Agentic Software EngineeringOverviewConceptual modelCurriculumLearning outcomesAssessmentTooling referenceGlossary
Glossary & references

The vocabulary of the discipline

The terms central to this curriculum, the adopted terms it builds on with attribution, and the literature the course draws on.

Terms central to this curriculum

18 terms
Agentic software engineering (ASE)
The discipline of structured, auditable human-agent workflows for building software. The human frames, specifies, and judges; agents execute. The discipline has its own principles, artefacts, failure modes, and ethics, and carries from one product to the next.
Agentic development environment (ADE)
The operational centre of ASE practice. An ADE determines what tools an agent has, what permissions it operates under, what context it can access, and what transparency the human receives in return.
ASE versus casual prompting
The foundational distinction that motivates the curriculum. A casual prompter gives an agent an instruction and hopes. An agentic software engineer equips the agent with the right problem, specification, context, tools, permissions and limits, and verifies the output against a defined standard before accepting it.
Evolutionary spiral
The synthesis at the heart of the curriculum: a rhythm in which intent and build co-evolve through a sequence of turns, each fast and loose inside but pinned down at named commit points between turns.
Co-evolution of intent and build
The claim that the problem space and the solution space evolve together rather than in sequence. Each step of the build teaches the framer something about what they actually wanted, so intent must be allowed to move.
Commit point
The decision, at the end of a turn of the spiral, about what to lock: what goes back into the specification, what remains open, what is named done.
Drift detection
The judgment skill of recognising when a spiral is failing: going in circles, expanding scope, losing intent, accumulating mess. Paired with commit point.
Trajectory management
The skill of recognising signals during a single agent session that the work is drifting from the task, and applying the corresponding intervention. Operates within one agent invocation; drift detection operates across turns.
Auditability principle
The discipline’s defining commitment: a workflow that cannot be inspected, replayed, or evaluated after the fact is craft, not engineering. Engineering is accountable and reproducible.
Audit trail
The frozen record linking intent to specification, to context, to plan, to execution, to verification. Sufficient that someone other than the original engineer can reconstruct what was decided, why, and what happened.
Verification before trust
The principle that every agent output is a hypothesis until checked. No agent output enters the project without passing through a defined verification gate.
Verification theatre
Verification gates that exist but catch nothing. A named failure mode students learn to recognise in their own and others’ workflows.
Legibility
The property of a system, interface, codebase or document that makes it readable, understandable, and evaluable by someone who did not build it. A control mechanism in agentic engineering.
Minimal footprint
The principle that an agent should have only the permissions required for the task at hand, no more. The engineering response to blast radius.
Blast radius
The scope of damage if an agent with given permissions acts incorrectly. The organising concept for permission design.
Context engineering
The discipline of deliberately curating what the agent encounters in its context window. A first-order engineering activity with its own principles, artefacts (CLAUDE.md, AGENTS.md, lessons-learned files), and failure modes.
Cognified products
Software products in which language-model intelligence is part of the user-facing value, distinct from products built using AI as a development aid.
Agent economics
The cost, latency, and architectural constraints that follow from inference being a metered, billable operation. Model choice, caching strategy, and latency budget shape what the product can promise.

Adopted terms

With attribution
Spiral model
The iterative software process proposed by Boehm (1988), in which risk-driven turns replace the linear waterfall.
Co-evolution (original sense)
From design research: Maher and colleagues (1996) and Cross and Dorst (1999) describe the problem space and the solution space evolving together in creative design.
Rationalism versus empiricism axis
Brooks (2010, Chapter 8): the opposition between believing that careful thought yields a flawless design before any code is written, and believing that humans are fallible and design must proceed by build-evaluate-revise cycles.
Better wrong than vague
Brooks (2010): the practical discipline of making implicit user models explicit, even by guessing, so they can be debated and corrected.
Five-criteria specification
Knuth (1968): an algorithm is defined by finiteness, definiteness, input, output, and effectiveness. The formal benchmark a natural-language specification approximates.
Jagged frontier
The pattern in which AI capability is non-uniform across tasks, vastly outperforming humans on some and failing on others that seem easier. Term in common use following Ethan Mollick.
Human sandwich
The pattern in which humans frame the work at the start, judge it at the end, and the agent does the typing in between. From Klaassen (2026) at Every.
Tacit knowing
Polanyi (1958, 1966): a practitioner knows more than they can tell. The parts of expertise that resist articulation are precisely the parts that resist replacement.
Merge-Readiness Pack (MRP)
Hassan et al. (2025): a structured verification standard with five criteria — functional completeness, sound verification, SE hygiene, rationale and communication, and full auditability.
SASE, and the ACE/AEE workbench split
Hassan et al. (2025): a research vision for agentic software engineering and the architectural distinction between agent-controlled execution environments and human-controlled review environments.
Prompt injection
Malicious content in the agent’s environment embeds instructions that the agent treats as commands rather than as data. The agent has no built-in mechanism for distinguishing instruction from content.
Five workflow patterns
Prompt chaining, routing, parallelization, orchestrator-workers, evaluator-optimizer. Vocabulary established in Anthropic’s engineering writing on agent workflows.
N-version programming
A reliability technique in which the same task is solved by multiple independent implementations whose outputs are compared. Applied here as a counterintuitive use of the parallelization pattern.
Six-pillar ADE architecture
A composite framework for reading any ADE as an instance of six concerns: the bare LLM, tool augmentation, knowledge and memory, learning from experience, multi-agent coordination, and computer use.
Autonomy level framework
A typology of agent autonomy spanning manual coding, task-agentic assistance, goal-agentic assistance, specialized domain autonomy, and general domain autonomy.
AgentOps
The operational practices of running fleets of agents in production: deployment, monitoring, version control of agent definitions, incident response, and the management of permission and policy at the organisational level.

Extended reference list

Foundational works in design, problem-solving, and software engineering

  • Brooks, Frederick P. Jr. (2010). The Design of Design: Essays from a Computer Scientist. Addison-Wesley.
  • Brooks, Frederick P. Jr. (1975; anniversary edition 1995). The Mythical Man-Month: Essays on Software Engineering. Addison-Wesley.
  • Polya, George (1945). How to Solve It: A New Aspect of Mathematical Method. Princeton University Press.
  • Knuth, Donald E. (1968; third edition 1997). The Art of Computer Programming, Volume 1: Fundamental Algorithms. Addison-Wesley.
  • Schön, Donald A. (1983). The Reflective Practitioner: How Professionals Think in Action. Basic Books.
  • Simon, Herbert A. (1969; third edition 1996). The Sciences of the Artificial. MIT Press.
  • Alexander, Christopher (1964). Notes on the Synthesis of Form. Harvard University Press.

Software process models

  • Royce, Winston W. (1970). “Managing the development of large software systems.” Proceedings of IEEE WESCON, 26.
  • Boehm, Barry W. (1988). “A spiral model of software development and enhancement.” IEEE Computer, 21(5), 61–72.
  • Maher, Mary Lou; Poon, Jurgen; Boulanger, Sylvie (1996). “Formalising design exploration as co-evolution.” In Advances in Formal Design Methods for CAD. Springer.
  • Cross, Nigel; Dorst, Kees (1999). “Co-evolution of problem and solution spaces in creative design.” In Computational Models of Creative Design IV, University of Sydney.
  • Raymond, Eric S. (2001). The Cathedral and the Bazaar. O’Reilly Media.
  • Mills, Harlan D. (1987). “Cleanroom software engineering.” IEEE Software, 4(5), 19–25.
  • Parnas, David L. (1979). “Designing software for ease of extension and contraction.” IEEE Transactions on Software Engineering, SE-5(2), 128–138.

Academic research on agentic software engineering

  • Hassan, Ahmed E.; Li, Hao; Lin, Dayi; Adams, Bram; Chen, Tse-Hsun; Kashiwa, Yutaro; Qiu, Dong (2025). “Agentic Software Engineering: Foundational Pillars and a Research Roadmap.” arXiv:2509.06216.
  • Meske, Christian; Hermanns, Tobias; von der Weiden, Esther; Loser, Kai-Uwe; Berger, Thorsten (2025). “Vibe Coding as a Reconfiguration of Intent Mediation in Software Development.” IEEE Access, 13, 213242–213259.
  • Taibi, Davide; Muccini, Henry; Vaidhyanathan, Karthik; Kalinowski, Marcos; et al. (2026). “A Research Agenda on Agents and Software Engineering: Outcomes from the Rio A2SE Seminar.” arXiv:2605.11720.
  • Dong, Tao; Shi, Sherry; Sampath, Harini; Macvean, Andrew (2026). “From Correctness to Collaboration: A Human-Centered Taxonomy of AI Agent Behavior in Software Engineering.” CHI EA ’26.
  • Otoum, Nesreen; Elkhalili, Nuha (2026). “Methods and Techniques of Agentic Software Engineering: A Systematic Literature Review.” IEEE Access, 14.

Empirical research on AI in software work

  • Eloundou, Tyna; Manning, Sam; Mishkin, Pamela; Rock, Daniel (2023). “GPTs are GPTs: An early look at the labor market impact potential of large language models.” arXiv:2303.10130.
  • Anthropic (April 2025). “Anthropic Economic Index: AI’s impact on software development.”
  • Massenkoff, Maxim; McCrory, Peter (March 2026). “Labor market impacts of AI: A new measure and early evidence.” Anthropic.
  • Brynjolfsson, Erik; Chandar, Bharat; Chen, Ruyu (2025). “Canaries in the coal mine? Six facts about the recent employment effects of artificial intelligence.” Stanford Digital Economy.
  • Hampole, Menaka; Papanikolaou, Dimitris; Schmidt, Lawrence D.W.; Seegmiller, Bryan (2025). “Artificial intelligence and the labor market.” NBER Working Paper 33509.
  • Acemoglu, Daron; Autor, David; Hazell, Jonathon; Restrepo, Pascual (2022). “Artificial intelligence and jobs: Evidence from online vacancies.” Journal of Labor Economics, 40(S1), S293–S340.
  • Gans, Joshua S.; Goldfarb, Avi (2025). “O-Ring Automation.” NBER Working Paper 34639.
  • Autor, David H.; Thompson, N. (2025). “Expertise.” NBER Working Paper 33941.
  • Handa, Kunal et al. (2025). “Which economic tasks are performed with AI? Evidence from millions of Claude conversations.”

Practitioner accounts and reference documentation

  • Shipper, Dan (May 2026). “After Automation.” Every.
  • Klaassen, Kieran (April 2026). “You’re the bread in the AI sandwich.” Every.
  • OWASP. “Top 10 for Large Language Model Applications” (2025) and “Top 10 for Agentic Applications” (December 2025).
  • Anthropic. Claude Code documentation. Available at docs.anthropic.com.

Mikael Alemu Gorsky

Mikael Alemu is an educator and researcher, and the author of two programs: Agentic Software Engineering, on building software with AI agents, and Building AI-Native Agentic Systems, on building software that thinks.

He teaches at the Holon Institute of Technology, near Tel Aviv, where Agentic Software Engineering runs as a credit-bearing course. He is an educator and researcher.

Nine published works, 76 citations. A 350-page textbook under contract with a major academic publisher.

Teaching and programs

Agentic Software Engineering — program, preprint and textbook

The discipline of structured, auditable human-agent workflows for building software. The human frames, specifies and judges. The agent executes. Nineteen modules in four parts, built on a running project called Tribunal, a web application in which agents argue opposing sides of a case and a judge agent decides. Taught for credit at the Holon Institute of Technology.

Agentic Software Engineering curriculum

Building AI-Native Agentic Systems — program, paper and book in writing

How to build systems that hold a language model as a working component, and treat that component as what it is: stochastic, slow and metered. Fourteen modules in four parts, about seventy hours. The running project is the Observatory, a news agency that watches sources, selects what matters and publishes on a cadence.

Research and analytics

Publications — journals and proceedings

Nine works, 76 citations. Research on artificial intelligence in education, with Ilya Levin and Alexei Semenov.

The AI Pravda — LinkedIn newsletter

Critical analysis of artificial intelligence and its effect on work and society. 5,500+ subscribers. The complete archive of 103 issues (2023–2026) is published in full at mgorsky.net/theaipravda.

Subscribe to The AI Pravda on LinkedIn

Pro bono

AI for seniors — free workshop

Helping older adults use everyday AI tools. Delivered to Russian-speaking communities in Israel.

For older adults, artificial intelligence is about preserving quality of life, maintaining autonomy, and sustaining the feeling of independence that defines dignified aging. For seniors who have emigrated, AI becomes a bridge: it can translate documents, explain official letters, help compose emails in the local language, and guide users through government websites. The workshop has been delivered to Russian-speaking communities in Israel, where participants — many of them in their 70s and 80s — discovered that AI could help them read Hebrew documents and communicate with Israeli institutions.

Startup competitions — unpaid time

Judging and mentoring early-stage ventures. Helping teams clarify their value proposition, assess technical feasibility, and prepare for the realities of scaling an AI product.

AC/VC LinkedIn group — community

A group for developers and students working with coding agents. The community shares practical insights, code examples, tool comparisons, and honest assessments of what works in production.

Join the AC/VC LinkedIn group

Recent

Important Links