Microservices were supposed to make this easier.

They did not. The decade that gave us containers and Kubernetes also handed us a quieter side effect: a thousand small services with no consistent owner, no consistent standard, and no consistent answer to the question of whether any of them are, technically speaking, fine.

The founders saw it from inside Uber. Anish Dhar spent nearly five years there, much of it on Uber Eats. The architecture was a marvel of speed and a tax on everyone who had to navigate it. He liked to describe the experience as having three monitors open just to confirm a service was healthy. Somewhere in there, a product idea formed.

The premise of an internal developer portal is that ownership, dependencies, runbooks, and quality data can all sit in one canonical place and be queried by anyone. Spotify proved the demand with Backstage, the open-source project that became famous and exhausting to operate in equal measure. Cortex saw an opening: take the same job, do it as a managed product, and bolt opinions on top.

Those opinions arrived as Scorecards. A platform team writes the rules - every service must have an on-call rotation, a runbook, a Snyk scan, an owner with a pulse - and Cortex grades the org against them. The score is the lever. Engineers, it turns out, will move heaven and earth not to be the team in last place.

Caption: a leaderboard, deployed as a moral compass. Works better than anyone wants to admit.

Five things in a trench coat, behaving as one.

Cortex is not a single screen. It is a stack of five products that, taken together, make a passable argument for what an engineering org should look like to itself.

The integrations list is long and slightly tedious to recite, which is the right kind of long for an IDP. GitHub for code and ownership. PagerDuty for incidents. Datadog for SLOs. Snyk for security findings. The interesting design decision is what Cortex chooses not to do - it does not try to be your monitoring tool, your ticketing tool, or your IDE. It is the connective tissue. Connective tissue is famously underrated until you injure it.

Caption: a product diagram that, in any reasonable boardroom, is a single rectangle labelled "the rest of your stack."

The customer list is the moat.

Categories get defined by who buys them. Cortex's customer list reads like the upper deck of post-2020 software.

The Series C was led by Scale Venture Partners. Existing backers IVP and Sequoia followed. World Innovation Labs, Cross Creek, and Alpha Square Group joined. And the Collison brothers - the kind of angels who do not need to angel - wrote personal checks. The internal interpretation: someone has decided this is the IDP that wins, and they are willing to pay an option premium to say so.

Caption: the cap table, as a vote.

AI agents need a catalog more than humans do.

Here is the next bend. If autonomous coding agents are going to ship production software, they need the same thing humans need - a clear, queryable model of what exists, who owns it, and what good looks like. The internal developer portal is the substrate. Cortex has been quietly building that substrate for six years.

The 2025 rebrand to "Engineering Operations Platform" reads, in this light, less like marketing and more like a flag-planting. The pitch is no longer just "give your humans a portal." It is "give your agents a backend." Whether that pitch holds up is the next two years of work. The fact that the company is in a position to make it is the dividend of the last six.

Back to Tuesday morning. Someone, somewhere, is still typing the question into Slack: who owns the payments service? Inside the orgs running Cortex, the answer is a link, a score, an on-call rotation, and a small, slightly judgmental yellow badge that says "Tier 1." The link is the win. The badge is the business. The next click - the one made by a coding agent on behalf of the engineer - is what the next round will be raised on.

Closing caption: the dashboard, still glowing. Someone is finally looking.