What happened to microservices
The pendulum swung back and the useful part is the specific reasons, not the vibe.
tech, developers, and the code underneath
older dispatches from the README archive.
The pendulum swung back and the useful part is the specific reasons, not the vibe.
The analogy is exact except for the part that matters: there is no parameterized query for natural language.
A real relational database in WebAssembly with durable storage. What it makes possible and what it breaks.
Synchronized global demand, hard latency requirements, and no option to shed load. A look at how it actually works.
A language that ships small releases, breaks nothing, and refuses features. Its detractors and its users are describing the same thing.
Treat added dependencies like added latency: a number you spend deliberately, with a ceiling nobody may exceed quietly.
A short guide to the specific ways performance numbers lie, from someone who has published misleading ones.
Java's concurrency change looked incremental and turned out to reshape how services get written. A field report.
Availability percentages are a compliance artifact. Here's what to measure if you want to know whether users are having a bad time.
WebAssembly's most important feature isn't running fast in a browser. It's letting you run someone else's code safely.
CRDTs got good, sync engines got boring, and the offline-capable app stopped being a research project.
Six-week releases, no breakage, and a backlog that is finally shorter than it was. What's left is the hard part.
Datacenter power demand is showing up in residential rates, and the regulatory fight is now the main constraint on compute supply.
The biggest storage-layer change in years is in production now. Here's what it actually did to real workloads.
We spend years learning to write and approximately zero hours learning to read. Then we spend our careers reading.
Harvest-now-decrypt-later makes this urgent for anything with a long confidentiality horizon. The tooling is finally ready.
Producing code got cheap. Knowing whether it's right did not. Everything downstream follows from that asymmetry.
Another year of AI in appliances, and one genuine trend hiding underneath: everything is now thermally constrained.
A convention nobody standardized became standard anyway. Here's what belongs in it and what doesn't.
Not predictions about model capability. Predictions about what your Tuesday looks like.