Nothing exists, and then you write something, and then it does something. That is the core of it. Software engineering builds working systems out of pure logic — no raw materials, no physical constraints, just a long chain of "if this, then that" that, once written, runs on its own and keeps running when nobody is watching. A thing you build can be used by one person or a billion, and the marginal cost of the billionth copy is essentially zero. No other field has quite that property, and it is the reason a single person at a kitchen table can build something that reaches the whole world.
It is tempting to call the structural pull Creation, and Creation is genuinely half the life of the work. But the honest version of the field, the one a 17-year-old needs to hear, is that the dominant daily pull is Resolution. Software does not stay built. It breaks — constantly, in ways nobody predicted, across layers of systems other people wrote years ago and documented in a hurry. The running joke in the field is that you spend twenty percent of your time writing new code and eighty percent figuring out why existing code does not do what it claims. The defining act, the thing that fills most working hours and separates the people who last from the people who burn out, is the move from broken to functional: tracing a failure through a system you only partly understand, forming a hypothesis, testing it, and making the thing work again. The people who thrive are the ones who find the debugging as satisfying as the building, who like the puzzle as much as the product.
🦊
There's a guide here if you want one
Kitsune can talk through anything on this page — whether it might suit you, what to do next, questions this page doesn't answer. Everything here is yours to read either way.
Technical skill barrier (vibe coding to agentic engineering)
Current fact
The barrier
Programming skill required years of learning; domain experts excluded from building
What changed
AI coding tools translating natural language to functional code; vibe coding market $4.7B growing 38% annually
Behaviours involved
Rapid idea-to-test cycles in hours not weeks
PrototypingSolving
Designers implementing designs without developer handoff
BuildingDesigning
What this is based on
63% vibe coding users are non-developers
Lovable $400M ARR
Cursor $2B ARR, 7M MAU
SaaS build cost $200K→$5K
44% of profitable SaaS solo-founded
How this could age
Low risk on direction; moderate risk on anyone-can-build-anything framing
What it does not cover
AI code has 1.7x more issues, 2.74x more security vulnerabilities. 45% contains security flaws. Works for simple-to-moderate; complex systems still need engineering discipline.
Assessed June 2026
Scale constraint on small teams
Current fact
The barrier
Production-scale software required large specialised teams
What changed
AI makes individuals productive across the full stack; $40/mo common stack enables solo operation
Behaviours involved
Solo builder reaching large audiences
Building
Solo founder builds entire technical product
Founding
Full-stack generalist breadth becomes superpower with AI handling depth
DesigningEngineeringSolving
What this is based on
55% faster task completion (GitHub)
$40/mo Cursor+Claude Code stack
Solo SaaS founders competing with 10x teams
How this could age
Low risk
Assessed June 2026
Testing and quality assurance barrier
Current fact
The barrier
Comprehensive testing was expensive and tedious; most teams under-tested
What changed
AI-generated test suites, automated security scanning, AI-powered code review
Behaviours involved
Quality engineer shifts from writing tests to judging what to test
Pattern-findingTesting
Security-conscious developer runs deep analysis previously prohibitively expensive
SecuringSolving
What this is based on
50% faster unit test generation
Snyk/SonarQube AI becoming CI/CD standard
How this could age
Low risk
What it does not cover
Paradox: AI code has more bugs (60% silent logic failures), so more testing needed even as test generation automated. Net effect on QA roles ambiguous.
Assessed June 2026
Knowledge-access barrier for junior developers
Current fact
The barrier
Learning codebases and debugging required mentors or painful trial-and-error
What changed
AI code explanation, error tracing, always-available tutoring; 34% of Claude.ai conversations are computer/mathematical tasks
Behaviours involved
Accelerated learner productive from week one; learning-through-doing rather than before-doing
Building
What this is based on
25% speed increase for new-to-codebase developers
Early-career developers heaviest AI explanation users
How this could age
Low risk on learning; moderate risk on assuming learning solves hiring problem
What it does not cover
STRUCTURAL PARADOX: AI makes juniors more productive AND reduces junior hiring demand. Employment for ages 22-25 dropped ~20% from 2022 peaks. Both true simultaneously.
Assessed June 2026
Jobs that did not exist five years ago
These are real jobs that exist now and did not exist before the current wave of AI.
New title
AI/ML Platform Engineer
Builds and maintains AI model training, deployment, and monitoring infrastructure. Bridges DevOps and ML Engineering.
340% growth in job postings requiring AI coding tool experience (Jan 2025-Jan 2026) · current fact
New title
AI Code Reviewer / Quality Engineer
Evaluates AI-generated code for correctness, security, and maintainability at scale.
45% AI-code vulnerability rate; 71% refuse merge without manual review · current fact
Familiar title, new shape
Agentic Engineering Lead
Senior engineer directing AI agents, reviewing output, making architectural decisions. 79% of Claude Code conversations are automation — this is the emerging interaction model.
Anthropic Economic Index 79% automation rate; SWE-bench performance levels enabling supervised workflows · current fact
Familiar title, new shape
AI-Augmented Solo Founder Developer
One person builds entire product. SaaS build costs $200K→$5K. 44% profitable SaaS solo-founded.
Stripe 2024 data, vibe coding platform economics · current fact
New title
Vibe Coder / Non-Developer Builder
Domain expert building functional software without programming background. 63% of vibe coding users.
Vibe coding platform user demographics, Lovable/Cursor user data · current fact
New role titles in SWE form faster than in regulated fields because there is no credential infrastructure to slow them. The configurations are already deployed at scale.
Bars above the line are the parts of this work that still need a person. Bars below it are what AI can already do. Tap any column to see the actual work behind it.
high ground · holds stronglydeep water · reaches furthest
yours, by strengthAI reach, by depth
The honest read. The robustly-human dimensions are concentrated at senior levels. Junior developers have thinner protection — their productive contributions (boilerplate, test writing, documentation) are directly in the high-exposure zone. This structural feature is what makes SWE entry-level employment the most disrupted in the corpus.
Software engineering is unique: the primary medium of work (code/text) is exactly the medium AI is most capable in. This creates a higher ceiling of AI relevance than any other professional field. The capability-to-deployment gap is narrower here than anywhere else. 79% of Claude Code conversations classified as automation, not augmentation — the highest automation share of any tool measured.
How AI is changing the way in
Much harder to enter than it was, and that applies fairly evenly across the ways in.
That is everything we currently know about AI in Software Engineering. It shows where things are moving so you can choose which way in suits you.
People drawn to Software Engineering are often drawn to these. Most sit in a different part of the terrain.