Oracle laid off another 30,000 people this week. Let me tell you about Oracle.
I got cut in September 2025. Principal engineers targeted. Exactly one week before my first stock vesting. One week. “It is what it is” became my mantra fast.
It’s code scaffolded on code from the last decade, scaffolded on a 90s-era DSL nobody outside a shrinking circle can read. And they’re firing the shrinking circle. The institutional knowledge walking out the door is irreplaceable. The whole tower exists to generate one thing: predictable support revenue. As long as that number holds, the bleeding is a line item.
And it was all sold — wait for it — as low-code/no-code.
That’s the joke at the center of this entire industry. LCNC was supposed to free us from developers. Instead, it spawned an entire caste of them: SAP Developers, Salesforce Developers, ServiceNow Developers, FuckDay Developers. Whole careers built on the premise that value lives in proprietary languages and internal schemas instead of, you know, engineering.
| Company | Ecosystem | Rant |
|---|---|---|
| SAP | ABAP | The OG in vendor-lock-in, the GOAT pioneer |
| ServiceNow | GlideScript | It’s JavaScript! No. It isn’t. |
| Pega | Pega Rules | Rule sets? You mean Java shoved in custom steps |
| Workday | XpressO | Metadata of Metadata. You’ve seen their ATS, you know. You. Know. |
| MS Dynamics 365 | Dataverse | C# with your hands tied behind your back |
| Epic | The Human Centipede | Hospitals are locked into Epic, which is locked into ObjectScript, which is InterSystems IRIS/MUMPS. No one can leave; everyone eats shit. |
A rant inside the rant: don’t build a DSL. Don’t ship LCNC. Don’t.
You can’t hire for it. The docs are whatever your company wrote on a Tuesday in 2014. There’s no tooling. Testing is a prayer. End-to-end debugging is a séance. Your engineers become polyglot archaeologists, tracing how a change in one layer mutates a DSL in the next, which compiles into a third, forever.
The real cost isn’t the tech debt. It’s the cognitive load. Seventy percent of your engineers’ mental budget goes to reverse-engineering proprietary layers just to figure out what the system is doing. The actual clinical problem — the thing a hospital is paying you millions to solve — gets the leftovers.
Mine looked like this:
- Dozens of clinical thin clients in C++, C#, Java, whatever was in fashion that decade.
- The same capabilities re-implemented across every practice, each library missing the exact thing you needed. Extremely-Programmed™ abstractions stacked on abstractions, each layer fighting the ones above and below. I once found a working solution in another language, in another clinical area, written by a team whose entire PR feedback was “get fucked.”
- An archaic, vendor-locked, custom messaging framework. Of course.
- And underneath all of it, the proprietary DSL. Broken tooling. No support. No docs. Sold with “it’s just like SQL!” — a sentence I will hear on my deathbed.
The plan? Lift all of it — the thin clients, the duplicated libraries, the messaging layer, the DSL — into “modernization” by rewriting it with some AI panel capable of approximately dick.
This was supposed to be an “I told you so” about Oracle, minus the telling. It turned into a list of everything that gives me the ick about enterprise software. I regret nothing.
Oracle is awful. The model is awful. The whole genre is a slow-motion hostage situation dressed up as a product category.
It is what it is.
Leave a Reply