mw888

Those most-intelligent of humans who carefully memorized verbally the most important civilizational facts known had the same reaction to the invention of writing.

You are clearly a wizard of a prior era. But it is a new era.

Tangential: I have no issue with it, but this seems by README writing style to be AI-assisted. As time goes on, whether AI was used or not will become far less a point of interest - if the project is good it is good.

The unceremonious use of (what I speculate is) AI-assistance is becoming more normal. I think a lot of people who were die-hard skeptics will just quietly protest less and less. Also, for people who use it well and poorly alike, the urge to not advertise any AI-use at all is common today. People who know how to use it well don't feel the need to explain themselves to unreasonable absolutist skeptics.

I'm not convinced a problem is actually here. I certainly wouldn't believe LLMs, an optional tool, make any part of this fundamentally worse. Maybe superficially for people who can't adapt.

That's the type of moral stance which doesn't encode well into reality. I don't even agree with it, but it is surely consistent enough from your end. The real issue is that it crosses so many practical legal boundaries, mainly around intent that it will never survive as representative of your actual morals—through the grinder that is the modern legal system.

You've conceded your point and picked another.

> they'll raze the ground to ashes before anyone can use it for legitimate means.

You were so cynical you thought this one category of people would literally prevent any fruitful use. Now you simply argue with me whether the category exists at all.

> I really have no idea what you're on about. There's TONS of examples of people abusing this already.

Grossly dishonest in spirit, you are.

> I can't wait to see Apple Verified™ photos of UFOs flying over the Golden Gate Bridge.

While I'm on board with you about the inabsolute security of this (relative to what's typically expected of cryptographic systems), the fact that their 'verified' state requires a live certification and can be revoked means that the sensor responsible for obviously faked images will see those images and that device no longer certified.

It all relies a lot on trust in Apple, and integration with Apple, and relatively unmotivated attackers.

Certainly not always. There's a hedonic adjustment which happens however, where some tasks go very smoothly without much specification and a lot of "you know what I mean" to the LLM, while others then require you to get painfully specific after it badly misinterprets your intent.

Or maybe you can just get too spoiled with it grokking your intent, then become so vague that your vague ideas are actually just bad ideas. Certainly has happened to me.

> In my work with LLM-included software, I built a tool that evaluates text output relative to a baseline of what's expected. It helps to ensure things don't drift over time.

Is that hotel example real? Curious how exactly you employ this technique—my naive idea was, if talking software development, a sort of 'sanity-check auto-linter agent' catch errors on a regular basis (every 10 seconds, every write, w/e).

Part of the boon of this will just be dealing with less employees.

Not so much reducing cost by reducing headcount, but reducing liability; it's the legal labyrinth which is the barrier to entry to scale business. While it is present elsewhere, it's universal in employment.

The real question is "how big can a business be before it requires a legal department?" That an AI can automate many rote tasks and have some expertise means the benefits of scale with less of the risk.

Oh please. Not only is this needlessly cynical but it demonstrates no ability to think for yourself. "Broad categories I don't like are excited about this, therefore it won't work."

Of course you'd feel no need to justify all the exceptions when people you don't like are excited about a movie, food item or video game. Yuur justification for cynicism is vacuous, which will obscure realistic concerns.

So are you, dear reader, for intellectual property laws or not. I see a lot of "anti-this" sentiment confused with "anti-AI [data collection]."

The consistent view is one regarding IP in general - not reverse engineering your morals based on the case.

'Prediction' gets overloaded with optimization. Predictions are binary, optimizations are fuzzy.

If you're saying it's predicting, then each result should be falsifiable.

The result of an LLM output should be able to be scored against what it is supposedly predicting. Of course, that isn't possible, because it isn't predicting anything when giving novel outputs, otherwise that thing would exist independently.

Your argument of course sounds nice and fails under your inability to define "gambling."

I view favored house-odds, on arbitrary games or not, as unethical and predatory. This is the classic casino slot machine and related. However, Poker doesn't have house odds, it is a fair game.

Move one level up: prediction markets don't have house-odds if implemented plainly. On non-game events, they also have a positive externality, which already contradicts your claim: prediction markets predict quite well.

Yet another level up: investing in the stock market. The same authentic gambler who burns money into a slot machine can play the stock market to similarly disastrous ends. But if you define this as gambling and want to outlaw or limit it to 'professionals' then you bar people from capital markets which is absurd.

It does have to be said that if LLMs keep becoming better coders at some point the bottleneck on quality is prompting. Good ideas have many hidden assumptions you think are procedural but often are pivotal to your broader vision.

I find that when I give an LLM my full handcrafted codebase, it does very well. It follows my conventions, sees the intent and can coherently build within its scope. It writes much better code than a 'vibe' prompt.

It is always tempting and I myself will continue pushing the boundaries, but when you keep an LLM in reasonable scope (that may be one line, function, file at a time, depending on your idea of reasonable), you, by definition, can get sound utility out of them.

The intent was to collapse this meaning.

In the context of 'decentralized tech' I've come to believe literal 'decentralization' has unavoidably poor network effects, while 'accountability' gives what people envision from 'decentralization' without the tradeoffs—by being a more fundamental property (decentralization can emerge from it if required).

You can choose relays, and everyone using one (centralized) relay avoids the issues. And network effects push everyone to the same one.

Email requires a message to reach a specific, limited set of participants. Nostr is a public forum, which means an unbounded number of participants who can engage.

Ordering a conversation between an unbounded set across independent nodes leads to misordered interactions not possible if you just centralize. It seems like a small issue, but at scale quality really suffers when some universal, fine-resolution ordering can't be agreed upon.

You can devise many clever schemes to try to avoid centralization and achieve coherency, but if one service is more up to date with more info, it is preferable and offers a more coherent public conversation. Network effects still dominate despite the protocol.

All that said, if I wasn't very aligned with the spirit, I wouldn't care so much to think about it. I'm also aligned with the spirit of Bitcoin but the culture present in both doesn't help it. The fact that Nostr is Bitcoin aligned, meaning it rejects the technical options of protocols taking more risks, hurts it.

The way it handles public keys and identities is very simple and good and to me is very obviously reflective of a good future state. But it's not unique to Nostr.

Nostr breaks down along the longstanding ambiguity around what 'decentralization' means.

In the literal sense, centralization leads to economies of scale and the alleviation of coordination issues. Those coordination issues are what make physically decentralized networks so complicated, inefficient and fractured. These networks usually just find ways to centralize despite.

People really want decentralization of power—let's just replace the word with a better word: accountability. One should be able to enjoy economies of scale (centralized infrastructure) but have the cryptographic mechanisms in place to ensure the infrastructure must reveal its use and abuse of power, and can be easily replaced.

Nostr, like most 'decentralized' tech will switch between the meanings of both. Either way, it fits the description above: a single relay can work well, multiple have awkward coordination. The sense it which it is accountable is only that it can be replaced, but network effects favor the largest, so that's a problem.

ChatGPT subscriptions do the same weekly allowance, which the discounts apply to.

There's an amortization aspect as well.

If I'm going to be iterating on one document or idea in an extended manner with feedback from the chatbot, I will make an effort to setup the decorum it should follow, because it's a small proportion of the time in that chat.

But the guy just trying to get a report or email out quickly? Not so much.

The article reads like AI, which is a shame because the topic is at least in-theory compelling.

'Crypto' adopts to some extent a principle of being 'decentralized' and now has to square what it means for that type of hardware to compete with economies of scale which are naturally centralized.

"Decentralized" is just a bad metric. The real value-add of 'crypto' done right is accountability; Bitcoin mining is just the simplest example: any miner who builds less utilitarian blocks (either worse transaction selection or building on a bad history) is accountable by any other miner with the same margins who will maximize that utility, thus their income.

The urge to 'decentralize' is always in tension with that efficiency optimization. It is fine for an agent with concentrated power (under certain bounds) to be in a system, so long as they are accountable, usually by being easily replaceable.

Unfortunately a lot of 'crypto' intuition is just based on the idea of a wide breadth of servers or nodes equals some meaningful distribution of power, which is naive.

Generally constraining scope and providing enough existing material until the LLM is productive.

The common counter-argument is that specifying to the sufficient level is more work than just not using an LLM at all. I find that is not a universal rule.

Paraphrasing:

> Advertisement serves to increase purchase of the inferior product.

> The best product, which previously didn't require ads [on Amazon] now must pay Amazon to compete for eyeballs.

Advertisement, when successful, increases the purchasing of any product, generally. That the best product is subject to this effect and competition has nothing to do with Amazon or technology.

Consider also the utility of a weaker technology: Succinct Non-interactive Arguments of Knowledge. Theses can be ZK, but even if not they can take an expensive verification, like a hundreds-wide multisignature, and make it cheap.

I won't overclaim but look at "Binius" which uses fields, instead of prime orders, of orders of powers of two. The very intuitive notion is that computers are good at 2s, thus explaining their massive performance gains.

You're appealing to ambiguity. All you've said is you have failed—how is anyone supposed to know what went wrong?

Predict multiple outcomes, induct across them, refine.

Software developers are famous for lamenting the high time preferences of business forcing them to make poor engineering decisions. The dynamic is common.

Therefore one should not expect higher quality out of many businesses using AI, but you should certainly expect such businesses to highly value AI. The arguments in favor of quality didn't seem to work before, I don't think they will now.

This could allow a very opinionated and tailored translation, which I think is the appeal.

That said, I'd like to see the prompt(s) having such an effect on the subjective choices for a project like this.

I like Pi, but I didn't end up using it. I tried OpenCode, Pi, Zed Editor and some NVIM packages. I wanted open source, featureful and easy to use. Specifically I wanted to easily edit the agent prompt.

I ended up on VS Code. I'm very critical of Microsoft generally, but VS Code is a very good editor and my favorite agent harness.

For headless, Pi might be the way.