> ## Content Index
> Fetch the complete content index at: https://www.thecraft.page/llms.txt
> Use this file to discover other available public pages before exploring further.

# Fellow software developers, we need to fast track to acceptance.
- URL: https://www.thecraft.page/fellow-software-developers-we-need-to-fast-track-to-acceptance/
- Published: 2026-09-27T20:48:54.000Z
- Updated: 2026-09-27T20:48:54.000Z
- Author: Felipe Ramos

For over fifty years the business side of any company put a lot of money into getting rid of developers and with AI it looks like they are almost there. Each iteration of the models gets rid of prior grievances, silly mistakes, and lowers the barrier of entry (for the most part). Don’t get me wrong as someone that has been writing code, solving problems, and architecting systems for over twenty years there are still some issues here and there, but if I am honest, as we all should be, I can probably avoid writing code and direct AI to do everything needed.

We are past the argument of AI capabilities writing code. A business and a user don’t necessarily measure the quality of the code by how clean it is, how well architected, how adaptable it is – a user and by extension the business measures it by the value it provides. If a software produces a correct result, lines of code, difference of seconds and milliseconds won’t tip the balance. My argument was like many in our community, it generates bloat, it hallucinates magical libraries/packages, it fixes areas I didn’t ask it to fix and finds things that make no sense. Many of those issues are still present today but they are becoming less of a problem **behind capable hands**.

 The truth is lately I only write code of interest to me, code that I don’t want to forget, code that I enjoy writing, or code that challenges me. The rest of the code is a great portion written by whatever model I identified to produce better results for the problem. I do keep my changeset window open because I want to catch potential issues and to spot check larger modifications across many files. While I can undo changes as needed, reorient the agent to produce different results, or provide newer specs the job remains the same – I am a glorified babysitter (those roles will most likely reverse).

 Agents’ keeper has become a central role of current day development and as many of you have struggled and are still struggling, I believe it’s time to let go and learn to make the best of what we have. There are several reasons for my belief.

 The first one is ethical in nature. Accountability and ownership are severely lacking in all these AI labs and untimely consumers, and infrastructure may pay the price. I believe businesses for the most part will do what they have always done, put profits ahead of everything and without a **vested party witnessing** what this new breed of software looks like we might end up in worse place than we have ever been. It is hard to keep up with so much generated code that constantly mutates but we should have an idea of what it does, the rules it breaks, and at least know what the new code looks like around areas concerning security, privacy, and decision-making.

 Second, AI is going to take our jobs statement is no longer a statement. AI already took all our jobs even if you are still working in the same place, odds are the job has shifted, mutated, mandated, or won’t take long before it does. Many of us, me included, have enjoyed solving problems for many years, we learn the domain but don’t invest enough in the business itself. A business and its domain are different things and usually one decides how to operate regardless of the domain it is in. For example, how many banks have gone under or faced legal repercussions due to their business practices. I guess that’s a better term—business practice is determined by a handful of individuals often with conflicting motivations; their missions say one thing, but their actions say another. The longer we resist the new status quo, the longer it will take us to learn what’s needed to survive and still be part of the cog—an **internal witness**.

 Third, speaking a different language for so long has created a disconnect between what we know to be true and what the common knowledge tells everyone is the truth. As engineers/developers we are not always the best communicators regarding technical matters. We lost this battle a long time ago when AI was the intermediary for anyone with an idea. Regular people don’t want to know about cloud providers, containers, app service, caching, messaging systems, optimizations, and cost, to name a few. Most people want to enter a prompt and see something generated that would have taken effort, knowledge, and lots of money to build. **Can’t reason with someone without the knowledge to reason** about the potential issues with that approach.

 Finally, accepting this new normal without resisting means our efforts can go into preserving our craft and its evolution. As I said before, I still write code by hand, have physical books, and worry about practicing the fundamentals. To me the foundations are essentials that every developer needs to have, and while the line will blur increasingly, that understanding will still give you an edge even in the age of AI. Run a simple experiment, give someone any model to build an app without any software development experience. Build the same app and compare the results.

 We could potentially get into a future (meaning a few years from now) where everything we care about is run by AI and the humans still part of the process is there to meet a very real human quota. Let’s keep every edge we have and avoid eroding our knowledge and desire to push forward with or without AI.

 By accepting early, we can be/remain the capable hands.