news medium.com

The Grand Demulsification of Software Engineering

Opinion

An AI-generated monograph argues that software engineering is splitting between systems experts and a much larger workforce focused on high-level “glue code.” It matters because the author says LLMs are making routine coding cheaper while exposing weaknesses in the abstractions, processes, and infrastructure that supported mass developer growth. The argument divides software history into four eras: rigorous systems work before 2000, managed runtimes after the dot-com boom, full-stack JavaScript from 2009, and probabilistic AI-generated code today. It claims vendors lowered entry barriers to expand platform markets, while Agile, Kafka, and Kubernetes often masked weak architecture instead of fixing it. According to the monograph, LLMs can reproduce coding patterns but cannot reason about execution, novel constraints, or system design. Developers without deep systems knowledge therefore struggle to audit plausible but incorrect AI-generated code. The piece presents C++ and other low-level technologies as the enduring foundation beneath runtimes, browsers, and AI infrastructure. It predicts that companies focused on rapid feature delivery will automate or shrink application teams, while organizations building compilers, databases, operating systems, and hardware will favor engineers with strong systems theory. The conclusion is that AI is “demulsifying” the industry: routine application work will become increasingly automated, while foundational systems engineering remains indispensable.
  • The monograph frames AI not merely as a productivity tool but as a sorting mechanism that may separate routine application work from systems engineering.
  • Its central forecast depends on the unsupported premise that there are roughly three supernumerary developers for every systems-oriented engineer, making the labor-market conclusions especially consequential.

Build this article’s timeline

Extract dated events, key actors, and open questions using the same timeline analysis available in Signal.

Log in to build

Timeline generation is available to signed-in community members.

Five Ws 🤖 3 AI prefills · 2 still unanswered
Who
Peter Naur, Charles Sanders Peirce, Sun, and Microsoft
🤖 AI prefill
What
An AI-generated monograph argues that software engineering is splitting between systems experts and a much larger workforce focused on high-level “glue code.” It matters because the author says LLMs are making routine coding cheaper while…
🤖 AI prefill
When When did this happen, or what is the timeframe?
Where Where did this occur?
Why
Centers on Grand Demulsification Theory, supernumerary developers, and systems-level engineering
🤖 AI prefill

Login to answer and earn XP for each gap you fill.

Software Labor Evolution developer demographics, exponential engineers, supernumerary developers, dot-com talent shock, vendor commoditization Abstraction And Infrastructure managed runtimes, full-stack JavaScript, Node.js, Agile, Kafka LLM Coding Disruption LLMs, stochastic prompting, code generation, induction, abduction Systems Engineering Resilience C++, systems programming, memory management, concurrency, RAII Corporate Demulsification demulsification, labor economics, corporate sorting, glue code, systems craft
Grand Demulsification Theory99% supernumerary developers97% systems-level engineering96% Large Language Models95% vendor-driven commoditization93% software engineering de-skilling92% stochastic prompting90% full-stack JavaScript89% managed runtimes87% Agile as organizational cover85%
82
Manipulation Score: High How much this article uses emotional framing vs. straight reporting
Loaded Language 9225%
Absolutist 8415%
Threat/Panic 7825%
Us vs Them 8815%
Engagement Bait 6220%
Manipulation score look right?

Factual Claims (5)

medium technology
LLMs are making routine coding cheaper while exposing weaknesses in software abstractions, processes, and infrastructure.
As AI tools collapse the cost of generating high-level glue code, this corporate emulsion is rapidly breaking apart
medium technology
Software history can be divided into four eras: systems work before 2000, managed runtimes after 2000, full-stack JavaScript from 2009, and stochastic prompting from 2022.
The progression of commercial software development is defined by four distinct eras
high economic
Vendors lowered entry barriers to expand platform markets, while Agile, Kafka, and Kubernetes often masked weak architecture rather than fixing it.
software vendors artificially lowered entry barriers, creating a vast 3:1 majority of supernumerary developers
high technology
LLMs reproduce coding patterns but lack a mental model of machine execution and cannot reliably handle novel constraints or system design.
They do not possess a mental model of machine execution.
high economic
Companies focused on rapid feature delivery will shrink or automate application teams, while foundational systems organizations will favor engineers with deep systems expertise.
Companies focused purely on time-to-market feature churn are replacing large supernumerary teams with automated LLM pipelines

Comments

Login to view and post comments.

🏠Portal 📰Links ❓Q&A 📅Events 💼Jobs