[spprashant]

I am not ashamed to admit this whole agentic coding movement has moved beyond me.

Not only do I have know everything about the code, data and domain, but now I need to understand this whole AI system which is a meta skill of its own.

I fear I may never be able catch up till someone comes along and simplifies it for pleb consumption.

[philbo]

I think this and other recent posts here hugely overcomplicate matters. I notice none of them provides an A/B test for each item of complexity they introduce, there's just a handwavy "this has proved to work over time".

I've found that a single CLAUDE.md does really well at guiding it how I want it to behave. For me that's making it take small steps and stop to ask me questions frequently, so it's more like we're pairing than I'm sending it off solo to work on a task. I'm sure that's not to everyone's taste but it works for me (and I say this as someone who was an agent-sceptic until quite recently).

Fwiw my ~/.claude/CLAUDE.md is 2.2K / 49 lines.

[d4rkp4ttern]

Indeed Anthropic’s best practices suggest keeping the CLAUDE.md relatively small.

[cruffle_duffle]

I’ve personally decided that cursor agent mode is good enough. A single foreground instance of cursor doing its thing is plenty enough to babysit. Based upon that experience I am highly highly skeptical people are actually creating things of value with these multi-agent-running-in-the-background setups. Way to much babysitting and honestly writing docs and specs for them is more work than just writing parts of the code myself and letting the LLM do the tedious bits like finishing what I started.

No matter what you are told, there is no silver bullet. Precisely defining the problem is always the hard part. And the best way to precisely define a problem and its solution is code.

I’ll let other people fight swarms of bots building… well who knows what. Maybe someday it will deliver useful stuff, but I’m highly skeptical.

[hoechst]

Much of it is just "put this magic string before your prompt to make the LLM 10x better" voodoo, similar to the SEO voodoo common in the 2000s.

just remember that it works the same for everyone: you input text, magic happens, text comes out.

if you can properly explain a software engineering problem in plain language, you're an expert in using LLMs. everything on top of that people experimenting or trying to build the next big thing.

[gdulli]

It's also possible to put in enough hours of real coding to get to the point where coding really isn't that hard anymore, at least not hard enough to justify switching from those stable/solid fundamental skills to a constantly revolving ecosystem of ephemeral tools, models, model versions, best practices, lessons from trial and error, etc. Then you could bypass all of this distraction.

Admittedly that stance is easiest to take if you were old enough, experienced enough already by the time this era hit.

[paweladamczuk]

"There exist developers whose performance cannot be boosted by an LLM" is a really strong statement.

[gdulli]

The point is that it takes significant time and attention to keep up with the treadmill of constantly learning the new tool/model/framework of the month, so there's significant opportunity cost. I have continued putting 100% of my attention on the direct problems I'm solving.

I don't see the coding as the hard or critical part of my work, so I don't put effort into accelerating or delegating that part.

[_se]

Not really. It's on the people asserting the positive (that LLMs do improve productivity for sufficiently experienced engineers) to prove it. In the absence of proof, the null hypothesis is the default.

[evanmoran]

To give you a process that might help:

I’ve found you have to use Claude Code to do something small. And as you do it iterate on the CLAUDE.md input prompt to refine what it does by default. As it doesn't do it your way, change it to see if you can fix how it works. The agent is then equivalent to calling chatgpt / sonnet 1000 times a hour. So these refinements (skills in the post are a meta approach) are all about how to tune the workflow to be more accurate for your project and fit your mental model. So as you tune the md file you’ll start to feel what is possible and understand agent capabilities much better.

So short story you have to try it, but long story its the iteration of the meta prompt approach that teaches you whats possible.

[lcnPylGDnU4H9OF]

I haven't really done much of it but my plan is just to practice. This seems like a powerful thing to start with.

[benhurmarcel]

> till someone comes along and simplifies it for pleb consumption

Just give it a few months. If some technics really work, it’ll get streamlined.