--- title: I have not typed code for months. Here is how it feels. description: What happens when AI writes all your code? An honest look from Technical Lead János Farkas at the loss of developer flow, late-night burnout, and the guardrails needed to protect both the code and the engineer. image: https://8west.ie/hubfs/8%20West%20Consulting%20-%20Article%20-%20Janos%20Farkas%20-%20I%20have%20not%20typed%20code%20for%20months.%20Here%20is%20how%20it%20feels.png --- [![Navigate to: 8 West Consulting website Homepage](https://8west.ie/hubfs/SEO_optimized_media/8_west_logo.webp)](https://8west.ie/) [![8 West Consulting: leftward chevron icon](https://8west.ie/hubfs/SEO_optimized_media/8-West-Consulting-leftward-chevron-icon.png) Back to main insights page](https://8west.ie/insights) ## ARTICLE # I have not typed code for months. Here is how it feels. AI & innovation What happens when AI writes all your code? An honest look from Technical Lead János Farkas at the loss of developer flow, late-night burnout, and the guardrails needed to protect both the code and the engineer. ![8 West Consulting - Article - Janos Farkas - I have not typed code for months. Here is how it feels](https://8west.ie/hubfs/8%20West%20Consulting%20-%20Article%20-%20Janos%20Farkas%20-%20I%20have%20not%20typed%20code%20for%20months.%20Here%20is%20how%20it%20feels.png) ![Avatar picture of the author: Janos Farkas](https://8west.ie/hubfs/SEO_optimized_media/8-West-Consulting-Member-Janos-Farkas.png) ### Janos Farkas #### Technical Lead, 8 West Consulting SHARE ON [![Click to share this article on Linkedin](https://8west.ie/hubfs/SEO_optimized_media/8-west-Consulting-icon-logo-linkedin.png)](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2F8west.ie%2Finsights%2Fi-have-not-typed-code-for-months.-here-is-how-it-feels) October 13, 2026 I write about AI often, mostly about tools, methods and ideas. In private, developers and managers often talk seriously about the mental side of this work. So I collected what research and several well-known engineers say about it.   This article is about something that no benchmark shows: how developers feel when they build software with AI, and what we can do about these feelings. It is also about the Friday nights when I cannot stop, and the rules that I try now, in chapter 7.   In this article, I cover:   1. How I use AI at work, and at home 2. How each mode feels 3. What is missing: the Flow 4. What research says 5. Managers know this feeling already 6. What we can do 7. Protect yourself, not only the code 8. My take 9. Appendix: the scaffolding in detail     ## 1. How I use AI at work, and at home   I use AI in two very different ways, and each way gives me different feelings. **At day work: controlled and reviewed.** AI development is a controlled process, reviewed and slow on purpose. I split each story into small tasks that a person can review and test, and I control the output. The company expects AI usage, and the process works well. However, I noticed that I have not typed a line of production code myself for a long time. I still design the solution, split the work and read every change that goes in. So what is my role now? I design the solution and I judge the result, and both are important work. However, the part in between, where I built the thing with my own hands, now belongs to the AI, and that part gave me always the most joy.   **At home: augmented coding for my own tools.** I have Claude Code and Codex subscriptions, and I build tools for myself that make meticulous, boring manual tasks easier. Often I take an existing idea, such as a vector graphic program or a data processing method, and I change it into a very specific tool that only I need: I pick the features that I want, and I change them. These tools are not final products, so I care most about the input, the UX and the output, and I do not review 10,000 changed lines anymore. However, this is not vibe coding - I use an augmented approach: tests, rules and automatic checks watch the code instead of my eyes. Chapter 6 and the appendix explain how this works.     ## 2. How each mode feels   At my day job, the work is good and the results are safe, but my part of the work has changed. I read, I judge, and I approve or reject, so at the end of the day I have made many correct decisions, but I have not made anything with my own hands in the code. In his post "[We Mourn Our Craft](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fnolanlawson%2Ecom%2F2026%2F02%2F07%2Fwe-mourn-our-craft%2F&urlhash=4LlT&trk=article-ssr-frontend-pulse_little-text-block)", Nolan Lawson wrote that he did not want programmers to become glorified TSA agents who check AI code for dangerous content. I understand his point very well.   Now see what happens on a Friday evening, after a normal workday, when the house has gone silent. I am excited to sit down to work on my pet projects. I have complete freedom there, but surprisingly, it is not better! I built one tool for my accounting tasks and once a specialized vector graphic program. I describe what I want, I look at the result, I try it, and I ask for changes. I know this loop well, because I did the same work in lead roles at several companies, but here the full cycle takes a single hour instead of several weeks. The output is great and almost immediate, but I still become impatient and then frustrated, and when I stop, I feel empty. The waiting is the worst part, because I sit and watch the AI work, and I forget to stand up sometimes for hours.   I try to do something else in the meantime, such as watching random videos, Netflix or playing a game, but I must stay near in case the AI asks for something. Each new result gives me the feeling that I am getting somewhere, like the small wins that keep a gambler at the table, and the great results give me new ideas, so I continue for hours, sometimes until dawn. It feels almost like an addiction, and in the meantime my body aches, so at the end I feel physically unwell.     ## 3. What is missing: the Flow   Neither mode gives me the satisfaction that coding gave me before, and that satisfaction has a name: [flow](https://en.wikipedia.org/wiki/Flow_(psychology)?trk=article-ssr-frontend-pulse_little-text-block), as the psychologist Mihaly Csikszentmihalyi called it.   I remember evenings when I sat down with one problem, and I looked up only when it was dark outside. The code grew line by line, each green test pulled me deeper into the problem, and I held the full structure in my head. When I stood up, I was often not tired at all, I had built something, and I knew every part of it. Fred Brooks described this feeling more than 50 years ago in *The Mythical Man-Month*, in the chapter "[The Joys of the Craft](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fcs%2Ecalvin%2Eedu%2Fcourses%2Fcs%2F106%2F25fa%2Fresources%2Fbrooks%2F&urlhash=-XRy&trk=article-ssr-frontend-pulse_little-text-block)". He named five joys:   - making things, especially things of your own design, - making things that are useful to other people, - building complex puzzles and watching them work, - always learning, - and working in a medium that is as flexible as thought.   He wrote that "the programmer, like the poet, works only slightly removed from pure thought-stuff." Today, when I read his list, I ask how many of these five joys AI leaves to me?     ## 4. What research says   I wanted to know if other developers working with AI feel the same. The research points in the same direction.   **Motivation goes down after AI help.** [A 2025 study](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fwww%2Enature%2Ecom%2Farticles%2Fs41598-025-98385-2&urlhash=4FcT&trk=article-ssr-frontend-pulse_little-text-block) in *Scientific Reports* by Wu and colleagues ran four experiments with more than 3,500 people. With generative AI, people did better work, but when they then worked alone, their intrinsic motivation fell and their boredom increased.   **Developers find the work less fulfilling.** [A 2026 field study](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Farxiv%2Eorg%2Fpdf%2F2607%2E02337&urlhash=Y47h&trk=article-ssr-frontend-pulse_little-text-block) found that nearly half of the developers did not agree that AI makes their work more fulfilling. The authors give a possible reason: AI automates the parts that developers enjoy, such as hands-on coding, and developers then spend more time on problem definition and on checks of AI output.   **The workflow itself breaks flow.** Mihaly Csikszentmihalyi, who named the concept, gave three conditions for flow:   - a clear goal, - fast feedback, - and a challenge that matches your skill.   [A 2026 longitudinal study](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Farxiv%2Eorg%2Fpdf%2F2605%2E23135&urlhash=9bx0&trk=article-ssr-frontend-pulse_little-text-block) says that AI coding assistants can break flow by design, because each suggestion needs an evaluation and each generated result needs verification. Thus the cycle of prompt, review and retry is an interruption inside the workflow.  [Interruption research](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fshiftmag%2Edev%2Fdo-not-interrupt-developers-study-says-5715%2F&urlhash=V67L&trk=article-ssr-frontend-pulse_little-text-block) shows that after a break, a developer needs 10 to 15 minutes to start to edit code again, and up to 45 minutes to recover the full context. That research is about human interruptions, but I think the wait for an agent works in the same way.   **Impatience and emptiness have names.** One study of vibe coders has the title "[Building Software by Rolling the Dice](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Farxiv%2Eorg%2Fpdf%2F2512%2E22418&urlhash=Sfwp&trk=article-ssr-frontend-pulse_little-text-block)", and I read this title as a description of my impatience: when each prompt is a new roll of the dice, I want the next roll immediately. The feeling of emptiness matches [the IKEA effect](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fdash%2Eharvard%2Eedu%2Fserver%2Fapi%2Fcore%2Fbitstreams%2F7312037d-2473-6bd4-e053-0100007fdf3b%2Fcontent&urlhash=_Ouh&trk=article-ssr-frontend-pulse_little-text-block), which Norton, Mochon and Ariely described in 2012: people value things more when they put their own effort into them. For this reason, a good tool that cost me one hour of prompts feels more like a purchase than like my own work. **Not all developers feel this.** [A Microsoft Research study](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Farxiv%2Eorg%2Fpdf%2F2509%2E12491&urlhash=Doa7&trk=article-ssr-frontend-pulse_little-text-block) describes vibe coding as conversation, co-creation, flow and joy. It also found that trust decides where a developer stands between delegation and co-creation. This explains my own case, because at first, in my one-hour sessions, I delegated large steps and then waited, so I managed the work, but I was not part of it.     ## 5. Managers know this feeling already   Managers will read this and say "welcome to management", and they are right. A developer has many small wins every day, while the wins of a manager are big, slow, shared with the team, and dependent on the success of other people. Charity Majors, who writes about moving between engineering and management, [describes this well](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fcharity%2Ewtf%2F2019%2F01%2F04%2Fengineering-management-the-pendulum-or-the-ladder%2F&urlhash=amdL&trk=article-ssr-frontend-pulse_little-text-block): a manager may never again feel the quick satisfaction of learning or fixing something, and most people describe the joy of management as a much quieter version of the thrill of building. Every manager who came from engineering knows this feeling.   In *[The Progress Principle](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fwww%2Elibrary%2Ehbs%2Eedu%2Fworking-knowledge%2Fhow-small-wins-unleash-creativity&urlhash=X9wQ&trk=article-ssr-frontend-pulse_little-text-block)*, Teresa Amabile and Steven Kramer showed that small daily progress in meaningful work is the strongest source of good feelings at work. As a developer, I got this progress from each passing test and from each bug that I found. As a manager, I had to learn to find it in the progress of other people, and that took me years. But never got used to it.   The new part is the speed and the scale. This year, Kent Beck, the creator of Extreme Programming, [tried a multi-agent setup](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fnewsletter%2Ekentbeck%2Ecom%2Fp%2Fgenie-lessons-nobody-wants-agents&urlhash=XRoQ&trk=article-ssr-frontend-pulse_little-text-block), and he noticed that he was managing it: he watched which agent did what, and he held the state in his own head. With AI, every developer becomes a manager, but nobody asks them first, and nobody trains them for it.     ## 6. What we can do   ### Keep the skills that you value   Gergely Orosz [does not use AI for his writing](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fnewsletter%2Epragmaticengineer%2Ecom%2Fp%2Fthe-pragmatic-engineer-ama&urlhash=TRfA&trk=article-ssr-frontend-pulse_little-text-block), because he wants to keep that skill sharp. For coding, he uses AI and accepts that his hand-coding skill will get weaker. His advice is to keep using the skills that you value, even when AI can do the work. For me, this is the first step, because it means that I decide which skills I keep.   ### Know which kind of AI coding you do: augmented or vibe coding   Kent Beck uses the term [augmented coding](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fnewsletter%2Ekentbeck%2Ecom%2Fp%2Faugmented-coding-beyond-the-vibes&urlhash=uFWJ&trk=article-ssr-frontend-pulse_little-text-block) and he separates it clearly from vibe coding. In vibe coding, you care only about the behavior of the system, and when an error occurs, you give it back to the AI and hope for a fix. **In augmented coding, you still care about the code, its complexity and the tests,** so the values are the same as in hand coding, but you do not type most of the code. We are getting to the point where AI is so capable that it produces huge change sets, and we cannot review 10,000 changed lines anymore. However, this is not vibe coding if we still care about the code and have a method to check it.   ### Check with scaffolding that you trust   Uncle Bob, the author of *Clean Code*, [now says](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fstartupfortune%2Ecom%2Funcle-bob-martin-says-he-no-longer-reads-ai-generated-code-and-the-developer-world-is-split%2F&urlhash=B3ZN&trk=article-ssr-frontend-pulse_little-text-block) that he does not read the code that his AI agents write. At first, this looks like the end of craftsmanship, but knowing him I think that he moved his craft one level higher. He surrounds his agents with strict constraints, such as tests, complexity limits, dependency analysis and mutation testing, and he relies on them instead of reading the code.   Not everybody agrees with him though. [Grady Booch says](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fstartupfortune%2Ecom%2Funcle-bob-martin-says-he-no-longer-reads-ai-generated-code-and-the-developer-world-is-split%2F&urlhash=B3ZN&trk=article-ssr-frontend-pulse_little-text-block) that metrics cannot find the security holes and dead code that an experienced engineer sees immediately. I think that both of them are correct, and the stakes decide which approach to use. In my regulated day job, I still review the code, but for my own tools, I trust the scaffolding and ask the proper questions.   Birgitta Böckeler calls this [harness engineering](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fmartinfowler%2Ecom%2Farticles%2Fharness-engineering%2Ehtml&urlhash=yFn-&trk=article-ssr-frontend-pulse_little-text-block): guides steer the agent before it works, and sensors check the result after it works. My harness has acceptance tests that I own, tests first in small steps, mutation and property-based testing, architecture and complexity rules, and hard gates that the agent cannot skip, and the appendix at the end of the article explains each part with tools.   Building this scaffolding also brings back part of the flow. Tests first give each step a clear goal, small steps give fast feedback, and the design of the harness is a challenge that matches my skill. So the three conditions of flow from chapter 4 could come back, only at a different level.   Vibe coding trusts the output. Augmented coding with this scaffolding checks everything, but the rules that I designed do the checks, not my eyes. The design of a good test or a hard rule is real engineering work, so each green run is a small success that I earned. What I understand now: the method, the harness and the checks above are part of our craftsmanship. A carpenter who builds his own jigs is still a carpenter, even when a machine makes the cut. In the same way, when I design the tests, the rules and the gates, and I improve them after each mistake, I still craft software, only at one level higher. This work gives back part of the joy, because the harness is mine, I built it myself, and it gets better every week.     ## 7. Protect yourself, not only the code   The scaffolding protects the code, but it did not protect me on those Friday nights. The joy comes from building and improving the harness, not from watching it run, and the harness only helps me if I trust it enough to leave the screen. When I just sit and watch it work, the session slides back into a slot machine.   When I started to read about this, I found that I am far from alone.   In April 2026, [Axios reported](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fwww%2Eaxios%2Ecom%2F2026%2F04%2F04%2Fai-agents-burnout-addiction-claude-code-openclaw&urlhash=jMo2&trk=article-ssr-frontend-pulse_little-text-block) that Quentin Rousseau, the CTO of Rootly, could not sleep for months after he switched to agentic coding, and in the end he needed sleep medication. He compares the tools to slot machines. Simon Willison, a developer with 25 years of experience, [said](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fwww%2Edigitaltoday%2Eco%2Ekr%2Fen%2Fview%2F45448%2Fdevelopers-hooked-on-ai-coding-tools-report-sleep-disorders-cognitive-overload&urlhash=e7uw&trk=article-ssr-frontend-pulse_little-text-block) that human cognition has limits, and that to give up sleep for agent coding is not sustainable. Steve Yegge wrote in his post "[The AI Vampire](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fsteve-yegge%2Emedium%2Ecom%2Fthe-ai-vampire-eda6e4f07163&urlhash=sbPY&trk=article-ssr-frontend-pulse_little-text-block)" that building things with AI takes a lot of human energy and that we should assume exhaustion is the norm, so he suggests short, focused working sessions. [LeadDev](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fleaddev%2Ecom%2Fai%2Faddictive-agentic-coding-has-developers-losing-sleep&urlhash=tep0&trk=article-ssr-frontend-pulse_little-text-block) also wrote about a principal engineer who codes late, not because somebody tells him to, but because he cannot get up from the computer. One [survey of 305 developers](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fcoddy%2Etech%2Fblog%2Fcoding-for-beginners%2Fai-coding-addiction-report&urlhash=tpuV&trk=article-ssr-frontend-pulse_little-text-block) found that 43% continue to code with AI after hours when they planned to stop, and 32% have put off sleep to continue. The research is still early, and "AI coding addiction" is not a medical diagnosis, but the pattern is clear enough to take seriously. I think the reason is simple. With hand-coding, my own tiredness was the natural end: I got stuck, I got tired, and I stopped.   The agent never gets tired, so the session ends only when I decide to end it. Each result is a small, unpredictable reward, like a pull on a slot machine, and each near-miss feels like progress.   The waiting is not rest either, because my eyes stay on a screen while half of my attention stays on the agent.   So, what can we do?   These are the rules that I try now:   1. **Decide the end before I start.** I set an end time or one finished goal, such as "this test passes", before the session begins, because during the session every result gives me a reason to continue. 2. **Stop at a green point and write down the next step.** An unfinished task stays in the mind, and psychologists call this the [Zeigarnik effect](https://en.wikipedia.org/wiki/Zeigarnik_effect?trk=article-ssr-frontend-pulse_little-text-block). Research also shows that a specific plan for the unfinished task [stops the intrusive thoughts that it causes](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fwww%2Epsychologytoday%2Ecom%2Fus%2Fbasics%2Fzeigarnik-effect&urlhash=rbQA&trk=article-ssr-frontend-pulse_little-text-block). 3. **Use the harness to leave the room.** The gates stop bad changes, and Claude Code can send a notification when the agent needs input. So I can give the agent a well-specified job, walk away, and come back when my phone calls me. 4. **Make the waiting a real break.** I stand up, walk, stretch or get water, instead of switching to another screen.   This is not a solved problem for me. Some nights the pull still wins, but now I recognize the problem earlier.     ## 8. My take   I use AI every day, and I use it by choice, because I do not want to go back to the old speed. The market expects AI speed, and AI has much wider domain knowledge than I have. With frontier models inside a good harness, the code quality can be trusted, and often it is better than what I would write by hand. So I no longer look for the old joy of typing. I look for it in the harness that I design and improve, and in the rules that protect my evenings.   At work, I find the joy in a surprising place: the stricter frame. Small tasks, reviews and rules that I cannot skip give each day clear goals and a clear end, and this gives me more joy than the full freedom semi-unattended agentic work at home.   Some developers also have already found their joy again. Matt Morgis moved from engineering manager back to staff engineer, and [he said](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fnewsletter%2Epragmaticengineer%2Ecom%2Fp%2Fai-engineering-in-the-real-world&urlhash=8AoZ&trk=article-ssr-frontend-pulse_little-text-block) that work with AI brought back his joy in coding. I do not feel this every day yet, but now I know where to look.   My question to you: where do you find your small wins now? Did you find flow again with AI, or did you find it somewhere else?     ## 9. Appendix: the scaffolding in detail   This chapter is for readers who want to build their own harness. It explains each part of the scaffolding from chapter 6, with tools for popular languages.   1. **Acceptance tests that I own.** These tests describe in plain sentences what the system must do, for example: "When an order has no date, the system rejects it." A builder does not need me to watch each brick, but he needs a clear contract for the house, and these tests are that contract. If I review only one thing, I review these tests. 2. **Tests first, in small steps.** The agent writes one failing test, then only the code that makes it pass, and then the next test. This stops the agent from building things that I did not ask for, as [Kent Beck found in his own project](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fnewsletter%2Ekentbeck%2Ecom%2Fp%2Faugmented-coding-beyond-the-vibes&urlhash=uFWJ&trk=article-ssr-frontend-pulse_little-text-block), and each green test is a small win again. 3. **Mutation testing.** A test can pass and still check nothing, and AI writes such tests easily. A mutation tool puts small bugs into the code on purpose, for example it changes \> to \>=, and if the tests still pass, they cannot see that part of the code. It works like a fire drill, because I learn if the alarms work before a real fire starts. Good tools exist for most popular languages: [PIT](https://www.linkedin.com/redir/redirect?url=http%3A%2F%2Fpitest%2Eorg%2F&urlhash=kFBr&trk=article-ssr-frontend-pulse_little-text-block) for Java, [Stryker](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fstryker-mutator%2Eio%2F&urlhash=RRUU&trk=article-ssr-frontend-pulse_little-text-block) for JavaScript, TypeScript and C#, [mutmut](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fgithub%2Ecom%2Fboxed%2Fmutmut&urlhash=6t6F&trk=article-ssr-frontend-pulse_little-text-block) for Python, [cargo-mutants](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fmutants%2Ers%2F&urlhash=Q_CY&trk=article-ssr-frontend-pulse_little-text-block) for Rust, [Infection](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Finfection%2Egithub%2Eio%2F&urlhash=0BGh&trk=article-ssr-frontend-pulse_little-text-block) for PHP, [Mutant](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fgithub%2Ecom%2Fmbj%2Fmutant&urlhash=Y16f&trk=article-ssr-frontend-pulse_little-text-block) for Ruby, [Mull](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fgithub%2Ecom%2Fmull-project%2Fmull&urlhash=SWKM&trk=article-ssr-frontend-pulse_little-text-block) for C and C++, and [Gremlins](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fgithub%2Ecom%2Fgo-gremlins%2Fgremlins&urlhash=-buM&trk=article-ssr-frontend-pulse_little-text-block) or [go-mutesting](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fgithub%2Ecom%2Fzimmski%2Fgo-mutesting&urlhash=uolB&trk=article-ssr-frontend-pulse_little-text-block) for Go. 4. **Property-based testing.** A normal test checks one example, but a property test checks a rule, for example: "When I export a shape and import it again, the shape stays the same." The tool tries hundreds of random inputs to break the rule, and it finds edge cases that the AI and I did not think of. There are mature libraries for most languages: **[QuickCheck](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fhackage%2Ehaskell%2Eorg%2Fpackage%2FQuickCheck&urlhash=zAp5&trk=article-ssr-frontend-pulse_little-text-block)** in Haskell, [jqwik](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fjqwik%2Enet%2F&urlhash=l8NL&trk=article-ssr-frontend-pulse_little-text-block) for Java, [Hypothesis](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fhypothesis%2Ereadthedocs%2Eio%2F&urlhash=DP-f&trk=article-ssr-frontend-pulse_little-text-block) for Python, [fast-check](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Ffast-check%2Edev%2F&urlhash=GWnJ&trk=article-ssr-frontend-pulse_little-text-block) for JavaScript and TypeScript, [FsCheck](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Ffscheck%2Egithub%2Eio%2FFsCheck%2F&urlhash=Ixgr&trk=article-ssr-frontend-pulse_little-text-block) for C#, [proptest](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fgithub%2Ecom%2Fproptest-rs%2Fproptest&urlhash=fOiB&trk=article-ssr-frontend-pulse_little-text-block) for Rust, [rapid](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fgithub%2Ecom%2Fflyingmutant%2Frapid&urlhash=lZZf&trk=article-ssr-frontend-pulse_little-text-block) for Go, and [RapidCheck](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fgithub%2Ecom%2Femil-e%2Frapidcheck&urlhash=q0U8&trk=article-ssr-frontend-pulse_little-text-block) for C++. 5. **Architecture rules.** I write rules about which part of the code can call which part, and a tool such as [ArchUnit](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fwww%2Earchunit%2Eorg%2F&urlhash=ymI1&trk=article-ssr-frontend-pulse_little-text-block) changes these rules into tests. The agent can move the furniture, but it cannot remove a load-bearing wall. 6. **Complexity and size limits.** Hard limits on method complexity and file length force the agent to split tangled code. Thus the code stays readable for the day when I must read it. Common tools for these limits are [Checkstyle](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fcheckstyle%2Eorg%2Fchecks%2Fmetrics%2Fcyclomaticcomplexity%2Ehtml&urlhash=slZI&trk=article-ssr-frontend-pulse_little-text-block) and [PMD](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fpmd%2Egithub%2Eio%2F&urlhash=miVw&trk=article-ssr-frontend-pulse_little-text-block) for Java, the [ESLint complexity rule](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Feslint%2Eorg%2Fdocs%2Frules%2Fcomplexity&urlhash=iV_Z&trk=article-ssr-frontend-pulse_little-text-block) for JavaScript and TypeScript, [Ruff](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fdocs%2Eastral%2Esh%2Fruff%2Frules%2Fcomplex-structure&urlhash=LWve&trk=article-ssr-frontend-pulse_little-text-block) for Python, [gocyclo](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fgithub%2Ecom%2Ffzipp%2Fgocyclo&urlhash=Q4NW&trk=article-ssr-frontend-pulse_little-text-block) for Go, [Lizard](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fgithub%2Ecom%2Fterryyin%2Flizard&urlhash=33Wp&trk=article-ssr-frontend-pulse_little-text-block) for many languages, and [SonarQube](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fdocs%2Esonarsource%2Ecom%2Fsonarqube-server%2Fglossary&urlhash=EUQR&trk=article-ssr-frontend-pulse_little-text-block) as a platform that also tracks cognitive complexity. 7. **Static analysis and security scanners.** Linters, strict types and scanners find known bad patterns, such as unsafe input handling or unused code. They find part of the problems that Grady Booch is worried about, and they do not get tired. 8. **Hard gates.** The checks run automatically on each change and before each merge, so the agent cannot skip them. The agent can say "done" many times, but only the gate decides. Both main agents support this directly: Claude Code has [hooks](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fcode%2Eclaude%2Ecom%2Fdocs%2Fen%2Fhooks-guide&urlhash=eVqs&trk=article-ssr-frontend-pulse_little-text-block) (see the [hooks reference](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fcode%2Eclaude%2Ecom%2Fdocs%2Fen%2Fhooks&urlhash=wbic&trk=article-ssr-frontend-pulse_little-text-block)), and Codex has [hooks](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fdevelopers%2Eopenai%2Ecom%2Fcodex%2Fhooks&urlhash=QGDp&trk=article-ssr-frontend-pulse_little-text-block) too, so a check can run before or after each tool call and block the change when it fails. 9. **Guides before the work.** A file with my conventions, my architecture and my forbidden patterns gives the agent the rules before it starts. A good briefing causes fewer failed gates, and this is also true for people. 10. **Spot checks.** When all checks are green, I still read a random sample from time to time, as factories do: an inspector takes random pieces from the production line and examines them. Quality engineering calls this [acceptance sampling](https://en.wikipedia.org/wiki/Acceptance_sampling?trk=article-ssr-frontend-pulse_little-text-block), and it gives good confidence without checking every piece. When I find a problem that the scaffolding missed, I do not read all the code again, but I add a new rule.     ## Sources   - Wu, Liu, Ruan, Chen, Xie: [Human-generative AI collaboration enhances task performance but undermines human's intrinsic motivation](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fwww%2Enature%2Ecom%2Farticles%2Fs41598-025-98385-2&urlhash=4FcT&trk=article-ssr-frontend-pulse_little-text-block), Scientific Reports, 2025 - [Developers' Experience with Generative AI Beyond Productivity Assessment](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Farxiv%2Eorg%2Fpdf%2F2607%2E02337&urlhash=Y47h&trk=article-ssr-frontend-pulse_little-text-block), arXiv, 2026 - [The Impact of AI Coding Assistants on Software Engineering: A Longitudinal Study](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Farxiv%2Eorg%2Fhtml%2F2605%2E23135v1&urlhash=dZCD&trk=article-ssr-frontend-pulse_little-text-block), arXiv, 2026 - [Building Software by Rolling the Dice: A Qualitative Study of Vibe Coding](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Farxiv%2Eorg%2Fpdf%2F2512%2E22418&urlhash=Sfwp&trk=article-ssr-frontend-pulse_little-text-block), arXiv - Pimenova, Fakhoury, Bird, Storey, Endres: [Good Vibrations? A Qualitative Study of Co-Creation, Communication, Flow, and Trust in Vibe Coding](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2509%2E12491&urlhash=Qktz&trk=article-ssr-frontend-pulse_little-text-block), Microsoft Research - Breaking the Flow: A Study of Interruptions During Software Engineering Activities, [summary on ShiftMag](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fshiftmag%2Edev%2Fdo-not-interrupt-developers-study-says-5715%2F&urlhash=V67L&trk=article-ssr-frontend-pulse_little-text-block) - Fred Brooks: The Mythical Man-Month, 1975, chapter 1, [The Joys of the Craft](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fcs%2Ecalvin%2Eedu%2Fcourses%2Fcs%2F106%2F25fa%2Fresources%2Fbrooks%2F&urlhash=-XRy&trk=article-ssr-frontend-pulse_little-text-block) - Mihaly Csikszentmihalyi: Flow, 1990 - Teresa Amabile, Steven Kramer: [The Progress Principle](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fhbswk%2Ehbs%2Eedu%2Fitem%2F6685%2Ehtml&urlhash=wJ6z&trk=article-ssr-frontend-pulse_little-text-block), 2011 - Norton, Mochon, Ariely: [The IKEA effect: When labor leads to love](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fdash%2Eharvard%2Eedu%2Fserver%2Fapi%2Fcore%2Fbitstreams%2F7312037d-2473-6bd4-e053-0100007fdf3b%2Fcontent&urlhash=_Ouh&trk=article-ssr-frontend-pulse_little-text-block), Journal of Consumer Psychology, 2012 - Nolan Lawson: [We Mourn Our Craft](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fnolanlawson%2Ecom%2F2026%2F02%2F07%2Fwe-mourn-our-craft%2F&urlhash=4LlT&trk=article-ssr-frontend-pulse_little-text-block), February 2026 - Kent Beck: [Augmented Coding: Beyond the Vibes](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fnewsletter%2Ekentbeck%2Ecom%2Fp%2Faugmented-coding-beyond-the-vibes&urlhash=uFWJ&trk=article-ssr-frontend-pulse_little-text-block) and [Genie Lessons: Nobody Wants Agents](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fnewsletter%2Ekentbeck%2Ecom%2Fp%2Fgenie-lessons-nobody-wants-agents&urlhash=XRoQ&trk=article-ssr-frontend-pulse_little-text-block) - Gergely Orosz: [The Pragmatic Engineer AMA](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fnewsletter%2Epragmaticengineer%2Ecom%2Fp%2Fthe-pragmatic-engineer-ama&urlhash=TRfA&trk=article-ssr-frontend-pulse_little-text-block) and [AI Engineering in the real world](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fnewsletter%2Epragmaticengineer%2Ecom%2Fp%2Fai-engineering-in-the-real-world&urlhash=8AoZ&trk=article-ssr-frontend-pulse_little-text-block) (Matt Morgis) - [Uncle Bob Martin says he no longer reads AI-generated code](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fstartupfortune%2Ecom%2Funcle-bob-martin-says-he-no-longer-reads-ai-generated-code-and-the-developer-world-is-split%2F&urlhash=B3ZN&trk=article-ssr-frontend-pulse_little-text-block), Startup Fortune (with Grady Booch's reply) - Charity Majors: [Engineering Management: The Pendulum Or The Ladder](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fcharity%2Ewtf%2F2019%2F01%2F04%2Fengineering-management-the-pendulum-or-the-ladder%2F&urlhash=amdL&trk=article-ssr-frontend-pulse_little-text-block) - Birgitta Böckeler: [Harness engineering for coding agent users](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fmartinfowler%2Ecom%2Farticles%2Fharness-engineering%2Ehtml&urlhash=yFn-&trk=article-ssr-frontend-pulse_little-text-block), martinfowler.com, 2026 - [AI agents are scrambling power users' brains](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fwww%2Eaxios%2Ecom%2F2026%2F04%2F04%2Fai-agents-burnout-addiction-claude-code-openclaw&urlhash=jMo2&trk=article-ssr-frontend-pulse_little-text-block), Axios, April 2026 - [Developers hooked on AI coding tools report sleep disorders, cognitive overload](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fwww%2Edigitaltoday%2Eco%2Ekr%2Fen%2Fview%2F45448%2Fdevelopers-hooked-on-ai-coding-tools-report-sleep-disorders-cognitive-overload&urlhash=e7uw&trk=article-ssr-frontend-pulse_little-text-block), Digital Today, April 2026 - Steve Yegge: [The AI Vampire](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fsteve-yegge%2Emedium%2Ecom%2Fthe-ai-vampire-eda6e4f07163&urlhash=sbPY&trk=article-ssr-frontend-pulse_little-text-block), February 2026 - [‘Addictive’ agentic coding has developers losing sleep](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fleaddev%2Ecom%2Fai%2Faddictive-agentic-coding-has-developers-losing-sleep&urlhash=tep0&trk=article-ssr-frontend-pulse_little-text-block), LeadDev, 2026 - [The AI Coding Addiction Report](https://www.linkedin.com/redir/redirect?url=https%3A%2F%2Fcoddy%2Etech%2Fblog%2Fcoding-for-beginners%2Fai-coding-addiction-report&urlhash=tpuV&trk=article-ssr-frontend-pulse_little-text-block), Coddy Tech, August 2026 Enjoyed this article? Follow Janos Farkas for more insights and commentary on: [![Social Icon: Linkedin](https://8west.ie/hubfs/SEO_optimized_media/8-west-Consulting-icon-logo-linkedin.png) LINKEDIN](https://www.linkedin.com/in/janosfarkas/) [AI Enablement for Enterprise Development Teams](https://8west.ie/hubfs/SEO_optimized_media/Impact_Engagements_PDFs/8%20West%20Consulting%20-%20AI%20Enablement%20for%20Enterprise%20Development%20Teams.pdf) [BOOK A SCOPING CALL](https://8west.ie/about-us/contact-us) [![8 West Consulting: chevron icon](https://8west.ie/hubfs/SEO_optimized_media/8-West-Consulting-leftward-chevron-icon.png) ALL POSTS](https://8west.ie/insights) [NEXT ARTICLE ![8 West Consulting: chevron icon](https://8west.ie/hubfs/SEO_optimized_media/8-West-Consulting-leftward-chevron-icon.png)](https://8west.ie/insights/why-agentforce-stalls-inside-health-dental-payers-and-the-foundation-that-fixes-it) ## RELATED CONTENT ![Why Agentforce Stalls Inside Health & Dental Payers (and the Foundation That Fixes It)](https://8west.ie/hubfs/8%20West%20consulting%20-%20Article%20-%20Why%20Agentforce%20Stalls.png) #### ARTICLE [Why Agentforce Stalls Inside Health & Dental Payers (and the Foundation That Fixes It)](https://8west.ie/insights/why-agentforce-stalls-inside-health-dental-payers-and-the-foundation-that-fixes-it) ![Getting Started with Wayfinder](https://8west.ie/hubfs/8%20West%20Consulting%20-%20Article%20-%20Russ%20Painter%20-%20Getting%20Started%20with%20Wayfinder.jpg) #### ARTICLE [Getting Started with Wayfinder](https://8west.ie/insights/getting-started-with-wayfinder) 8 West Consulting © 2026 [Privacy Policy](https://8west.ie/privacy) [Sitemap](https://8west.ie/sitemap) [![Navigate to: 8 West Consulting Facebook page](https://8west.ie/hubfs/SEO_optimized_media/8-west-Consulting-icon-logo-facebook.png)](https://www.facebook.com/8westconsulting/) [![Navigate to: 8 West Consultng Youtube Channel](https://8west.ie/hubfs/SEO_optimized_media/8-west-Consulting-icon-logo-youtube-white.png)](https://www.youtube.com/@8westconsulting) [![Navigate to: 8 West Consulting Instagram business profile](https://8west.ie/hubfs/SEO_optimized_media/8-west-Consulting-icon-logo-instagram.png)](https://www.instagram.com/8westconsulting/) [![Navigate to: 8 West Consulting LinkedIn business profile](https://8west.ie/hubfs/SEO_optimized_media/8-west-Consulting-icon-logo-linkedin.png)](https://www.linkedin.com/company/8-west-consulting) ```json { "@context" : "https://schema.org", "@type" : "BlogPosting", "author" : { "@type" : "Person", "name" : "Janos Farkas", "url" : "https://8west.ie/insights/author/janos-farkas" }, "dateModified" : "2026-10-13T08:54:12.610Z", "datePublished" : "2026-10-13T08:54:12.000Z", "headline" : "I have not typed code for months. Here is how it feels.", "image" : [ "https://8west.ie/hubfs/8%20West%20Consulting%20-%20Article%20-%20Janos%20Farkas%20-%20I%20have%20not%20typed%20code%20for%20months.%20Here%20is%20how%20it%20feels.png" ], "mainEntityOfPage" : { "@id" : "https://8west.ie/insights/i-have-not-typed-code-for-months.-here-is-how-it-feels", "@type" : "WebPage" }, "publisher" : { "@type" : "Organization", "logo" : { "@type" : "ImageObject", "url" : "https://8west.ie/hubfs/Imported%20images/logos/8West.png" }, "name" : "8 West Consulting" } } ``` ```json { "@context" : "https://schema.org", "@id" : "https://8west.ie/#organization", "@type" : "Organization", "areaServed" : [ "Europe", "United States", "Global" ], "contactPoint" : [ { "@type" : "ContactPoint", "availableLanguage" : [ "English" ], "contactType" : "sales", "email" : "info@8west.ie" }, { "@type" : "ContactPoint", "areaServed" : "IE", "contactType" : "sales", "telephone" : "+353 21 4925100" }, { "@type" : "ContactPoint", "areaServed" : "IE", "contactType" : "support", "telephone" : "+353 21 4925155" } ], "description" : "8 West Consulting is an ISO 27001-certified software partner delivering enterprise systems for complex, regulated Health InsurTech and ecommerce environments.", "knowsAbout" : [ "Software development", "Cloud engineering", "Data engineering", "Artificial intelligence", "Dev Ops", "Cybersecurity", "Ecommerce", "Health insurance technology (Health Insurtech)", "Digital marketing" ], "logo" : { "@type" : "ImageObject", "url" : "https://145696986.fs1.hubspotusercontent-eu1.net/hubfs/145696986/SEO_optimized_media/8-West-Consulting-Logo-250x250.png" }, "name" : "8 West Consulting", "sameAs" : [ "https://www.facebook.com/8westconsulting", "https://www.youtube.com/@8westconsulting", "https://ie.linkedin.com/company/8-west-consulting", "https://www.instagram.com/8westconsulting/" ], "url" : "https://8west.ie/" } ``` ```json { "@context" : "https://schema.org", "@id" : "https://8west.ie/#website", "@type" : "WebSite", "name" : "8 West Consulting", "potentialAction" : { "@type" : "SearchAction", "query-input" : "required name=search_term_string", "target" : "https://8west.ie//search?q={search_term_string}" }, "publisher" : { "@id" : "https://8west.ie/#organization" }, "url" : "https://8west.ie/" } ``` ```json { "@context" : "https://schema.org", "@id" : "https://8west.ie/insights#1002", "@type" : "WebPage", "about" : { "@id" : "https://8west.ie/#organization" }, "description" : "This page collects the original articles, webinars, and white papers produced and published by the members of 8 West Consulting.", "isPartOf" : { "@id" : "https://8west.ie/#website" }, "name" : "Insights | 8 West Consulting", "url" : "https://8west.ie/insights" } ``` ```json { "@context" : "https://schema.org", "@type" : "BreadcrumbList", "itemListElement" : [ { "@type" : "ListItem", "item" : "https://8west.ie/", "name" : "Home", "position" : 1 }, { "@type" : "ListItem", "item" : "https://8west.ie/insights", "name" : "Insights", "position" : 2 } ] } ```