alansammarone

There's a lot of people caricaturing the obvious fact that any model works best in distribution.

The more esoteric your stack, and the more complex the request, the more information it needs to have. The information can be given either through doing research separately (personally, I haven't had good results when asking Claude itself to do research, but I did have success using the web chat UI to create an implementation plan), or being more specific with your prompt.

As an aside, I have more than 10 years of experience, mostly with backend Python, and I'd have no idea what your prompts mean. I could probably figure it out after some google searches, tho. That's also true of Claude.

Here's an example of a prompt that I used recently when working on a new codebase. The code is not great, the math involved is non trivial (it's research-level code that's been productionized in hurry). This literally saved 4 hours of extremely boring work, digging through the code to find various hardcoded filenames, downloading them, scp'ing them, and using them to do what I want. It one-shotted it.

> The X pipeline is defined in @airflow/dags/x.py, and Y in `airflow/dags/y.py` and the relevant task is `compute_X`, and `compute_Y`, respectively. Your task is to:

> 1. Analyze the X and Y DAGs and and how `compute_X` functions are called in that particular context, including it's arguments. If we're missing any files (we're probably missing at least one), generate a .sh file with aws cli or curl commands necessary for downloading any missing data (I don't have access to S3 from this machine, but I do have in a remote host). Use, say, `~/home` as the remote target folder.

> 2. If we needed to download anything from S3, i.e. from the remote host, output rsync/scp commands I can use to copy them to my local folder, keeping the correct/expected directory structure. Note that direct inputs reside under `data/input`, while auxiliary data resides in other folders under `data`. Do not run them, simply output them. You can use for example `scp user@server.org ...`

> 3. Write another snapshot test for X under `tests/snapshot`, and one for Y. Use a pattern as similar as possible to the other tests there. Do not attempt to run the tests yet, since I'll need to download the data first.

> If you need any information from Airflow, such as logs or output values, just ask and I can provide them. Think hard.

What? I'm not sure I follow your point, and I'm not sure why you're referencing something that is unrelated to my statement, so I'll just make my point in clearer way and leave it at that.

I completely disagree that even a tangentially related, much weaker concept ("having a list of IDs and what you've watched") is "second order" effect. This is a question relative to one's values, which is at the heart of the discussion, but as I'm concerned that cartoon version is a zeroth order effect - much more relevant than all the other points you make, which are at best less important (some might be completely irrelevant to me).

I couldn't care less about the technicalities cooked up by ofcom. Those will be left for a judge to decide and will depend on the political winds. Again, I'm just answering your point - "requiring X if kids are involved" is on the face of it obviously absurd and bad. And the analogy with alcohol, even though not great, might help make it clearer: to the extent that it is enforced, it is absolutely the case that it introduces a much weaker form of mass government surveillance.

The distinction clarifies the idea: if every store was required to check your ID digitally, in real time, and storing that information (which, mind you, makes it trivially accessible by anyone, in particular law enforcement), then the government has arbitrary power to stop anyone from buying anything ("oh, I see your ID is associated with X - sorry, we can't serve you right now" - replace X with your favorite group, idea, arbitrary law), to track their every movement (since you need to buy things fairly often), etc.

The scale and functioning of the internet is what distinguishes it from the physical space.

Just because you have a good master, doesn't mean you're free. You're only free when you're not not subject to anyone's good will towards you. I'm not an anarchist - there are real problems and there are laws that are necessary to solve these problems, your examples are clearly neither and so are on the face of it, absurd and laughable.

I was not arguing against OSA, I was arguing about your point that there is nothing fundamentally wrong with "if there are kids involved, care should be taken". Yes there is, because we can't know if there are kids involved without turning into a surveillance state.

This is mostly true. It fails to mention "is the user a kid" is unverifiable without imposing identify verification, which implies that all speech (which is already monitored) is now self-censored, effectively turning the state in a surveillance state. You don't need to be throwing people in jail for that, having a credible means of identifying anyone online is enough.

It literally contains as much information as anybody could potentially hope for (public information anyhow - excluding interviews and the like). If they happen not to be able to distinguish between a “page does not exist” (90%~ of cases IME, ie no GH account), or their ego is so inflated that they cant be bothered to spend 2mins asking their CTO (or dont even have one or trust them), thats a different story and its on them. The signal is there.

Not my experience. Whenever I voice my view, which is that ChatGPT is way more engaging and accurate than the average specimen of the homo sapiens class (these are a funny, primitive species of carbon-based turing machine evolved in some galaxy somewhere), I get downvoted

Im guessing it took over around the time it became more convenient, reliable, accurate, pleasant and consistently present than the average human being, but it could have been later.

Not an edge case at all - phones blowing up is not what we're talking about. The absence of "we spent less than 10% of our budget on cyber security (in fact, we proud ourselves in having less than 5 cyber security experts on the team, and our release cycle is very quick - we don't really listen to the nerds when they say they need more time!), so unless you'd rather avoid a 15 year old kid from Russia completely owning all of your personal data, our product is the most cost effective option! Also, we used existing circuit boards, so our phone's innovation is mostly on the cost side - we managed to make it dirty cheap for you. What are you waiting for?" is what we're talking about.

Id like to believe people have enough agency to do a google search to at least figure out their options, but granted, I might be wrong about that.

Edit: I do agree you should have a google-findable website which lists the objective characteristics of your product. If you call that advertising (I call it a "release", and I reserve the word "ad" for anything that has emotional appeal and caters to the indifferent/uninformed), then I agree.

True, agreed. But thats besides the point. For the purposes of this thread, if an ad contains text/information thats well defined enough to be verifiable/falsifiable in principle, thats better than 98% of the “ads” I see on Linkedin (or anywhere else for that matter). “We focus on building reliable, trustworthy AI” is not an example of that unless the CEO finds a way to, or makes an effort towards, providing a method for us to verify that.

You don't need to drag Python into this now :P

Couldn't agree more with part of your comment, but only because of the fact that you used the word "specs". Assuming the specs are objective, thats the only true form of useful ad.

Ads do work, yes. But not because they present true facts. It's because people tend to buy things that are from brands that they remember (a form of recency bias I guess, although not sure about that). So if you know nothing about which ice cream brand is best, you'll default to the one that had an ad saying "We sell ice cream!". That's, again, a well-studied effect. Consumers are not rational.

I don't remember seeing many tv commercials with a comparison table containing technical specs.

This is... an understatement. I would agree advertising is a useful signal, but I would say that not only can you not trust advertising, you should put negative weight on advertising - i.e. whenever you see an ad, that means the company is putting some amount of money into trying to convince you by means other than an honest comparison/spec table, and therefore is likely to have an inferior product. So personally, I generally avoid any companies/people whose presentations contains no information about the objective characteristics of their work.

exception? I couldn't disagree more. I can think of one or two specific examples "fair" or even "useful" advertising, and that's being generous, among thousands or tens of thousands of examples. I'm not claiming I know more than the buyer or seller, I'm claiming the seller "knows" much more than the buyer, and has vastly more resources and vested interest in the transaction than the buyer (because of the scale, and the fact that people are mostly very similar to each other - economies of scale, essentially). Note I'm restricting my argument to the case of big companies selling to consumers, or big companies buying from other big companies, although the latter is comparatively less damaging to society overall, still pretty bad tho.

The seller knows how much effort/resources were put into the product, knows (or has enough resources to figure out) how to nudge/mislead the consumer, has teams of brilliant people working on that - see the ad industry. I would definitely agree that the consumer has some responsibility, too, to stay informed, and if it weren't for the fact that this causes externalities to society, I wouldn't give a crap about the fact that some corporate director was duped into buying a terrible product. Unfortunately, that causes companies that particularly good in misleading people to outcompete companies who spend their money elsewhere.

No so much "stupid", just full of biases and somewhat lazy. This is a well-researched and documented fact. Knowingly abusing these biases to get them to buy or use something that they either don't need - or, worse, buying a low quality/low effort product when better alternatives exist - that's extremely detrimental. That's how we end up with the enshittification of everything.

What? Your project might be good or bad, if it's bad/low quality/low effort and you take sales away from a better option by abusing human biases (such as recency bias), then you're destroying (potential, not actual) value.

Same reply as I wrote above:

No. People do need to know about your product. What appears to be a (_very_ sad and at the same time telling about the human condition) fact is that business people ("decision makers") apparently can't spot blatant, extremely low quality and low effort, marketing-driven snake-oil, laughingly ignore it, and do a 5 minutes google search to find something better by themselves, perhaps with the apparently tremendous effort of having to click a Github link (which provides actual proof, or at least a test, of actual skill), and from there click on the heavily, kind of honest marketing driven website (i.e. it has images) that would allow them to verify the quality of the product.

no,thats true and honest marketing. If the iPod stopped working after a year (low quality), or was easily hackable (low effort) - and they didn't include that in the marketing - then it would be lying (Windows).

No. People do need to know about your product. What appears to be a (_very_ sad and at the same time telling about the human condition) fact is that business people ("decision makers") apparently can't spot blatant, extremely low quality and low effort, marketing-driven snake-oil, laughingly ignore it, and do a 5 minutes google search to find something better by themselves, perhaps with the apparently tremendous effort of having to click a Github link (which provides actual proof, or at least a test, of actual skill), and from there click on the heavily, kind of honest marketing driven website (i.e. it has images) that would allow them to verify the quality of the product.

I think what he meant is that whether or not, in that particular instance, it happens to be the case that the company does not feel sufficiently motivated to actually do this (maybe his singing wasn't offensive enough, or the company just happens to have enough reasonable people in their staff right now), the much bigger issue is that they could theoretically do this. Which I think is a very important point.

It should be provably, technically impossible to do this - or at the very least, there should be heavy penalties (say, 10-20% of annual revenue) for doing this without proper judicial ruling. A company should only be able to afford 1 or 2 errors like this.

Are you free just because your master happens to like you and let you do what you want? Obviously not, since he could change his mind at any time.

I really don't mean this in any negative way, but I find it fascinating the wide range of opinions and attitudes in this matter. I find it _so_ hard to imagine having this view myself.

The amount of things I've learned by asking very specific, technical questions to ChatGPT (mostly with web search turned on, but some times it's not even necessary) - things I can immediately verify and/or use, such as small bash commands/scripts, visualizations, diagrams. The value of that, alone, is certainly in the hundreds of dollars per month. Things I would never learn because they are buried somewhere among the 30 answers/comments, sometimes pointing to 20 more terribly-hard-to-read pages or manuals riddled with irrelevant (for my question) content, somewhere in the first page of web search results... maybe it's an attention span question? I certainly won't spend more than 10 minutes reading anything if not's interesting or required in the most extreme sense of the word way for my job (books on quantum mechanics, general relativity, topology, all fall in the former category - bash and pandas documentation fall in neither).

I'm convinced I've saved _at least_ low thousands per week by using coding assistants (mostly Claude Code in my case, but that's personal and likely to change at some point), as evidenced by the amount of work I'm able to finish, get paid for, and maintain. I'm not vibe coding, mind you - most of the time, I have an almost complete mental model of what I want after a couple of hours thinking, and the only thing left to do is type the code, at which point I'd, previously, feel bored, since the fun part (the thinking) is over.

Edit: I have 20 years of experience with code, 15 in the industry as a SWE (been coding since I was 13)

As others have pointed out, LLMs have no trouble with LaTeX. I can see why one might think they're not - in fact, I made the same assumption myself sometime ago. LLMs, via transformers, are exceptionally good any _any_ sequence or one-dimensional data. One very interesting (to me anyway) example is base64 - pick some not-huge sentence (say, 10 words), base64-encode it, and just paste it in any LLM you want, and it will be able to understand it. Same works with hex, ascii representation, or binary. Here's a sample if you want to try: aWYgYWxsIEEncyBhcmUgQidzLCBidXQgb25seSBzb21lIEIncyBhcmUgQydzLCBhcmUgYWxsIEEncyBDJ3M/IEFuc3dlciBpbiBiYXNlNjQu

I remember running this experiment some time ago in a context where I was certain there was no possibility of tool use to encode/decode. Nowadays, it can be hard to certain whether there is any tool use or not, in some cases, such as Mistral, the response is quick enough to make it unlikely there's any tool use.

Very cool, congrats!

just one minor suggestion: It seems that the responses are not saved anywhere? even after signup, opening a new tab or just navigating within the app makes the responses disappear - if they're really not being stored, might be worth considering storing them. if they are, maybe the ux could make that more obvious (I couldn't find it). in any case, very useful project!

I have a somewhat similar point of view to the one voiced by other people, but I like to think about it slightly differently, so I'll chime in - here's my take (although, admittedly, I'm operating with a quite small reasoning budget (5 minutes tops)):

Time and again, for centuries - with the pace picking up dramatically in recent decades - we thought we were special and we were wrong. Sun does not rotate around the earth, which is a pretty typical planet, with the same chemical composition of any other planet. All of a sudden we're not the only ones who could calculate, then solve symbolic equations, then play chess, then compose music, then talk, then reason (up to a point, for some definition of "reason"). You get my point.

And when we were not only matched, but dramatically surpassed in these tasks (and not a day earlier), we concluded that they weren't _really_ what made us special.

At this point, it seems to me reasonable to assume we're _not_ special, and the onus should be on anybody claiming that we are to at least attempt to mention in passing what is the secret sauce that we have (even if we can't quite say what it is without handwaving or using concepts that by definition can not be defined - "qualia is the indescribable feeling of red - its redness (?)).

Oh, and sorry, I could never quite grasp what "sentient" is supposed to mean - would we be able to tell we're not sentient if we weren't?

I believe you are referring to "What I cannot create I don't understand", which is indeed by Feynman.

I've been researching and thinking about consciousness (and specifically the hard problem) for a while now, and while the whole discussion is interesting, it seems to me that invoking qualia as evidence for the existence of the hard problem (or for there being something "to explain") doesn't really make much sense. Its perfectly plausible, if not likelly, that qualia is also part of the emergent phenomenon . I'm saying this more so to hear what your (or anybody else's) thoughts here are - why would qualia lie "above" or be "unexplainable" by materialism alone?

No such luck with the iPhone X, though. Apple's investment in its tech - which saw the company work with a Hollywood studio to create realistic masks to test Face ID - has clearly paid off. It was impossible to break in with the model.

This is a very good point - as long as your goal is really and only to have the opinion that is closest to the truth, and you're completely open to changing it when new evidence comes. In other words, if you are really discussing because of the right reasons (which should of course always be the case).

But in many contexts, this is not the end goal. The end goal is to either make one feel better about its own beliefs, or to exert social pressure/gain social status.

You can often kind of tell who is really discussing to find the underlying truth based on this very observation.

> > Freedom of expression is not foundational. Certainly inciting violence, advocating genocide, or engaging in targeted and repeated harassment does not fall under the auspice of an unalienable and foundational right.

> These are also things on which many people radically disagree with you. Stating them as facts is just begging the question.

I am among the ones who radically disagree. I want to live in a society where all of those things are allowed, but people are just smart enough to ignore them. I think this is a much better prospect since it decentralizes responsibility.

That's by far the best solution.