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

> I believe in the full utilitarian calculus it’s better to have a few people get verbal lashings than have many people lose their livelihood.

But this is a false dilemma. You can have both civilized discussion instead of verbal abuse and have people keep their jobs. It's just as simple as "You risk losing your job if..." rather than "I'll rip your arms off..." etc.

I worked with both these kinds of PHBs and in general the latter had worse staff as the talent preferred to work in more civilized environment. (But these were engineers, maybe middle management doesn't care.)


The reason why Google was #1 was they had a product that was ultrafast and without ads, so anybody who saw it switched instantly, there was no reason to look for another option. As for office suites, Microsoft won because of their aggressive tactics regarding format compatibility.

None of these apply to LLMs. I do enjoy cheap subscription to use frontier models, but if tomorrow I need to switch to the so-called open models, I won't regret it that much.


The lock-in you're missing here is that businesses use LLMs in conjunction with collaboration. They're building one-off apps that live inside the Claude GUI and sharing them with teammates.

To move that to an open model, you need to replicate the software surface (collaboration, functionality, permissions, etc.) and then provide a way to migrate the existing artifacts/data/history. Migrating developers is the easy version of the problem, migrating everyone else is potentially more difficult.

And then, it has to be trustworthy. Do I trust the provider with my sensitive business assets? If we are spending (say) $50k/yr on this, is it worth me risking my job by switching the company to a no-name brand to save $25k/yr?

The lock-in effect is stronger because so many engineers don't think it exists at all.


>That doesn't seem to make much sense

Neither does SIM cards presented as The Sims games:

https://nypost.com/2022/04/25/russia-appears-to-confuse-the-...

But it doesn't matter, they do it anyway and are quite effective as the far right has considerable support in various parts of Europe.


Not entirely. In the past they would go from free to paid Figma account when in need of more professional approach. Now they can't unlock Claude Whiteboards become a professional tool, so they still need to migrate to Figme when their needs cross a certain threshold.

> I'd argue it is the other way around - LLMs are great at pure creativity

Interesting because pure creativity is the most prominent thing they can't do by design.


There's no such thing as "pure creativity," which I presume you mean "creative output not directly derived by prior inputs." The very concept is utter nonsense.


turn up the temperature, the outputs will be very "creative" ;)


It would be great net result for our brains, however too much money has been poured into this business to give up on that. My belief they will push for SWE layoffs trying to position LLM-based agents as a cost efficient alternative to human programmers. Which is extremely stupid and short-sighted but that's the strategy.


They will learn certain lessons the hard way for sure. As much as I want to spin up 100 agents in a loop on our codebase and let them loose, I also know its not feasible as they can easily go down the wrong rabbit holes or over-engineer the wrong solution. They are amazing but cannot be left 100% autonomous just yet.


> just yet.

How long, in your estimates, do you think you have before your career ends?


There are different names for this, such as The Last Mile Problem or the Ninety-Ninety Rule, but I wouldn't hold my breath.


Who’s “them” and is there really one homogenous strategy?

Companies will do different things, just like they did with mobile, the internet, and computers in general.

Some strategies will work, some will fail.


"Them" are the management of frontier labs, especially OpenAI and Anthropic as they have no other products than their LLMs and they do have a fairly consistent strategy aimed at convincing large investors that (1) something that doesn't exist yet is bound to exist, and that (2) their products are far better than they really are.

This is fairly common SV strategy, used by Tesla and others, and it works extremely well in spite of people pointing out the obvious weak points.


I mean isn’t that any startup strategy in a nutshell? Why would investors invest in something uncertain if the product already exists and the new company can’t add anything to it?


> The problem I find is that no one writes C# libraries that actually do anything

Yes, because when you think about it, why should they?

MS has showed many times how unreliable and even user-hostile they can be, so the community at large is very wary of their ecosystem in spite of real advancements.


at least supply chain issues are less problematic if everyone depends on MS :)


I assume 17% is roughly the percentage of people that use LLMs globally (the numbers vary between 15 and 20 percent).


Amazingly low, given potential impact. Where is this use concentrated? Who has the best data on use right now.


It's one of these rare cases where intention is not what is the most important - the net benefit for consumers and companies outside of the USA is indisputable.


Developers with enterprise credentials sprinkled throughout their environment running anything from the Internet deserve what they get.

But I assumed the intended audience are home users with entry level macbooks/minis with 128 GB RAM where this patch actually helps them.


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

Search: