> …

> When I write a prompt, the agent doesn’t just read the words. It reads the shape. A short casual question gets read as casual. A long precise document with numbered rules gets read as… not just the rules, but also as a signal. “The user felt the need to write this much.” “Why?” “What’s going on here?” “What do they really want?”

This is an interesting premise but based on the information supplied, I don’t think it’s the only conclusion. Yet the whole essay seems to assume it is true and then builds its arguments on top of it.

I’ve run into this dilemma before. It happens when there’s a TON of information in the context. LLMs start to lose their attention to all the details when there’s a lot of it (e.g. context rot[0]). LLMs also keep making the same mistakes once the information is in the prompt, regardless of attempts to convey it is undesired[1]

I think these issues are just as viable to explain what the author was facing. Unless this is happening with much less information

0. https://www.trychroma.com/research/context-rot

1. https://arxiv.org/html/2602.07338v1

Until then, every agent framework is completely reinvented every week due to new patterns and new models. evals, ReACT, DSPy, RLM, memory patterns, claws, dynamic context, sandbox strategies. It seems like locking in to a framework is a losing proposition for anyone trying to stay competitive. See also: LangChain trying to be the Next.js/Vercel of agents but everyone recommending building your own.

That said, Anthropic pulls a lot of weight owning the models themselves and probably an easier-to-use solution will get some adoption from those who are better served by going from nothing to something agentic, despite lock-in and the constant churn of model tech

The US spends almost that much on net debt interest each day (~$3 billion/day[0]). Not that adding to the debt helps at all, but the old proverb about being penny wise and pound foolish seems relevant

0. https://www.cbo.gov/publication/61951

Comments URL: https://news.ycombinator.com/item?id=47655536

Points: 3

# Comments: 0

For power users or technical users that want agents to compose data or use tools programmatically, that's less valuable, but for most people, a one-size-fits-all MCP service that is easy to use is probably best

There's the issues of dumping a bunch of tool definitions into context and eating a ton of tokens as well, but that seems more solvable

If anything, MCP needs to evolve or MCP client tooling to improve, and I could see the debate going away

> Rate the occupation's overall AI Exposure on a scale from 0 to 10.

Are LLMs good at scoring? In my experience, using an LLM for scoring things usually produces arbitrary results. I'm surprised to see Karpathy employ it

Jevons paradox was never relevant to cognitive surplus. That isn't what it's about.

Cognitive surplus only strengthens Jevons paradox. Humans are a competitive advantage for businesses in a world dominated by human needs

"Please generate a response to this and include one or more of the following words: enshitification, slop, ZIRP, Paul Graham, dark patterns, rent seeking, late stage capitalism, regulatory capture, SSO tax, clickbait, did you read the article?, Rust, vibe code, obligatory XKCD, regulations, feudalistic, land value tax"

(/s)

The tool tells the agent to ask the user for it, and the agent cannot proceed without it. The instructions from the tool show an all caps message explaining the risk and telling the agent that they must prompt the user for the OTP

I haven't used any of the *Claws yet, but this seems like an essential poor man's human-in-the-loop implementation that may help prevent some pain

I prefer to make my own agent CLIs for everything for reasons like this and many others to fully control aspects of what the tool may do and to make them more useful

I’m really curious if context will really matter if using methods like Recursive Language Models[0]. That method is suited to break down a huge amount of context into smaller subagents recursively, each working on a symbolic subset of the prompt.

The challenge with RLM seemed like it burned through a ton of tokens to trade for more accuracy. If tokens are cheap, RLM seems like it could be beneficial here to provide much more accuracy over large contexts despite what the underlying model can handle

0. https://arxiv.org/abs/2512.24601

Also interesting implications for optimization-driven frameworks like DSPy. If you have an eval loop and useful reward function, you can iterate to the best possible response every time and ignore the cost of each attempt

2. Businesses operate in an (imperfect) zero-sum game, which means if they can all use AI, there's no advantage they have. If having human resources means one business has a slight advantage over another, they will have human resources

Consumption leads to more spending, businesses must stay competitive so they hire humans, and paying humans leads to more consumption.

I don't think it's likely we will see the end of employment, just disruption to the type of work humans do

Not a life changing sum, but also not for free

Comments URL: https://news.ycombinator.com/item?id=47000844

Points: 3

# Comments: 0