The underlying issue here is the same as why I am generally opposed to large governments, entitlement programs, etc.

In the case of jobs, we are already overly dependent on individual incomes just to get by. We've collectively outsourced nearly everything a person needs to actually survive, choosing to pay for everything rather than know how to do it ourselves or go without. A tiny fraction of people today are involved in food production, and most of us don't know how the food is produced, processed, or stored. We don't know how to make our own cloths, fix an electrical or plumbing issue in our own home, or maintain our own vehicles yet we depend on all of these.

Insurance programs are much the same, though when those leave you high and dry its generally much more impactful than when you can't get a toilet fixed in short order. To their credit, some Democrats tried to warn us of the risks of tying health care to jobs and many Democrats tried to design a better system even if they didn't or couldn't explain the risks necessitating it. Now the tech industry is feeling the pain of all this centralization and dependence.

The story this article begins with is tragic, though the fact that we collectively are okay with, and even feel entitled to, being so dependent on various insurance programs is similarly tragic in my opinion.

We need to change the core of what our systems are based on today for any meaning, long lasting change to happen. We can keep duct taping the tears along the edges but it will continue to fail, and usually the failures become more painful and more frequent when we just look for more quick fixes.

Uh, you realise what you are actually opposed to is sedentary civilisation itself? Specialisation is old dude.

How so? If your argument or assumption is that civilization is directly linked to dependence and individual helplessness that seems like a terrible long term strategy.

I don't agree that civilization demands dependence, by the way. We can choose to buy food from someone else while knowing how we would do it ourselves, ideally with some experience. We can buy our food from someone local, reducing our dependence by shrinking the loop is a huge improvement on what we have today.

I wasn't arguing that specialization is bad or evil, but I would argue that too much specialization can lead to a fragile system.

I'm still not sure what safeguards they can be adding here. Unless they've suddenly solved alignment, at best isn't it a collection of system prompts saying what not to do and potentially some screening algorithms that try to catch key phrases in inputs/outputs?

> We will do all the model training ourselves

That's actually an interesting note. So you all will be managing the training runs on hardware you own or rent and manage?

That is slightly different though. The government says you must pay taxes, mandatory as you said.

PostHog here is saying they will train on your data but opting out is allowed. For the taxes analogy to work, PostHoh would not offer the opt-out option at all and you'd be doing something like hacking their system to filter your data out on their end.

At least to me, this seems like a pretty logical progression based on how education is handled today.

We teach children from a very early age that there's always a right answer and that someone smarter/older/etc knows it. They're told to ask that person and largely memorize the answer for a test.

With LLMs we're being told they are, or will soon be, as smart or smarter than any human. Its no surprise to me that people with access to LLMs that can already answer a question would just blindly use it and trust the response.

I assume it will be similar to when a person is out sick or on vacation. Another person on the team likely could take over the work for a day, but realistically it just sits until they're is back.

So work stops until Claude is back? What if Claude comes back and costs 10x the amount? The answer is obviously that you'll "bend over" and pay, because the AI vendor who convinced you that Claude is so great owns you, your codebase, and by extension your company now.

Or you point your Claude code at a different LLM provider. It's not complicated and there are lots of vendors (and in the open-weights space multiple vendors serving the same models competing on price). Sure DeepSeek 4 isn't quite Opus at the moment. But it's plenty good to do the work. We've got different competing front-end tools and different competing back-end providers. No one 'owns' your company. Maybe that will change as the market evolves and one of the frontier tools become so much better than one vendor will own the market. But that's not where we are now.

I didn't realize you could swap out the underlying model used by Claude Code. Aren't all of the Claude tools tied directly to Anthropics models, their authentication and billing, etc?

What happens when your engineer realizes they can make 10x more at another company? They leave and work stops. You then hire someone else or raise your pay to get better, more reliable engineers. The analogies keep going because AI is a tool, not a replacement. If it's a tool used by a non-technical person, so be it, but it's still just a tool.

For substandard developers, yes, work stops.

I have seen many many times in microcontroller forums posts from first timers in the liking of "hello sirs i have problem please show how to do this", followed by their own reply a few hours later asking again because they were holding up, where "this" was usually something really trivial, you just needed to read the docs and the rightful answer was "did you really not try anything in 6 hours?"

In such a scenario, are you assuming Anthropic has a monopoly? Or are all LLM providers callusing on prices?

Not really, realistically speaking it's now possible to use an agent to read code and make sensible summaries of a codebase faster than ever before, and it's exactly the thing you'd use to onboard yourself or someone else on the team.

The OP was asking what happens to productivity when your LLM is offline, I'd assume it isn't available yo onboard anyone at that time either.

More importantly I think, if devs become dependent enough on LLMs that they just put it aside when the model isn't available, they wouldn't be able to onboard quickly or at all.

It takes experience and a pretty deep understanding of programming in your language of choice to pick up a new code base and quickly understand how it works, the architecture(s) and pattern(s) being used, etc. Those skills would likely have been lost long before a dev simply can't work without the LLM.

> Whatever AGI is, it "AGI" is not glueing a load of text prediction machines together.

K don't think it would be that simple either, but for now we simply don't know.

I would like to think that what I consider my intelligence will always be distinguishable from a cleverly built harness wrapped around text prediction, but I can't say for sure that's guaranteed.

Yes we do know. We’ve fully explored the possibilities of LLMs and they are nowhere.

The latest efforts like agents are clearly showing the limitations and are nowhere near AGI.

We’ve now reached the buzzwords and bullshit stage of the bubble where they cast around for problems shaped like the solution.

Oh this bubble started in the buzzword and bullshit stage, no argument there. I'm just not sure how you can be so certain LLMs will have no place if a system we'd consider AGI, most inventions do require a long list of failures and tweaks before it finally works.

Sure seems like we as a country are heading further down the isolationist, nationalist road. I expect we won't be the last western country to batten down the hatches as it were, for better or worse it seems like most countries are preparing for something much larger.

Based on the article here, and Firefox's mythos article, they had found bugs with Opus 4.6 as well but mythos is finding more that it missed.

That would align with the curl feedback you linked, they aren't using mythos but are finding bugs with other models. Presumably the expectation would be that with mythos they'd find more that were missed by other models already used.

> Based on the article here, and Firefox's mythos article, they had found bugs with Opus 4.6 as well but mythos is finding more that it missed.

It's not quite apples-to-apples. It was Opus on Firefox 148, Mythos on 150. A better test of Mythos vs Opus would have been to apply Mythos to Firefox 148. Or also re-apply Opus to Firefox 150.

Do we know all the Opus+Firefox 148 bugs are fixed in Firefox 150? Do we know the number of new bugs introduced per Firefox release?

> Do we know all the Opus+Firefox 148 bugs are fixed in Firefox 150? Do we know the number of new bugs introduced per Firefox release?

That may be parsable from their bug tracker, though I don't know of all bugs raised by mythos are public.

I'd be particularly interested in how many of the bugs found existed in 148. Assuming most or all of them weren't newly created bugs added in 149 or 150, the comparison should still hold even though Opus and Mythos looked at different releases.

To be fair my understanding at the time was that Ukraine was well aware the invasion was coming and the government was publicly making claims that helped avoid panic and allowed them to move military assets around more easily.