← all systems · see it in The System
Software engineering
Software engineering is a branch of both computer science and engineering focused on designing, developing, testing, and maintaining software applications (src:wp:software-engineering); 'programming integrated over time' (Software Engineering at Google, as quoted there). Whether it is engineering at all is contested (C-2260, C-3228).
Properties published scienceour proposal
Sourced fields, and our own readings.
| born at | 1965-1968: the term in print (June 1965 Computers and Automation; August 1966 CACM) and the NATO conference, Garmisch, October 1968 |
| born precision | a few years; contested (C-3227) |
| boundary | Inside: the people, practices, records and processes by which software is specified, designed, built, tested, released and maintained (requirements, design, construction, testing, version control, process and management). Outside: the software product once running (sys:software), the machines (sys:computer), the users and domain it serves, the theory of computing (sys:computer-science). |
| nested level | A branch of both computer science and engineering (rung 12), carried by developers in organisations; it runs on people and organisations and produces software. |
| scale | social |
Measures (19) published science
| OS/360 release interval | 90 days |
| Agile iteration (sprint) length | 1-4 weeks |
| Errors corrected in one OS/360 release | 1074 errors (plus 16 changes) |
| OS/360 corrections rate between versions | over 11 corrections per day |
| OS/360 development effort | at least 5000 man-years |
| Maintenance share of project cost | 40-80 % |
| Communication paths in a team of n people | about n^2/2 paths |
| Professional software engineers worldwide | 26.9 million (2022); 21 million (2016) people |
| US software developers employed (BLS 2018) | 1,365,500 people |
| Standish 1994 project outcomes (contested) | 16 success / 53 challenged / 31 failed % |
| IT projects in the cost-overrun sample | 5,392 projects; overruns follow a power law, no calculable mean |
| Annual US cost of software bugs (NIST 2002) | 59.5 billion US dollars per year |
| Therac-25 overdose relative to normal | up to 250 times |
| Change in task time with early-2025 AI tools (experienced open-source developers) | +19 (forecast -24, felt afterwards -20) % completion time |
| Copilot trial: speed-up on an HTTP server task | 55.8 (95% CI 21-89) % faster |
| Causes of developer unhappiness identified | 219 (from 1,318 complete responses of 2,220 developers) factors |
| Group size range of superlinear open-source productivity | 5 to a few hundred developers |
| Scientists writing software who are primarily self-taught | 90 or more % |
| Languages with a significant defect association after reproduction | 4 (of 11 claimed) languages |
Open contentions (9) open contention
Where sources or readings disagree. Both stay on the map.
- C-413 Where do thoughts come from?
booked: No theory is proven, so the map lists the main ones side by side and doesn't pick one. It shows where thinking is handled in the brain, not where thoughts start.
the other reading: positions as listed under C-413 - C-1509 Do organisations shape what they build, or the reverse?
booked: The map books the match between the two, and keeps all three directions open.
the other reading: organisation shapes design; design shapes organisation; both ways - C-2260 Is computer science a science?
booked: The map keeps the readings: science, mathematics, engineering or art.
the other reading: Dijkstra: 'the doomed discipline', 'How to program if you cannot'; Knuth: programming as art and science - C-3176 Do AI coding tools make developers faster or slower?
booked: The map keeps the competing readings; none is ruled.
the other reading: effect depends on task, codebase maturity and developer familiarity; not ruled - C-3214 Does the programming language affect how many bugs code has?
booked: The map keeps the competing readings; none is ruled.
the other reading: the original GitHub study's 'small but significant relationship between language class and defects'; NATO 1968: high-level languages credited with fewer bugs - C-3227 When was software engineering born?
booked: The map keeps the competing readings; none is ruled.
the other reading: phased development presented by Benington in 1956 for SAGE; Hamilton naming the discipline during Apollo - C-3228 Is software engineering really engineering?
booked: The map keeps the competing readings; none is ruled.
the other reading: not engineering but should be (McConnell); a misnomer for 'radical novelty' (Dijkstra); licensing: NCEES exam 2013-2019 ended for lack of participation; ACM and IEEE judged licensing inappropriate in the 1990s - C-3233 How often do software projects fail?
booked: The map keeps the competing readings; none is ruled.
the other reading: Eveleens and Verhoef 2010: the definitions are misleading, one-sided and give meaningless figures; Flyvbjerg 2022: overruns follow a power law with no calculable mean (a different measure: risk in the tail) - C-3236 Do bigger teams get less done per person?
booked: The map keeps the competing readings; none is ruled.
the other reading: the two may describe different settings (late commercial projects vs open-source communities); not ruled
Partners and parts published science
Built from / builds (8)
- Programmers and language communities
- Code review
- Establishing design requirements (problem definition)
- Software design
- Software testing
- Version control
- Software development process
- Software maintenance
Sits inside / holds (3)
- Engineering (design)
- Organization
- Computer science (the field: the practice and the computation it studies)
Trades with (valence) (12)
- Human (Homo sapiens) via sys:sensory-environment · Software engineering gain
- Science (as a human social and cultural system) via sys:written-record · Software engineering gain
- Engineering (design) via sys:written-record · Software engineering gain
- Software via sys:written-record · Software engineering gain
- Large language model via sys:written-record · Software engineering mixed
- Human society (from local community up to societies) via sys:human-society · Software engineering gain
- Computer via sys:computer-hardware · Software engineering mixed
- Market and trade via sys:human-society · Software engineering mixed
- Programming language via sys:written-record · Software engineering gain
- Open-source software via sys:written-record · Software engineering gain
- Organization via sys:human-society · Software engineering mixed
- Computer science (the field: the practice and the computation it studies) via sys:written-record · Software engineering gain
expressed by (22)
- Generation/Synthesis
- Recombination/Composition
- Scaffolding/Enablement (Meta-Tooling)
- Templating / Standardization
- Quality Improvement/Refinement
- Interoperability/Transduction (Bridging)
- Encapsulation & Modularity (Abstraction)
- Symbiosis/Co-Creation (Collaborative Specialization)
- Evolvability/Plasticity Engineering (Design for Future Variation)
- Internal Coordination/Regulation
- Sensing/Scanning
- Hypothesis/Goal Selection
- Mapping/Modeling
- Uncertainty Reduction/Learning
- Risk-managed Exploration
- Active Experimentation / Probing (Optimal Experimental Design)
- Trail-Marking / Externalized Memory (Stigmergy)
- Social Learning / Public Information Use
- Self-Monitoring
- Validation of Effectiveness
- Triage & Prioritized Allocation
- Quality Control & Error Correction
floor filled by (17)
- Human (Homo sapiens)
- Written record (durable external marks: texts, archives, canon)
- Software
- Large language model
- Human society (from local community up to societies)
- Computer
- Programming language
- Organization
- Programmers and language communities
- Code review
- Establishing design requirements (problem definition)
- Software design
- Software testing
- Version control
- Software development process
- Software maintenance
- Computer science (the field: the practice and the computation it studies)
Kinds (4)
- Engineering (design)
- Computer science (the field: the practice and the computation it studies)
- Secure software development
- Professional (software engineering) ethics
Notes published science
- In a 2025 randomised trial, 16 experienced open-source developers took 19% longer to finish real tasks on their own projects when allowed AI tools, yet afterwards they believed the AI had made them 20% faster (src:becker-2025-metr). A 2023 trial on one fresh task found the opposite: 55.8% faster (src:peng-2023-copilot).
- The famous 'waterfall' picture comes from a 1970 paper whose author called that single-pass plan 'risky and invites failure' and argued for feedback and doing the job twice (src:royce-1970).
- Reading code does not light up the brain's language network: an fMRI study found code comprehension runs mainly on the fronto-parietal 'multiple demand' system used for maths and logic, for both Python and the children's block language ScratchJr (src:ivanova-2020-code).
Layers
The ways it takes in its world, where they meet, and how it acts back, each from L1 (where it meets the world) up.
building and keeping software over time, as software engineers describe it: from the world the software serves, through requirements, records, design, construction and process, to software released into use and evolved (reader's view)
- L1: The world the software serves and runs in: users and their activities, the machines, and the part of the world an E-type program mechanizes
- L2: Intake: requirements elicitation, tests run, failures and change requests reported from the field
- L3: Records: requirements specifications and backlogs, source code under version control, issue trackers, documentation
- L4: Deciding what to build or change: requirements analysis and design (interface, architectural, detailed)
- L5: Construction and verification: coding, code review, unit and integration testing, debugging
- L6: Process and coordination: the methodology (waterfall, spiral, agile), project management, estimation and the team's communication
- L7: Release and evolution: software delivered into use and kept fit through maintenance and continuing change; the discipline's purpose, reliable software obtained economically
the boardroom (shared floors)
- L4: Deciding what to build or change: requirements analysis and design at interface, architectural and detailed levels
- L5: Construction and verification: coding, review, testing, integration and debugging
- L6: Process and coordination: methodology, project management, estimation, the team's communication structure
- L7: Release and evolution: software delivered into use and continually adapted; the discipline's purpose, reliable software obtained economically
way in: AI assistants: code suggested by language models
- L1: Code suggested by an AI pair programmer
- L2: Suggestions accepted, edited or rejected in the editor
way in: Code read: source text seen on screens
- L1: Source text shown on a screen
- L2: The developer's eyes (Human's sight channel)
- L3: Program comprehension in the brain's multiple demand system, not the language system
way in: Emotion: happiness, unhappiness and being stuck
- L1: Work events: being stuck, time pressure, bad code, broken code
- L2: Unhappiness measured with a validated instrument
- L3: Affect perceived to raise or lower productivity
way in: Field: failures and change requests from software in use
- L1: Software in use, in an environment that keeps changing
- L2: Errors reported by users and installations
- L3: Errors and requests queued for the next release
way in: History: the repository's record of past changes
- L1: Source files and other hand-made files
- L2: The version control system records each revision
- L3: The history graph: trunk, branches and merges
- L4: Comparing and restoring past versions while changing code
way in: Literature and training: methods, bodies of knowledge, research
- L1: Written methods, standards and research
- L2: Education: degrees and training
- L3: The body of knowledge a graduate engineer is expected to master
way in: Money and schedule: budgets, estimates, deadlines
- L1: Funding and contracts
- L2: Estimates of cost, time and functionality
- L3: Overrun recorded against the estimate
- L4: The economic constraints among the requirements
- L5: Measuring progress
way in: Needs: what users, customers and the domain require
- L1: Users, customers and the part of the world the program mechanizes
- L2: Elicitation: interviews and customer collaboration
- L3: Requirements specified: functional, non-functional and domain requirements
way in: Review: colleagues' comments on a change
- L1: A proposed change and colleagues' written comments
- L2: Comments on defects, design and style
- L3: Understanding the change and its context
- L4: Knowledge transfer and team awareness
way in: Team talk: what everyone else is doing
- L1: Speech and messages between team members
- L2: Meetings: the daily stand-up's coded questions
- L3: Keeping in sync: knowing what everyone else is doing
- L4: Communication paths: about half the square of the number of people
- L5: The organisation's communication structure, which resembles its administrative structure where protocol restricts it
way in: Tests: the program run against expectations
- L1: The program executed on a machine
- L2: Test cases and harnesses compare behaviour with expectation
- L3: Failures and regressions recorded
- L4: Debugging: locating the cause of a failure
way in: Thought: the problem held in mind
- L1: The problem and the program held in one mind, within the limited size of one's skull (where ideas come from: contested, C-413)
- L4: Abstraction and intellectually manageable programs, within the limited size of one's skull
way out: Documenting and specifying
- L1: Documents and specifications
- L4: A design to hand on
way out: Publishing methods and bodies of knowledge
- L1: Published manifesto, standards and papers
- L7: Ways of developing software worked out in practice
way out: Releasing: software delivered into use
- L1: Running software among its users
- L6: Releases on a cadence (OS/360: every 90 days)
- L7: Early and continuous delivery of valuable software
way out: Talking: stand-ups, reviews and design discussion
- L1: Speech and messages
- L6: Information to share with the team
way out: Writing code: typed changes committed to the repository
- L1: Source text in files
- L2: Typing at a keyboard (Human's movement outlet)
- L3: Incremental changes committed as revisions
- L5: A change designed and ready to make
Sources (33) published science
Waterfall model (Wikipedia, Wikimedia Enterprise snapshot 2026-11-01 via tools/wme) (behind its links)
Naur & Randell (eds.) 1969, Software Engineering: Report on a conference sponsored by the NATO Science Committee, Garmisch, 7-11 October 1968 (behind its links)
Lehman's laws of software evolution (encyclopedia) (behind its links)
Cunningham 1992, The WyCash Portfolio Management System, OOPSLA '92 experience report (behind its links)
Beck K et al. (2001) Principles behind the Agile Manifesto (behind its links)
Conway ME (1968) How Do Committees Invent? Datamation, April 1968 (author's reprint) (behind its links)
Software crisis (encyclopedia) (behind its links)
Bacchelli A, Bird C (2013) Expectations, Outcomes, and Challenges of Modern Code Review. ICSE 2013 (authors' copy) (behind its links)
Computer science (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Agile software development (Wikipedia, Wikimedia Enterprise snapshot 2026-11-01 via tools/wme) (behind its links)
Eveleens JL, Verhoef C (2010) The Rise and Fall of the Chaos Report Figures. IEEE Software 27(1):30-36 (authors' copy) (behind its links)
Software testing (encyclopedia) (behind its links)
Therac-25 (encyclopedia) (behind its links)
Becker J, Rush N, Barnes B, Rein D (2025) Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity. arXiv:2507.09089 (behind its links)
Peng S, Kalliamvakou E, Cihon P, Demirer M (2023) The Impact of AI on Developer Productivity: Evidence from GitHub Copilot. arXiv:2302.06590 (behind its links)
Graziotin D, Fagerholm F, Wang X, Abrahamsson P (2017) On the Unhappiness of Software Developers. EASE '17 (arXiv:1703.04993) (behind its links)
Wilson G et al. (2014) Best Practices for Scientific Computing. PLoS Biol 12(1):e1001745, doi:10.1371/journal.pbio.1001745 (behind its links)
Berger ED, Hollenbeck C, Maj P, Vitek O, Vitek J 2019. On the impact of programming languages on code quality: a reproduction study. arXiv:1901.10220 (TOPLAS) (behind its links)
Version control (encyclopedia) (behind its links)
Brooks, No Silver Bullet: Essence and Accidents of Software Engineering, UNC technical report TR86-020 (1986) (behind its links)
Ivanova et al. 2020, Comprehension of computer code relies primarily on domain-general executive brain regions, eLife 9:e58906, doi:10.7554/eLife.58906 (behind its links)
Dijkstra EW (1972) The Humble Programmer (EWD340), ACM Turing Award Lecture, Communications of the ACM 15(10) (transcription) (behind its links)
Raymond ES (1998) The cathedral and the bazaar. First Monday 3(3). doi:10.5210/fm.v3i2.578 (behind its links)
Software development process (Wikipedia, Wikimedia Enterprise snapshot 2026-11-01 via tools/wme) (behind its links)
Royce WW (1970) Managing the development of large software systems. Proceedings, IEEE WESCON, August 1970 (behind its links)
Beck K et al. (2001) Manifesto for Agile Software Development (behind its links)