in the sense that in a couple of years, outside of categories like computer science, information about the non-digital world will be closely guarded, and thus will make AI much dumber about physical world subjects.
1
2
130
What an amazing irony Now we get our flying cars *and* our 140 characters from the same guy
but we still don't have flying ca....
2
109
Charles Cushing retweeted
Computer scientist Sara Hooker on underestimating difficulty:
58
2,174
136
21,183
391,503
Feels like a seed fundraising announcement<$50M these days is just an invitation for someone to clone you / fast follow. "Hey guys we found a new pool of demand and are undercapitalized, come after us!"
54
Most apps were built around a push notification from the app itself: app > user. This is a horrible UX and is what led to notif fatigue. The next wave will be built around push from your agent: app > agent > user. Many opportunities to differentiate here.
1
74
A nuance in the "software is dead" debate that hasn't come up enough: Maintaining internal software is MUCH easier than building a SaaS tool you sell to others. Building a SaaS = constantly innovating on features, supporting different platforms, manifold user stories, upselling, cross-selling etc. Building your own software, its possible to get to stability and just... stop building. Because it's good enough.
1
1
90
Not to be a downer, but it really feels like Moloch is winning and will continue to win. Like, is 'Pace the frontier' not just the last cry of a civilization about to be devoured by Moloch?
1
47
Was Bending Spoons suddenly a bad idea? Has anyone tried vibecoding Airtable lately?
There is no moat
1
1
128
I'm not gonna make it
39
Update: it did not disappoint
I hit the Linear free tier issue cap tonight. Instead of upgrading I'm just cloning the features I need into a new app. Agents will run overnight and I fully expect a working build when I wake up. Crazy times.
1
51
Charles Cushing retweeted
Your software needs one of these or it’s worthless: 1. A network effect 2. Corporate relationships 3. Physical atoms
32
18
10
483
39,087
I hit the Linear free tier issue cap tonight. Instead of upgrading I'm just cloning the features I need into a new app. Agents will run overnight and I fully expect a working build when I wake up. Crazy times.
1
1
113
Just spent a million tokens trying to simulate a yeti doing the worm AMA
42
Not a good name guys
2
70