Espressosaurus

My experience is that 6.1 Sol is a little behind Opus 5.5 but it’s in spitting range. Their cost per task is generally higher, but if you’re on a quota plan the difference in cost doesn’t really matter until you’re paying API rates.

Both get the job done these days. I still hate Opus 5.5’s output, 6.1 Sol does an acceptable job of doing a second cleanup pass to it.

They were on their downswing but LLMs put a gun to their head and executed them. I don't use Stack Overflow anymore, just an LLM and cppreference.

Damn sad how quickly that habit changed for me.

Never asked a question, but got some good info from people who had asked questions and the people who replied.

If we assume none of it is front-loaded (which seems, uhhhh unlikely) we're looking at ~$100 billion a year for the first 4 years of commitment. That's significant.

Yeah, but they've been running layoffs (sometimes more than even yearly) since what, 2022? 2023 at the latest.

I'm taking it to mean everyone below (or perhaps even including) VP, because I thought they topped out at 68 for normies.

It’s not clear yet. Index funds are now an enormous part of the market and players are starting to play them (see SpaceX, and given the door they opened OpenAI and Anthropic to follow).

Peak to peak, inflation adjusted, can be 10 to 15 years for the really big busts though. It's not entirely irrational if you moved in at the peak.

If you were in on the upside then you've got much less ground to retrace.

And it's painful to watch your holdings drop like you said if you're planning on retiring in the next 1, 2, 5 years.

So I get it.

If you're 25 you've got tons of room to maneuver. If you're 60 things are a lot less rosy (even if by now you should theoretically be cycling into less volatile assets. But bonds didn't do that great either over that period) .

We could have had a slowly declining empire and a smooth handoff.

Instead we decided to elect a bunch of people to burn it all down and speedrun the fall of Rome.

We'll still be a world power at the end of it--in the same way France and England are world powers.

My impression based on a cursory look at the research is that the caching algorithm is absolutely vital, non-obvious, model-dependent, and potentially prompt-dependent to make sure you're holding onto the right experts.

It doesn't look like a solved problem yet.

And Muse is what normal people use instead

OpenAI dropped a slop bomb on mathematics the same way people will drop slop bombs in codebases and the people here are saying “well drop everything and review it of course!”

I think it’s very reasonable to expect the people that are dropping this on other people for review to have reviewed it first, closely, made adjustments, etc, the same way I do before dumping my Claude/ChatGPT assisted PRs onto other people.

The reviewers shouldn’t be exerting more effort than the person that set the prompt.

This is how I can tell you aren’t an immigrant. Neither am I.

The money is what matters since it gives you resources you don’t have access to elsewhere. And it’s worth burning yourself out on because the situation is still better than the situation you left behind.

Happiness? Fulfillment? Ain’t nobody got time for that.

That’s for their kids or their grandkids.

Microsoft still was willing to sponsor (which many companies are not) and still pay an outsized wage (even if low by FAANG/Bay Area standards)

The US has been the beneficiary of the world's brain drain for the last...what, 50 years? 80 years?

They start the most businesses, they staff specialty roles like doctors, nurses, researchers, engineers. They put down roots and innovate. After 2-3 generations of course their kids are the same as every other American.

But first and second generation immigrants are far more driven and that shows up. They'll do the dirty jobs if they don't have an education, and if they do have an education they're going to be going for high impact roles.

Since we're going nativist long term I expect we're going to see innovation and productivity drop. It's not a one year change, but over generations we can expect to see the impact. Additionally immigration has been helping stave off our demographic rollover since our birth rate is relatively low. Without immigration which both increase working age population and tend to have more kids per capita than the native born Americans, we can expect the same demographic problems other countries are already dealing with.

How will we manage our giant deficit as growth slows? What if it stops? We can't rely on international integration that has given our giant companies worldwide profits given the way we've shattered the global order.

Also ZERO chance I’m connecting my bank account to a gas station. Gas stations are where I’ve had the most fraud attempts originate on my credit card.

Connecting my checking account to that? Fuuuuck that.

The suggestions in the box where I put my text push me out of flow. Bad enough it says “want to do X?”, but when it’s in my own text box it screws me up. It’s not more efficient. It actively makes things worse for me.

Waymo at least seems to take a reasonable duty of care. Tesla clearly doesn't.

Well I would feel slightly irate if I had to find out about the bear trap before I decide I don’t want the car.

30 years of money banked is quite a bit no matter which way you slice it.

This is for American schools, unfortunately. And it's not a good development.

Similarly, algebra needs to be automatic to do calculus.

Everything builds on everything else.

No, they think they learn, then when they need to use that knowledge it turns out they learned nothing.

As far as I can tell (based on university and high school level instruction forums) institutions increasingly are not backing the teachers that want to let kids that have earned that F fail the students.

I've heard of places where the minimum points for an assignment is 50% even if you never turn it in which is insanity.

I'll second that. I tried the trial for a few months then started paying money for it. At $100/year I've found it useful for modeling and gaming out different scenarios for retirement thresholds.

Nothing's certain, but if you like to play with data and you don't want to build the whole damn thing yourself, it's useful.

Still waiting for them to get support for options.

It’s not like ChatGPT isn’t doing similar. I’ve been hit by cybersecurity strikes before while working on an internal codebase that I had to appeal. Anthropic hasn’t done that to me yet. ChatGPT also regularly does that “thinking for a long time while we check if your chat is rule breaking” thing a lot for me when doing model identification without even interacting with external codebases or services.

The real answer is local instantiations where you don’t have to worry about poorly tuned guardrails screwing you over while you try to work.

Until eventually the Chinese models get good enough/the strategic balance shifts and they start locking everything behind closed weights the same way the US companies are doing.

Anything with long context quickly gets dominated by cache reads. Especially for interactive sessions I’ve got cache read % between 95% and 98%.

I work in embedded systems. I have seen the same thing happening day by day from Opus. Some days it’s okay to use and performs well. Other days I have to correct it repeatedly and remind it of information already in the prompt earlier (before compaction!) and still other times it’s infuriatingly stupid.

It’s a slot machine for what they’re actually giving us behind the opaque paywalls.

Yes, I’m on a business subscription plan.

A pity Supernova was never finished. Stars! Without all the Econ micromanagement would have gotten rid of one of the bigger mechanical advantages someone with a lot of patience had.

Yeah, but this time it's ON THE INTERNET.

I mean it's WITH AI!

Depends on the jurisdiction. No matter how you slice it's it's an interception, it may be a transmission (to the iPhone if nowhere else), and there is an argument to be made that the text extracted is a recording.

It is in no way settled law especially in places with 2 party consent laws like Florida or California. Nevermind Oregon. Or the Europe. A cursory search suggests France's laws have some teeth to them.

It also doesn't address the inadvertent listening problem.