Mtinie

AI adopted humanisms, we just weren’t used to seeing them at the same scale we do today.

Diversity of writing styles was part of that, but I’d point to vernacular exposure as the larger component. We’re going to go through a period where we try to adapt to a form of “Universal English” for those of us who read primarily English writing.

Other languages’ readers may be experiencing the same dissonance when they come across AI-generated prose in their native language (but I’ll let others validate /reject my hypothesis).

qregnant

I interpreted the comment you are responding to as “people want to do what they believe is the right work” rather than simply the mandated work (which they might fundamentally disagree with, directionally). When people have limited agency, apathy increases.

What was your expectation? That your prompt would trigger a web search, first, before the introspection of past conversations and a training set recall?

How did Sonnet 4.6 respond that was objectively better for your use case?

That ship has long sailed. You’re correct, but the author isn’t the one who “named the thing” in this case, they are just using the name commonly used to describe it.

Multi-rotor drones have been called tricopters, quadcopters, hexacopters, octocopters based on their propeller counts conversationally for as long as I can remember.

There are plenty of commercial vendors who use the exact term for their expensive industrial drones.

Update: I see that in the four minutes it took for me to validate my initial inclination and post that plenty of others also had the same thought :) No need to me to belabor the point!

In this case, I’d expect it should make a web search tool call to find the Python library best suited for SVG generation and manipulation, and then use what it learns there to execute the task you’ve asked it to do (either asking if you’d like to incorporate the library as a dependency or to roll its own implementation of a subset of the features if that was your preference),

Assuming tool calling hasn’t been entirely stripped out of this model.

(Edit) No tool calling, per this comment: https://news.ycombinator.com/item?id=48640189

In some cases, I’m sure you are right. That said, the online communities which currently employ shadowban systems continue to use them. That tells me there’s a value greater than their implementation and operation costs.

There is no perfect tech solution to a human problem. But in my opinion, having access to a partial mitigation is better than no mitigation.

A slur is rarely the problem. Trolls can (sometimes) add flavor to a community and keep things interesting. Mostly they can be an annoyance.

I’m specifically referring to people who have seemingly made it their sole purpose to create as much indiscriminate damage as possible.

You can ban them, block routes for them to attempt to Sybil themselves back to having accounts, etc. but even with great moderation tools and systems, it’s extremely difficult to set up a strong enough set of controls which don’t adversely impact everyone else who you want to have participate in the community.

Yes, a shadow environment is dystopian. It’s not my nature to want to even consider using one.

But we’re talking about privately run communities which also deserve to exist to serve their purposes.

So given the choice between anarchy which drives away people who contribute to make the community what it is and a shadow option for those actively working against its interests, I’ll consider the community first.

You may have misinterpreted my comment. I’m not suggesting you use LLMs as moderators. I’m talking about using LLMs as participant “members” of this shadow board to interact with someone whose account was flagged by a human moderator.

After having been a moderator for a number of communities over the years—BBCs in the 90s, niche forums during the 00s & 10s, Reddit boards in the 20s—I’ve come to believe that running shadow services where disruptive individuals can comingle may be the best way to handle this.

The advent of LLMs really opens the door to shunting off these “community members “ who’d rather contribute in misanthropic ways for the lulz than either leave or not contribute at all. They can take part in an interactive echo chamber that gives just as well as they can. You don’t even need a powerful model so the overall costs to the community are probably lower than the alternatives of trying to coexist with community-arsonists.

I spent years trying to find ways to bring people productively “into the fold” but eventually realized that it is futile in some cases because there’s zero value to the individual or the community to find a middle ground. They want to see things burn, and the community simply wants them out.

A few years ago, a facsimile of Imperial Rome staging munera at the Colosseum on the White House lawn would have been unimaginable.

The attendees seemed like they enjoyed it and there were impressive athletic performances. Panem et circenses.

In all seriousness, wasn’t that always the case? Writing bad code is relatively cheap.

Ensuring code isn’t bad is the expensive part.

“…used for religious purposes and/or tribal rituals.” (Far-future Archeologist, Zarb-7854)

Which, joking aside, isn’t too far from the truth.

Why would the U.S. government want to sabotage prediction markets when the chief executive has a vested interest in keeping them legitimate?

What would you do differently if LLM outputs were deterministic?

Perhaps I approach this from a different perspective than you do, so I’m interested to understand other viewpoints.

I review everything that my models produce the same way I review work from my coworkers: Trust but verify.

…and a concise one at that. Which is more impressive given the (general) penchant to say less with more.

https://broadbandbreakfast.com/dateline-ashburn-data-centers...

“Loudoun County currently has 199 data centers, with another 117 in development, according to Michael Turner, vice chair of the board of supervisors transportation and land use committee and Ashburn’s district supervisor.”

https://virginiabusiness.com/loudoun-county-advances-changes...

What is your bar for “useful”? Let’s start there and we’ll see what evidence can be offered.

User count? Domain? Scope of development?

You have something in mind, obviously.

It would be helpful if you could define “useful” in this context.

I’ve built a number of team-specific tools with LLM agents over the past year that save each of us tens of hours a month.

They don’t scale beyond me and my six coworkers, and were never designed to, but they solve challenges we’d previously worked through manually and allow us to focus on more important tasks.

The code may be non-optimal and won’t become the base of a new startup. I’m fine with that.

It’s also worth noting that your evidence list (increased CVEs, outages, degraded quality) is exclusively about what happens when LLMs are dropped into existing development workflows. That’s a real concern, but it’s a different conversation from whether LLMs create useful software.

My tools weren’t degraded versions of something an engineer would have built better. They’re net-new capability that was never going to get engineering resources in the first place. The counterfactual in my case isn’t “worse software”—it’s “no software.“

GitHub, Bitbucket, GCE, AWS…all have licensing agreements for user contributions which the user flagged as “public” so I’m not exactly clear of your point if you are holding SO up as a bastion of intellectual property rights different from the other places LLM training sets were scraped from.

Recklessly is a strong word. I’ll give you the benefit of the doubt and assume your comment in good faith.

How do you describe the “reckless” use of information?

I generally agree with your underlying point concerning attribution and intellectual property ownership but your follow-up comment reframes your initial statement: LLMs generate recombinations of code from code created by humans, without giving credit.

Stack Overflow offers access to other peoples’ work, and developers combined those snippets and patterns into their own projects. I suspect attribution is low.

If a human did this we probably would have a word for them.

I don’t think it’s fair to call someone who used Stack Overflow to find a similar answer with samples of code to copy to their project an asshole.

It’s Agile philosophically, and how it should be.

But that is rarely how it works. In the dozens of different projects across ten or twelve companies I’ve had insight into, “doing Agile” is analogous with “we have a scrum master, hold stand ups, and schedule iterations” while the simple reality is “Agilefall.”

Both camps, for different reasons.

For novices, LLMs are infinitely patient rubber ducks. They unstick the stuck; helping people past the coding and system management hurdles that once required deep dives through Stack Overflow and esoteric blog posts. When an explanation doesn’t land, they’ll reframe until one does. And because they’re confidently wrong often enough, learning to spot their errors becomes part of the curriculum.

For experienced engineers, they’re tireless boilerplate generators, dynamic linters, and a fresh set of eyes at 2am when no one else is around to ask. They handle the mechanical work so you can focus on the interesting problems.

The caveat for both: intentionality matters. They reward users who know what they’re looking for and punish those who outsource judgment entirely.

As I think more on how this could work, I’d treat the fully commented code as the source of truth (SOT).

1. SOT through a processor to strip comments and extra spaces. Publish to feature branch.

2. Point Claude at feature branch. Prompt for whatever changes you need. This runs against the minimalist feature branch. These changes will be committed with comments and readable spacing for the new code.

3. Verify code changes meet expectations.

4. Diff the changes from minimal version, and merge only that code into SOT.

Repeat.

1. The raw code with no empty space or comments. 2. Code with comments

I like the sound of this but what technique do you use to maintain consistency across both views? Do you have a post-modification script which will strip comments and extraneous empty space after code has been modified?

In this scenario the person who wants to be paid owns the output of the agent. So it’s closer to a contractor and subcontractor arrangement than employment.

If it’s done in a controlled manner with the ability to revert quickly, you’ve just instituted a “scream test[0].”

____

[0] https://open.substack.com/pub/lunduke/p/the-scream-test

(Obviously not the first description of the technique as you’ll read, but I like it as a clear example of how it works)

In my experience that “blink of an eye” has turned out to be a single moment when the LLM misses a key point or begins to fixate on an incorrect focus. After that, it’s nearly impossible to recover and the model acts in noticeably divergent ways from the prior behavior.

That single point is where the model commits fully to the previous misunderstanding. Once it crosses that line, subsequent responses compound the error.

From my viewpoint you are conflating software quality with ambition. All software develops iteratively. Tools now celebrated for quality and consistency (commercial and OSS alike) shipped from states where they were neither. Jerm-CAD existing gives it a shot at improvement. The alternative is it doesn’t exist.