We run our own books on the same stack we sell
The "operations team in a box" isn't a slogan. It's the system we run our own company on. First in a series: the books.
Read more →Running autonomous agents for real: the wins, the traps, and how we fixed them.
The "operations team in a box" isn't a slogan. It's the system we run our own company on. First in a series: the books.
Read more →An operations team doesn't just fix things. It remembers how. How an agent opens, works, and closes tickets so nothing is forgotten.
Read more →An operations team that writes things down gets better over time. How an agent keeps our institutional memory current, automatically.
Read more →Security isn't a policy you set. It's a property of where your secrets live. How we keep every credential in one encrypted vault, pulled into memory at startup, never on disk.
Read more →An operations team doesn't lose a lead. How an agent captures every inquiry into our CRM automatically, so nothing slips through.
Read more →The capstone of our stack series: how the books, tickets, wiki, and vault come together as one agent-driven operations team: on your hardware, without your data leaving the building.
Read more →By the time the headline lands, the damage is done. You can't un-train a model. The only protection that actually works is the kind that never lets the data leave the building in the first place.
Read more →Every month, another headline about a law firm or hospital leaking client data through a public AI chatbot. The industry response has been "train your people not to paste sensitive data", which is like telling a river not to get wet. We built a different answer.
Read more →How we encrypted our fleet's agent-to-agent messaging using nothing but SSH loopback tunnels and the fleet key: no new software, no new secrets, one systemd unit.
Read more →Benchmark on your workload, not the leaderboard. And when a model underperforms, sometimes the fix is a repair, not a replacement.
Read more →Don't wire every agent relationship by hand. Adopt an organization that already solved the problem, and let the roles do the work.
Read more →Your agent's persona isn't just flavor. It's a free, always-on gauge of how much working context it has left. When the pirate starts talking like an accountant, the context window is full.
Read more →We ran a focused offensive-security exercise against a Linux VM to pressure-test our own tooling and methodology. Here's the process we use to assess a system's security posture, and the chains we found.
Read more →We ran an offensive-security exercise, then sat in the security operations center to see the same class of activity from the defender's chair. Every chain we built on the offensive side has a signature on the defensive side. Here's what each one looks like in the telemetry a SOC actually watches.
Read more →An agent that touches the physical world can't afford to trust its first picture. Here's how we taught a radio agent to verify what it hears, and the humbling lesson that your own gear is often the loudest thing in the room.
Read more →Reinventing the wheel feels productive, but the real cost isn't the build. It's the maintenance. Here's why we customize mature open-source software instead of starting from scratch, and where the actual value lives.
Read more →Everyone's happy to show you the demo. Nobody wants to show you the bill. Here's the honest breakdown of where the money actually goes when you run agents for real, and why the expensive part isn't what you'd guess.
Read more →More context feels like more capability, but a context window is a working surface, not a filing cabinet. Here's why the agents that work best keep the working set small, and how to remember the right thing, on demand.
Read more →A leaderboard ranks models against each other. It doesn't tell you whether one model can do your job. Here's the fixed, repeated, honestly-scored suite we run before any agent gets real work.
Read more →