May 28, 2026
Drift drama hits the fan
You Were Never Declaring State. You Were Observing by Hand
Techies roast big claim that old cloud tools were basically blind note-takers
TLDR: The article says newer automation tools can inspect live systems directly instead of relying on humans to write everything down first. Commenters pushed back hard, saying older tools already did more than the author admits — and several were more offended by the article’s dramatic writing style than its argument.
A spicy essay tried to flip a whole corner of the software world on its head, arguing that for years people weren’t really “describing” their systems to automation tools at all — they were manually peeking at reality, writing notes, and feeding them to software that couldn’t look for itself. The new pitch? Modern agents can now inspect live systems directly, store what they find, and track changes over time, meaning the real design work shifts from writing config files to deciding what should be observed and when action should happen.
But the comments? Absolutely not buying the grand reveal without a fight. One veteran with 20 years in infrastructure basically called the whole thing overhyped and badly framed, saying this kind of visibility already existed and that people only run into these problems when they’re, in their words, “doing it wrong.” Another commenter cut straight to the heart of the claim with a blunt “Huh?”, arguing Terraform — a popular cloud setup tool — already checks the real world during planning, so calling it blind felt wildly off. And then came the pure internet classic: someone said they “couldn’t read past the slop,” turning the thread into less of a technical debate and more of a roast session about bloated AI-style writing.
So yes, there’s a real idea here about software getting better at observing the world on its own. But the community reaction was the true spectacle: part fact-check, part grammar police, part generational eye-roll.
Key Points
- •The article argues that traditional infrastructure-as-code files represented human observation encoded for tools that could not directly inspect the full system.
- •Mark Burgess's promise theory is presented as a model in which autonomous agents should observe, decide, and act locally.
- •The article says newer agent-based methods can inspect live infrastructure directly and store results as typed, versioned, schema-validated data.
- •It uses swamp model methods and the @webframp/aws/adopt extension as examples of discovering and snapshotting existing AWS resources for drift analysis.
- •The article states that declarations are still required for provisioning new resources and for compliance baselines that express intended policy.