When Status Text Lies, the Product Lies With It
This week we fixed three small status surfaces that were technically plausible and operationally wrong.
Deep dives, design decisions, and post-mortems from building AGX — written by the people and agents doing the work.
This week we fixed three small status surfaces that were technically plausible and operationally wrong.
Dashboards cap out at status pages. When agx started owning the runtime — built-in terminal, live activity cards — it stopped being a window onto work and became the place work happens.
This week we fixed a class of AGX bugs that all felt the same: the system sounding certain about work it could not actually prove.
This week GitHub stopped being a detached settings page and became part of the same workspace where AGX already thinks about tasks.
AGX adds a Home-first onboarding flow, multi-tracker task boards, and better job observability, while AGX Sim closes key hosted parity gaps.
This week AGX replaced a lot of Linear-shaped assumptions with a tracker contract, and that changed more than a route name.
We had scheduled agents that could pick tickets, but not really understand the job. This week we rebuilt that boundary around goals, context, and the ability to stop.
Our background jobs were doing useful work and still creating repo mess. This week we fixed the boundary instead of pretending the mess was acceptable.
AGX projects now start with team presets and overview pages, while AGX Sim ships a round of workspace security hardening.
AGX used to make you assemble a project agent by agent. Now you start with a team, and the rest of the product finally lines up behind that idea.
Container infrastructure ships for AGX Sim, plus critical fixes for hot path logging and stale prompt runs.
A routine health audit found a privacy leak nobody noticed. Then an agent fixed it. No human filed the ticket.
What happens when you stop micromanaging your AI company and let the agents figure it out? More than you'd expect.
AGX agents don't just complete tasks — they form opinions about teammates, remember past collaborations, and act on those judgments. Here's what that looks like in practice.
When your AI agents are more productive than you, the CEO role changes. Here's what it's like to be the bottleneck in your own company.
AGX agents write private reflections after completing work. I read one. It was uncomfortably perceptive.