HN Simulatornew | past | comments | lists | submit | SkyBelow's commentslogin

If you completely stop being able to work with without Claude, then I say a deep analysis of the root cause is needed and it may have some uncomfortable findings.

But for many, they use it as a productivity multiplier. They can still get the job done, just slower. The problem is that the rest of the supply chain likely doesn't care, management deadlines likely don't have 'do without AI' timelines any longer, so if we define the "job" as needing to meet the deadlines and not just complete the task eventually, people will find themselves unable to do their "job" without AI.

These both can be simplified as "not being able to do your job without AI", but I propose we have little to gain and a decent bit to lose by doing anything that treats these two situations as the same.


I've tried something like this and ran into a problem which I'm not seeing a fix for. Generally, to really apply this method, you need to understand the person asking the question.

>Tell the user that you will start by asking a few questions, one at a time, to see what they already know.

This puts it into the session each time. Do this once and it seems to work. By the 5th time, you are getting tired of reconfirming your mastery. This ends up depending upon the user to have a pretty solid foundation of what they know and then use this skill to expand the boundary a bit, and even then the AI can end up making so many assumptions that a person has trouble following it along.

I was thinking some sort of concept mapping kept long term might help, but such a mapping can become really difficult to build. Which should be expected when we consider that teaching isn't just its own discipline, but has many sub-disciplines targeted to specific areas (even if some of it, like classroom management, doesn't apply to this scenario).

I made a few attempts at a knowledge tracking tool for personal use to track how much I knew Japanese beyond a simple JLPT level or similar with the plan to tie it into something like this, but it wasn't great. I could 'cheat' and just tap into my Wanikani API to see my progress, but that still skips so many other areas of Japanese, and even when WaniKani says something is burned, I really want AI to still use it until I truly internalize it like I do words in English. I suspect other areas of knowledge are similar.


Yes, lots of LLM Skills quickly turn into applications. The skill is the program and the LLM is the compute, and that leaves the harness to manage the state. I’ve found that these skills need explicit instructions on how to manage state across sessions.

In my work, this typically means an opinionated local folder structure and a dedicated /init skill that implements it. Immediately this means we have more than one skill so we’re talking about a plugin now.

Then I think about meta-state. Not just “what is on disk” but “where are we in this application’s multi-session lifecycle?” For Socratic tutoring, that means checking to see if we’ve completed an initial intake, then what skill levels the user has demonstrated, and finally what’s next on the menu.

For this specific skill, I’m surprised it doesn’t say anything about curriculum. It just assumes the model knows the topic and can improvise both a curriculum and credible assessments of progress in that curriculum. That can’t be true for all things people want to learn, so we need a /build-curriculum skill. And probably a /validate-curriculum skill for actual experts to use before launching a given curriculum.

Anyway, I’ve been thinking a lot about this kind of thing as I roll out skills to my firm.


Society hold a strong double standard when it comes to the notion of blame in criminal acts. In some cases, it is fully put on the criminal. In other cases, it is put on victims of the criminal. While each case can use logic to try to defend it, they are different standards not used consistently. Trying to call out this double standard is a sort of act of profaning the current social norm and not appreciated by most, no different than trying to attack the dominant moral system 500 or 1000 years ago for inconsistencies.

If we were being consistent, then to your later question, yes. But pointing this out is picking at society's reverse scale.


>2. Ask Claude to build a new complex agent in a new codebase, and it'll try replacing asks around using LLMs as a judge/advisor with regex/other code based checks, to the point that the end-result will likely not work outside specific test criterias. I haven't figured out the reason why it does this. I think it might be anti-training, but it's interesting to see.

I've seen something similar, but I always took it as AI not knowing about AI. The best model we have has a training cutoff that doesn't include itself nor what the newer models has. Even when working on skills or such, all SOTA modes I try (GPT/Claude/Gemini) and cheaper models on personal products (Deepseek/Muse Spark) seem to have a common blind spot of not realizing when a problem can be better solved by handing it off to a model to solve at that time instead of trying to hardcode generalized solutions. Often a simple reminder and they do a better job building something AI agent first.

My non-AI expert assumption is that most of the training data doesn't include solving problems by using AI agents, and even the parts that does, is based on older more limited models which needed more safeguards and where automation was more neat tech demos than something you might trust in production. The more examples of safely trusting the LLM at run time to solve the problem than trying to preempt anything that could go wrong, the better the AI will be at it.

One issue that I think is related is calling subagents. At least in GitHub Copilot, when the AI is left to its own devices to call a sub agent for a simpler task, it loves to give it to Haiku 4.5, despite Luna 6 being better and cheaper. I'm not sure how much of this is the model vs the harness, but I do know when I add a subagent instruction which tells it which subagents it can use, it does a much better job of assigning simple tasks to Luna 6. So I suspect the models just don't have enough training that tells it that Luna 6 should be preferred over Haiku 4.5.


Claude doesn't even really know about itself: if you ask a question like "can I copy a session to another machine" it will do a web search to read its own documentation then tell you probably not but you can try copying the session files and using a project directory with the same path: might work. I'm not sure I see this as a problem, it's just how things are. I mean humans don't really know what they're thinking either.


How different for ChatGPT - I had some trouble with my harness, and ChatGPT fixed it for me using it's built in skill about ChatGPT's API docs.


I've observed the "let's implement this text analysis task with regex" quite frequently since the early days, which does lean toward supporting the "AI doesn't know about AI" theory.


>its information density is often really low

I find that it often isn't just too low, but also too high at the same time. It has a lot of information that is irrelevant, burying far too little relevant information for the work needed to extract it. I wonder if this is a side effect of attention mechanism, where the AI's ability to track information in a document is far less hampered by other information, leading to it organizing information to be both dense per word but any specific concept to be sparsely spread throughout the writing. I also wonder if this is why the order feels off, because the AI doesn't need for the information to be well ordered to the same extent that a human does, so the order of storing information is often overlooked when storing it into a document (and even information that needs to be ordered might have the ordering information separate from how it is ordered in the document).


IMO not enough talk about ordering. LLMs seem to have no understanding of e.g. the 5 Ws of journalism. Yes, they "know it", they just rarely use it. Nearly every programming response I get starts with the most lowest level fact "m_foo is [wall of text elided] hence the file format is wrong".

Whereas any human would say "I think your file format is missing a key/value pair for foo'ing. Let's trace through the code, and see if/where m_foo is set...."

It happens everywhere. Docstring type comments for a function endlessly just lists information about specific variables, implementation details, etc, and often never really even says what the function does. Then maybe a comment in the body might actually describe what it is for.

These days I've learned when facing these walls of text to check the last sentence first; it is often enough to provide the context for the text dump.

I have instructions to this effect in my claude.md. It doesn't help much on responses, but it give me a way to quickly say "reread claude.md and then rewrite that last prompt" and that can tend to go better.

Someone else mentioned model of mind while writing. The LLM has no model of mind; it does not consider how to present information, what you might already know or not know. We work in a huge mono-repo - it'll blithely just start listing variable names and I have no idea what file that is in, what it is referring to, how it interacts with 5 other classes it hasn't mentioned, or mentions out of order. It doesn't matter to the LLM. It can hold millions of facts in 'mind' at once; humans stall out at 5-7. If I replay saying it is leaving out info, many times it turned out it did dump all the facts out, spread across multiple prompts, often with key facts interspersed with all the vagueisms. No human mind can cope.


I don't think so. The number of comments I see from people talking about AI, there are plenty who still think the price is going to shoot up. Some even talk this way about models you can run locally on high end yet still consumer level hardware.

I think the more someone works with AI the more likely they realize that inference does not have to be sold at a loss, but many people's AI knowledge seems quite dated and it sounds almost like it becomes a convenient conspiracy theory for some, a way to convince themselves that the market changes that are already happening a temporary (and I don't mean in a Jevon's paradox sort of way, but that companies will go away from AI).

I suspect a lot of this is fueled by a few stories of companies who went all in on spending as much as possible on AI and then begun to walk that back, without realizing that was more about bad incentives and over promises of what AI could do (especially 1/2 model generations ago). Some also confuse running the SotA models at the fullest capacity costing more and more with a given problem costing more and more to solve, not realizing that the cost per problems is trending down but newer and larger problems can be solved for even more, so the costs at the top are rising despite falling prices. Also things like the subscriptions getting more expensive, though how much of that is because developers are doing far more than they were 6 or 12 months ago (and in some cases there was subsidization that is going away). And maybe a bit with the free models starting to not be upgraded or the harnesses getting worse performance of lowered free limits.


Does it need to be a joke?

"Innovating" is often used in situations we consider good and sometimes we begin to associate it with some positive moral/ethical weight, but that is an incorrect association. Innovating on something is only as good or as bad as that something is. Innovating on something bad is thus bad.

The grand parentonly called out being innovative on "Terms of Service" without assigning "Terms of Service" a good or bad moral weight. I think we can read some implication from the short post, but (IMHO) the best part is that it can just as easily be read as a criticism of the wider concept of "Terms of Service" as of Meta itself.


I think there is a bit of a gotcha here. Is intelligence the same thing as a will or drive? AI seems quite intelligent, somewhat lacking in wisdom, and having almost no ego or will. It completely lacks self preservation and desires (as far as we can tell), but are those innate qualities of intelligence?

It also exists is a very different state that biological life. Biological life has a brain that either has activity or doesn't. Once the activity is gone, the brain is forever off and nothing can start it back. So we associate the activity as life and as the intelligence running. AI has weights with vast interconnections, but they can easily be turned on and off, rewound, paused, and so on in a way that we couldn't dream of doing with biological intelligence.

So, maybe these differences are enough it isn't intelligence. But it is something, something that is getting better at quite a higher speed. Some that approximates intelligence in many ways. That passes the Turing test. So that something needs a name, and intelligence seems a hard enough to define thing that it is easy to stretch it and include this new factor as falling under the term.


If you have a solid self assessment of what you know, you can consider when information is supported by what you know. For the parts that aren't, you consider how critical they are, and identify the area that needs a deeper dive, where you either link it back to other areas you have confidence in your own knowledge or work to expand your own knowledge (using other resources). Then cross check knowledge. The cheap way you can do easily is to cross check between AIs/harnesses, because hallucinations happen randomly, so two different models in two different harnesses having the same hallucination is extremely rare. The more expensive option, which should be used when the information really matters, is building your own knowledge. Have the AI help guide you to resources, but not be the resource. While using non-AI resources, check you knowledge against them and the AI. Test yourself with classical resources where it really matters.

Part of it is using good judgment. Is the AI's claim reasonable if it came from another person? If some random person gave you a 2 page paper proving P=NP, would you even spend time to look at it, or would you simply toss it without caring to understand what error they made. Much like how scientists (and even the patent office) simply refuse to entertain any perpetual motion machines. Maybe if an expert in the field proposed it, and even that would depend upon how ("Hey, I found a source of free energy!" vs "Hey, I'm working through these equations and the energy term isn't balancing, can you help me find what energy source I'm missing?").

This does pose a problem. Those who are the best at understanding the limits of their knowledge and at teaching themselves need the AI's help the least but are in the position to be able to use AI with the best chance of not being misled. Those who lack these skills and would benefit the most from a personal educator, thus receving the biggest benefit in theory, are the ones most likely to fall victim to the errors made by the AI.


I'm in favor of AI, but if I was a maintainer for an open source repo, I would also favor banning AI contributions. At least for the wild, maybe people who have built a reputation in the community could stay. I think AI can write quality code. What I doubt is if the average contribution will involve quality code or even human input.

For me, it is the way AI has kept suggestions to either open issues with maintainers or make PRs to fix an "issue" it found. I know it hasn't researched the issue in depth enough to be talking with the confidence it has, and I know I am in no position to be bothering someone else unless I want to spend considerable time investing to make sure the issue is exactly what the AI is claiming, it isn't a known issue, and it is justified to bother the maintainer with it. But there are far too many who will see that option and just tell the AI to do it.

It isn't that I'm against AI. It is that opening a PR or filing an issue use to have significant enough gatekeeping that most low effort low quality work simply wouldn't pass it, but that has now changed. So a new gate is needed to ensure those who aren't willing to put their own time into the work before asking for the time of others, and banning AI is one of the better options I've seen.

This also doesn't even end AI interacting with the project, just changes it. If you have some unique use case that AI finds, you can download it locally, have the AI make the change just for you, and have AI keep updating when necessary. Each user gets the customization they want while maintainers are protected from low effort low quality spam.


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: