Generative AI coding tools and agents do not work for me

> As the reader I had to do too much work to discern your point and what was relevant.

Re-reading, I hope the first ~15 words make my main point:

> Either/or fallacy. There exist a varied set of ways to engage with the technology...

Was this part unclear? Something else?

> you think you're being open-minded...

"Open minded" can mean very different things to different people. I recommend the article "The Proper Use of Humility" by Yudkowsky [1] because it rings true to me. I'm open to hearing other people's points of view, up to a point, given enough time. (Everyone has their limit, whether we admit it or not.) When it comes to assessing truth, I care about good arguments and good evidence, and I heavily discount anything else. If someone says I'm not "open minded" because of what I just wrote, then my reply would be "what do you want me to be more open to?"

There is a gem from in a comment below the above article that deserves repeating:

> People often take open disagreement as a sign of intellectual arrogance, while it is a display of respect and humility; showing respect with the honest acknowledgment of your disagreement, and showing humility in affording the other person a chance to defend themselves and prove you wrong. To say nothing is to treat that person's beliefs dismissively, as if they don't matter, and then assume that discussion was futile because they're incapable of understanding the truth, and of course, couldn't possible have anything to teach you.

> ...but you offered a multiple choice question, where the choices are reductive...

I offered two specific categories (tone or substance) and a third option for "anything else". I'm not following why this feels reductive to you; it leaves space for someone to reply however they like.

> and it comes off as defensive.

I've thought about this word quite a bit. From dictionary.com defensive means "excessively concerned with guarding against the real or imagined threat of criticism, injury to one's ego, or exposure of one's shortcomings." I'm open to criticism and happy to learn. If I'm wrong, I strive to admit it and apologize where needed. At the same time, I am confident enough to push back, stand up for myself, and defend my ideas (which is a different sense of 'defensive').

Here is the backstory to my second comment. The comment I replied to did not strike me as kind, much less well-intentioned. It probably was intended to be an insult, but I replied anyway. I gave the benefit of the doubt while challenging the commenter to give constructive criticism. I strived for clarity and confidence without being defensive or going on a counter-attack. This is a hard balance to strike.

[1] https://www.lesswrong.com/posts/GrDqnMjhqoxiqpQPw/the-proper...

Analyzing crash dumps and figuring out what's going on is a pretty useful skill.

The later is not making any neuron embedding tradeoff when they hand of the slog to agents.

There’s a lot of software development in that latter category.

We ran into the same problem when rolling out AI-assisted code reviews and code generation pipelines. What helped us was adopting AppMod.AI's Project Analyzer: - Memory & context retention: It parses your full repo, builds architecture diagrams and dependency maps—so AI suggestions stay grounded in real code structure. - Human-in-the-loop chat interface: You can ask clarifying questions like, “Does this function follow our performance pattern?” and get guided explanations from the tool before merging. - Collaborative refactor tracking: It tracks changes and technical debt over time, making it easy to spot drift or architectural erosion—something pure LLMs miss. - Prompt-triggered cost and quality metrics: You can see how often you call AI, what it costs, and its success rates in passing your real tests—not just anecdotal gains.

It’s far from perfect, but it shifts the workflow from “LLM writes → you fix” to “LLM assists within your live code context, under your control.” Others have noted similar limitations in Copilot and GPT-4 based tools—where human validation remains essential .

In short: LLMs aren’t going to replace senior devs—they’re tools that need tooling. Blending AI insights with architecture-aware context and built-in human validation feels like the best middle path so far.