'the world should work this way' and 'i need this' are very different things.
you can spend years on the first, wondering why nobody gets it.
meanwhile, someone solving the second gets to decide how things work.
consumer expectations for apps are just absurdly high by default.
it’s no longer enough to make the core functionality work. users expect every edge case, interaction, and preference to be accounted for.
as well seen from million settings in codex.
the baseline for what feels like a complete product keeps rising, and it’s getting harder to keep up with the details unless you are a big team.
stuff like that needs to be all human cross checked.
maybe one reason why openai attempts to overwhelm human entity (like me) with million setting is to make me give up making apps and just build plugins..
figma is becoming an interface tax on design.
figma forces design intent through an intermediate representation that ai may no longer need.
makes more sense for figma to launch instrument panel to tune individual components visibly.
@sebonomics everyone wishes they started earlier, but they’re imagining their younger self with their current capabilities. some sort of survivorship bias mixed with hindsight bias.
a lot of consumer demand might be latent precisely because the cost of expressing and fulfilling it has historically been too high.
if agents collapse that friction, ai doesn’t just make supply cheaper. it makes previously uneconomic demand worth serving.
the opportunity might be less about finding new desires and more about making existing ones economically addressable.
@brandon_galang more likely the bottleneck is converting attention into actual demand.
even creators rarely capture most of their value from content itself. they monetize through products, services, and commerce.
content drives discovery. owning the transaction captures the value.
increasingly thinking of the agent stack as:
model → harness → (adapters) → UI
each layer has its own competitors and potential moats. you can already point to leading players at each layer.
for personal agents, harnesses are still challenging, but they’re becoming capable enough to support increasingly complex workflows and tool execution.
as components of this stack become more interchangeable, consumer differentiation will increasingly come down to user experience, and finding the right abstractions for workflows people actually care about.
@aksenHQ this is exactly the trap.
normal people don’t want to manage files or review diffs. most users don’t need huge parts of oldie abstraction.
they want the work done.
codex’s moat vs other coding agent is the interface.
it gets the abstraction right: one task per tab, tools like the terminal available when you need them, side chats without losing your place. it makes delegation feel possible and reliable.
but files, commands, and diffs are a developer’s way of understanding work.
pas need a shared execution layer equivalent to what codex did with interfaces shaped around the work. still feels wide open.
really cant believe how could slightly prettier chat gpt could raise 40m?
some sleek visual effect is not sufficient. would have to show some new kind of mental model beyond recents that consumers would adopt.
It's time to get out of stealth 👋
Today, we are launching @monogram_ai and announcing our $40m seed round led by DST and Lux Capital.
Monogram is the first AI app that was built around a visual interface, from the ground up.
We created a technology that generates an entire user interface on the fly, in just a few seconds. Ask anything, and instead of staring at a wall of text, you get an interactive visual response.
@robbiepetersen_ but there is a remaining question of what would the produced output from agents look like?
would that also eventually converge all into some dopamine inducing interface?