15:[[47511075,{"by":"zane__chen","descendants":15,"id":47511075,"kids":[47522413,47565301,47533061,47513327,47516285,47518176,47515534,47527451,47538412,47541393,47514076,47511320,47547965,47517438,47526873,47515200,47515473,47513168,47511451,47514680,47525046],"score":5,"text":"As titled.","time":1774395030,"title":"Ask HN: Do you feel less happy when coding with agent?","type":"story"}],[47493938,{"by":"jawerty","descendants":30,"id":47493938,"kids":[47494310,47495954,47496113,47496407,47501451,47500589,47499846,47507822,47496051,47495978,47498364,47494013,47495826,47520025,47552937,47563729,47565252,47538966,47500822],"score":32,"text":"I read on Reddit about a podcast where Karpathy described how he went from writing 80% of his own code to 0%, being in a constant state of “AI psychosis” because the possibilities feel infinite.

I’ve personally found that my workflow has become very “opportunistic”—I feel like I can do anything with AI, so I try everything. That might be good…or bad. I’d be curious to see what HN has to say, or whether anyone else has experienced something similar.

Here’s the Reddit post for context: https://www.reddit.com/r/ClaudeAI/comments/1s08r1c/karpathy_says_he_hasnt_written_a_line_of_code/

Anyone also feeling this way?? If not psychosis which may be an exaggeration then feeling more stressed, frazzled, whatever.","time":1774293682,"title":"Ask HN: Is anyone here also developing \"perpetual AI psychosis\" like Karpathy?","type":"story"}],[47543324,{"by":"midnight_eclair","descendants":6,"id":47543324,"kids":[47554257,47550813,47566325,47565183,47547067,47566001,47552233,47552935,47556097,47565208,47550703],"score":5,"text":"my current convictions are:

1. "agentic" development is here to stay, whether we like it or not

2. markdown prompts are a terrible way of driving such development, no matter how much structure and hierarchy you're trying to impose (whether chaining or fan-out into personas, etc)

3. jigsaw-puzzle approach is less terrible - declare the "boundary" of any component or subsystem in a normative language of your choice and let the agent fill the insides

it's fine if you think those convictions are incorrect, but i'm more interested in feedback from folks that more-or-less agree, with regard to the last point: what do YOU use for normative language to describe component boundary, function and cross-component interactions?

i've had mild success with openapi, but maybe there are better alternatives?","time":1774623042,"title":"Ask HN: What do you use for normative specs to drive AI agents?","type":"story"}],[47559293,{"by":"bhaviav100","descendants":32,"id":47559293,"kids":[47577331,47559616,47589330,47595082,47595187,47566274,47559476,47560688,47565058,47562067,47563774,47559326,47571565,47576138,47560929,47560802,47560786,47564748,47659063,47600210,47576063,47570777,47563358,47585228,47560734,47693441,47583024,47596566,47561547,47564825,47582833,47572882,47568131,47569988,47560578,47560973],"score":8,"text":"$16","time":1774743740,"title":"Ask HN: How are you keeping AI coding agents from burning money?","type":"story"}],[47511823,{"by":"makingstuffs","descendants":12,"id":47511823,"kids":[47515304,47520714,47518294,47521492,47516515,47513599,47512151,47512401,47512404,47512189,47520326,47547176,47515005,47515197,47529807,47518575,47528169,47512261,47515368,47513029,47515329,47513727,47512073,47512357],"score":13,"text":"$17","time":1774400427,"title":"Ask HN: How do you offload all coding to AI?","type":"story"}],[47494696,{"by":"noemit","descendants":22,"id":47494696,"kids":[47498409,47496895,47569769,47540979,47518087,47561130,47540200,47499371,47499342,47499000,47548754,47508348,47532964,47526878,47508063,47498203,47500805,47522625,47497365,47498274],"score":41,"text":"I ran an experiment to see if CLI actually was the most intuitive format for tool calling. (As claimed by a ex-Manus AI Backend Engineer) I gave my model random scenarios and a single tool "run" - i told it that it worked like a CLI. I told it to guess commands.

it guessed great commands, but it formatted it always with a colon up front, like \n:help\n:browser\n:search\n:curl

It was trained on how terminals look, not what you actually type (you don't type the ":")

I have since updated my code in my agent tool to stop fighting against this intuition.

LLMs they learn what commands look like in documentation/artifacts, not what the human actually typed on the keyboard.

Seems so obvious. This is why you have to test your LLM and see how it naturally works, so you don't have to fight it with your system prompt.

This is Kimi K2.5 Btw.","time":1774297908,"title":"LLMs learn what programmers create, not how programmers work","type":"story"}]]

user:	paulwelty
created:	March 25, 2026
karma:	2
about:	Philosophy PhD. Management consultant. Building AI agent systems. Writing about what happens when organizations meet AI: paulwelty.com