aggressively curating (see reposts) current: @superhuman Past: @copy_ai @futurism Watch my videos below 👇
New York City
Joined November 2011
- Tweets4.4K
- Following2.3K
- Followers2.2K
- Likes49.4K
Pinned Tweet
The artist creates markets.
What makes great art is a question too few people in business ask. Much of business operates through commercial incentives and commoditized thinking, both of which are often downstream of what artists first imagine.
Business has also become far too obsessed with next quarter’s results. The artist works on a different horizon, often without knowing whether a market for their work will ever exist.
Edwin Land, the inventor of Polaroid and one of Steve Jobs’ heroes, understood something essential about this:
“The important and nearly impossible projects such as we set for our goal require prolonged periods of intensive concentration.”
The ability to sustain attention gives each individual a chance to create something that previously seemed impossible.
Yet the sheer number of decisions our current technology forces upon us makes this anything but an artist’s paradise.
We have confused convenience with progress.
Technologists offer us the tooling, but the values guiding its design have gone terribly awry. If we’re approaching the renaissance Altman envisions, we ought to ask what kind of human activity our technology is making possible.
So many of our screens demand our attention simply to operate them. We spend our days clicking through business applications, reconstructing context, managing workflows, and performing what increasingly feels like cosplay of productive work.
All this activity, with remarkably little new creation to show for it.
Our values have become splintered, in part because the algorithms organizing our attention reward fragmentation. We’ve optimized for engagement, retention, and the repeated return to the interface.
Perhaps the next great innovation is the recovery and re-founding of the values that should guide our technology in the first place.
And I have a feeling those values begin with protecting human attention.
Because attention is what makes sustained intention, independent judgment, and ultimately creation possible.
Perhaps, in the end, we should measure technology by the meaningful decisions it enables per unit of human attention, rather than the amount of attention it retains.
Decision quality per unit of attention.
Maybe that’s a better metric than retention.
Austin Calvert retweeted
Nothing very fancy/complictaed. They all share an agents.md where 3 things are defined per workspace=agent
1. Sources to watch (eg Slack channels to read)
2. Part of the brain relevant and that they can write to
3. Section of the to do list they own
But for me the workspace=topic structure is important to define where the works lives. If I want to work on legal, I open the legal workspace. It has only relevant context for the running session, its own subagents, etc.
The orchestrator is a task manager and helps me prioritize and launch some of the task to the relevant workspace.
Austin Calvert retweeted
I've changed how I work over the last few weeks.
I now have one lead agent for each area of focus, running all the time, and a chief of staff agent that oversees them. Each lead manages its own sub-agents that write code, test it, chase reviews, and coordinate with team members.
Everything I need to decide or review comes into one prioritized queue, a simple local web app. Each product review includes screenshots and a video demo of the feature. I clear the queue in batches, mostly by voice and keyboard shortcuts. The leads also bring back ideas I hadn't considered.
Working from one queue, instead of switching between dozens of chats, has cut a lot of mental overhead. The system keeps learning my preferences and distilling my judgment into a "Brian brain," so its recommendations and first drafts get better before they reach me. The goal is fewer corrections per change.
Every change still goes through the same human review and security checks. It is eliminating tasks, not human judgment.
What an exciting time to push the frontier of productivity, especially as we build the economy for agents (agentic finance). We should all keep pushing in this direction.
Austin Calvert retweeted
Nikesh Arora's Strategy for Building a $268 Billion Company:
The fastest way to build a giant business is to sell more products to the customers you already have. And the fastest way to get those products is to acquire companies rather than build everything yourself.
Nikesh realized two important things when he became CEO of Palo Alto Networks:
1. The biggest companies sell many products to the same customers. It's much easier to grow revenue by selling additional products to customers who already know and trust you than by constantly finding new customers.
2. Building great products takes 4–7 years. If Palo Alto Networks wanted to expand quickly, it couldn't build every new product from scratch. The solution was to acquire companies that had already spent years developing great products.
Nikesh explains his thinking in this clip. This part is awesome:
Nikesh Arora: If you look at the largest businesses, there's Microsoft's enterprise business or Salesforce, or take the largest businesses, they sell a lot to the same customer.
That was the insight.
I'm like, well, if we’re going to be big at Palo Alto, we've got to find a way of selling a lot to a customer. We only had one thing to sell.
So you go knock on the door, you knock on 20 doors, you spend all the time, and 15 people say, "I'm sorry, I don't need it, I already have it," you're stuck with five.
If you knock on a door, you have a truck behind you saying, "What do you want? Do you want this? You don't want this? You want this? You want this?" It's like, the world's your oyster. What can I get you?
Armed with that insight, I'm like, well, we need to figure out a way that we can find adjacencies to our business, which the customers will trust to buy from us, so that we can keep expanding what we sell to customers, so we can go serve the top 10,000 customers in the world and increase the amount of money you can get for them. You only get that if you earn their trust, you only get that if your shit works, you only get that if it's best in class, because nobody wants to buy the second best product in cybersecurity. It kind of sucks.
So customers want to buy the best which means we’ve got to be the best. Now that's one part of the insight.
The other one of the insight is, which is still true, it takes four to seven years to build a good technology product. It doesn't matter who you are, what you are, where you are.
Whether it's an Uber, it's YouTube, it's Facebook, it's Google search, it took four to seven years, sometimes longer, to build a useful product that the customers found exciting and interesting. That's true in enterprise as well.
So I just joined this company. We don't have a large product pipeline. It's gonna take me four to seven years to build new products. What am I gonna do?
David Senra: Acquire them.
Nikesh Arora: That's right.
Austin Calvert retweeted
Bret Victor on possibilities to build a dynamic medium for human's capabilities.
Austin Calvert retweeted
What is an implicit state machine?
Your apps are full of state machines, which is the logic that says "when we're in this state, and this event happens, now we're in this state" (given, when, then).
Sometimes those state machines are clearly mapped out (switch statements, reducers, tables/maps, libraries like XState for the more complex ones, etc.) and the logic is explicitly enumerable.
And sometimes it's not. The ad-hoc logic is spread all over the place. That's how developers typically code, and that's how agents typically vibe-code (they learned from you). Those are implicit state machines.
What's wrong with implicit state machines? It's not clear what state transitions are possible, so you might have "allowed" states that should have been impossible without knowing it. Agents make the same mistakes.
You can get all tests passing and your code review agents (even the well-funded ones) can say "LGTM" and you can *still* have very bad bugs because an implicit state machine allowed unexpected or illegal behavior. (Remember the Apple FaceTime bug?)
So yeah, making your app logic explicit is important.
Austin Calvert retweeted
Btw the new meta is:
- Fable 5.2 for planning
- Opus 5.5 for execution
Trust me, and thank me later.
Austin Calvert retweeted
22 threads running. Nothing in my inbox. Feels great.
Idk why every other app shoves all the in-progress work in your face.
Austin Calvert retweeted
“Make it exist first, then you can make it good” is one of those nice sayings that’s meant to help you overcome inertia and perfectionism, but following this advice too bluntly is risky. Because once you make something exist, it’s tempting to focus next on making it *grow*. And by then you’re already trapped in the iteration you’ve shipped first, at least for a while.
Austin Calvert retweeted
the sooner we can move past agentic computer use, the better. it’s a hilariously inefficient interim hack. like buying a cattle ranch because you want a burger but Shake Shack isn’t open yet.
Austin Calvert retweeted
Apple really should redo its entire software team with people who are all the way spun up on software factories. Instantly all these legendarily buggy apps like TV or Calendar will be insanely good, redone by teams of a few people and a hundred codegen agents.
Austin Calvert retweeted
frontier models, with lots of test time compute, are becoming the engine of scientific discovery
