Ivan Stepanok — ArticlesField notes on AI automation, agent pipelines and autonomous software development.https://stepanok.com/enClaude Code vs Codex: why I stopped choosing and use bothhttps://stepanok.com/articles/claude-code-vs-codex/https://stepanok.com/articles/claude-code-vs-codex/Claude Code and Codex compared on hooks, sandboxing, headless runs, review and billing, and why the best setup uses one to write and the other to review.Fri, 06 Nov 2026 00:00:00 GMTClaude Code hooks examples: turning an assistant into a pipelinehttps://stepanok.com/articles/claude-code-hooks-pipeline/https://stepanok.com/articles/claude-code-hooks-pipeline/Practical Claude Code hooks examples: block risky commands, gate Stop on tests, call a second model to review, and the hook mistakes that break a pipeline.Wed, 04 Nov 2026 00:00:00 GMTWill AI replace software engineers? The work that moves, and what growshttps://stepanok.com/articles/will-ai-replace-software-engineers/https://stepanok.com/articles/will-ai-replace-software-engineers/What 2026 labour data says about AI and software jobs: BLS projections, Indeed postings, Stanford's early-career findings, and which engineering work grows.Mon, 02 Nov 2026 00:00:00 GMT"Would you like me to continue?" Why coding agents stop halfwayhttps://stepanok.com/articles/agents-that-stop-halfway/https://stepanok.com/articles/agents-that-stop-halfway/Coding agents finish a third of the task and ask permission to go on. Why that happens, and how stop hooks, checkable criteria and a reviewer end it.Thu, 29 Oct 2026 00:00:00 GMTThe true cost of AI coding: tokens are the small line on the billhttps://stepanok.com/articles/true-cost-of-ai-coding/https://stepanok.com/articles/true-cost-of-ai-coding/What AI coding really costs: subscriptions and tokens, plus review time, rework and long agent runs. A cost model and how to cut the parts that matter.Tue, 27 Oct 2026 00:00:00 GMTSpec-driven development with Claude Code: write intent before codehttps://stepanok.com/articles/spec-driven-development-with-ai/https://stepanok.com/articles/spec-driven-development-with-ai/Spec-driven development gives AI agents a spec, a plan and checkable criteria before any code. How to do it with Claude Code and what a good spec contains.Fri, 23 Oct 2026 00:00:00 GMTVibe coding vs agentic engineering: same models, different resultshttps://stepanok.com/articles/vibe-coding-vs-agentic-engineering/https://stepanok.com/articles/vibe-coding-vs-agentic-engineering/Vibe coding and agentic engineering use the same AI models. The difference is specs, checks and review. When each one fits, and how to move up.Tue, 20 Oct 2026 00:00:00 GMTWhy developers don't trust AI code, and why they're right tohttps://stepanok.com/articles/why-developers-dont-trust-ai-code/https://stepanok.com/articles/why-developers-dont-trust-ai-code/Trust in AI coding tools fell while use kept rising. Why the distrust is rational, and how to build trust from evidence instead of from the model.Fri, 16 Oct 2026 00:00:00 GMTRunning coding agents overnight: what must be true before you sleephttps://stepanok.com/articles/running-ai-agents-overnight/https://stepanok.com/articles/running-ai-agents-overnight/An overnight AI agent can turn eight idle hours into finished work, or into a mess and a bill. The checklist for unattended coding agents that actually works.Tue, 13 Oct 2026 00:00:00 GMTWhy AI agents fail in production: nine failure modes and their fixeshttps://stepanok.com/articles/why-ai-agents-fail-in-production/https://stepanok.com/articles/why-ai-agents-fail-in-production/The demo worked, production didn't. Nine ways AI agents fail once they meet real work, what causes each one, and the fix that actually holds up.Fri, 09 Oct 2026 00:00:00 GMTGuardrails for AI coding agents: permissions, sandboxes, budgets, stop ruleshttps://stepanok.com/articles/guardrails-for-ai-coding-agents/https://stepanok.com/articles/guardrails-for-ai-coding-agents/How to give a coding agent autonomy without the risk: permissions, sandboxes, budgets, stop rules and a trust ladder built on reversibility and blast radius.Tue, 06 Oct 2026 00:00:00 GMTThe dark software factory: levels of autonomy, autocomplete to lights-outhttps://stepanok.com/articles/dark-factory-levels-of-autonomy/https://stepanok.com/articles/dark-factory-levels-of-autonomy/What a dark factory in software development means, Dan Shapiro's six levels of AI coding autonomy, and what has to be true to move from one level to the next.Fri, 02 Oct 2026 00:00:00 GMTCan you take people out of code review? A staged path to automationhttps://stepanok.com/articles/remove-humans-from-code-review/https://stepanok.com/articles/remove-humans-from-code-review/A five-level path from humans reading every pull request to review with no person in the loop, with the conditions that must hold before each step.Tue, 29 Sep 2026 00:00:00 GMTThe writer can't be the reviewer: why a second AI model catches morehttps://stepanok.com/articles/independent-reviewer-model/https://stepanok.com/articles/independent-reviewer-model/Why a coding agent reviewing its own work fails, what correlated errors and self-preference mean for multi-agent workflows, and how to set up a reviewer.Tue, 22 Sep 2026 00:00:00 GMTAI coding agents benchmark: why you should run your own evaluationhttps://stepanok.com/articles/evaluate-ai-coding-agents-on-your-tasks/https://stepanok.com/articles/evaluate-ai-coding-agents-on-your-tasks/SWE-bench scores won't tell you which AI coding agent fits your codebase. How to build a small evaluation from your own tickets, and what to measure.Thu, 17 Sep 2026 00:00:00 GMTCode review is the new bottleneck: why AI hasn't sped up shippinghttps://stepanok.com/articles/code-review-is-the-new-bottleneck/https://stepanok.com/articles/code-review-is-the-new-bottleneck/Teams using AI heavily merge 98% more pull requests, but review time per PR grew 91%. Why review became the constraint, and what actually moves it.Tue, 15 Sep 2026 00:00:00 GMTClaude Code writes, Codex reviews: what an independent review loop changedhttps://stepanok.com/articles/claude-code-codex-review-loop/https://stepanok.com/articles/claude-code-codex-review-loop/The same game-rebuild task given to Claude Code alone and to a two-model pipeline. 3.7 hours became 1.5, and 389 M tokens became 34 M. Here's why.Thu, 10 Sep 2026 00:00:00 GMT"All tests pass" is a claim. Make your coding agent show evidencehttps://stepanok.com/articles/evidence-over-claims/https://stepanok.com/articles/evidence-over-claims/Coding agents report success they haven't earned, and some edit the tests. How to build a review pipeline that trusts exit codes the harness ran, not summaries.Wed, 19 Aug 2026 00:00:00 GMTWhat an AI automation engineer does, and how to hire onehttps://stepanok.com/articles/hire-an-ai-automation-engineer/https://stepanok.com/articles/hire-an-ai-automation-engineer/A buyer's guide to hiring an AI automation engineer: what the role covers, what an engagement looks like, interview questions that work, and red flags.Wed, 05 Aug 2026 00:00:00 GMTAI for iOS development: what agents do well on Apple platformshttps://stepanok.com/articles/ai-for-ios-development/https://stepanok.com/articles/ai-for-ios-development/AI agents for iOS development in 2026: what works in Swift and Xcode, what still doesn't, and how to give an agent a build, simulator and screenshot loop.Wed, 22 Jul 2026 00:00:00 GMTAI code gets copied, not refactored. What GitClear's data showshttps://stepanok.com/articles/ai-generated-code-quality-duplication/https://stepanok.com/articles/ai-generated-code-quality-duplication/GitClear's data shows refactoring falling and copy-paste rising as AI writes more code. Why agents duplicate, what it costs, and how to make them reuse.Wed, 08 Jul 2026 00:00:00 GMTFrom nine developers to one: a product team when agents write the codehttps://stepanok.com/articles/from-nine-developers-to-one/https://stepanok.com/articles/from-nine-developers-to-one/A reasoned model of how a nine-person product team can shrink to one engineer on that product while eight move to new ones, and what has to be true first.Wed, 10 Jun 2026 00:00:00 GMTAI-written tests that catch bugs, and the ones that only passhttps://stepanok.com/articles/ai-test-generation-that-catches-bugs/https://stepanok.com/articles/ai-test-generation-that-catches-bugs/Coverage tells you which code ran, not which bugs a test would catch. How Meta and Google filter generated tests, and a pipeline for AI tests that find bugs.Wed, 27 May 2026 00:00:00 GMTAcceptance criteria an AI agent can't talk its way aroundhttps://stepanok.com/articles/acceptance-criteria-for-ai-agents/https://stepanok.com/articles/acceptance-criteria-for-ai-agents/How to write acceptance criteria for AI coding agents: observable, executable, owned outside the agent, with an honest way out. Bad and good examples included.Wed, 29 Apr 2026 00:00:00 GMTLLM-as-a-judge for code: when it works and when it lieshttps://stepanok.com/articles/llm-as-a-judge-for-code/https://stepanok.com/articles/llm-as-a-judge-for-code/What the research says about LLM judges: where they match human agreement, where position, length and self-bias distort them, and how to judge code safely.Wed, 15 Apr 2026 00:00:00 GMTHallucinated APIs and packages: how they get into code and how to stop themhttps://stepanok.com/articles/ai-code-hallucinations-and-slopsquatting/https://stepanok.com/articles/ai-code-hallucinations-and-slopsquatting/AI models invent packages and APIs. 19.7% of packages in a USENIX study didn't exist. Real examples, why it happens, and the checks that stop slopsquatting.Wed, 01 Apr 2026 00:00:00 GMTIs AI-generated code safe? A threat model for teams that ship ithttps://stepanok.com/articles/is-ai-generated-code-safe/https://stepanok.com/articles/is-ai-generated-code-safe/AI code fails security tests in 45% of Veracode's tasks. A practical threat model: which risks are likely, which are severe, and the controls for each.Wed, 18 Mar 2026 00:00:00 GMTWhat the AI coding productivity studies actually measuredhttps://stepanok.com/articles/ai-coding-productivity-studies/https://stepanok.com/articles/ai-coding-productivity-studies/One study says AI makes developers 55.8% faster, another says 19% slower. Both are right about different things. A guide to what each study measured.Wed, 04 Mar 2026 00:00:00 GMTHow to reduce code review time when AI agents write the codehttps://stepanok.com/articles/small-diffs-for-ai-agents/https://stepanok.com/articles/small-diffs-for-ai-agents/AI agents write bigger pull requests: Faros measured +154% in size. Why review maths still favours 200–400 lines, and how to make agents ship small diffs.Wed, 11 Feb 2026 00:00:00 GMTWhere the hours go: mapping human wait time in a delivery pipelinehttps://stepanok.com/articles/where-the-hours-go-human-wait-time/https://stepanok.com/articles/where-the-hours-go-human-wait-time/Most of a feature's lead time is waiting for people, not work. How to map wait time in your delivery pipeline and decide which human gates to automate.Wed, 28 Jan 2026 00:00:00 GMTAmdahl's law for software teams: why faster coding barely helpshttps://stepanok.com/articles/amdahls-law-for-software-teams/https://stepanok.com/articles/amdahls-law-for-software-teams/If coding is 16% of a developer's time, making it 10 times faster speeds delivery by about 17%. The maths behind the AI productivity paradox, and the fix.Wed, 14 Jan 2026 00:00:00 GMT