I've been disappointed with Sanders lately, his age is showing and how he got convinced to give these LLM companies a massive federal bailout is bonkers.
I’ve had a good experience with a greenfield project.
The single thing that seems to have helped is that we all agreed to use OpenSpec early on, and to commit the specs alongside the code.
I have no affiliation with OpenSpec and I don’t suspect it’s doing anything unique here, but having the intent develop alongside the code in the repository seems to have ensured that agents have a more holistic view of the project.
It’s a night/day difference when I use an agent against this codebase that integrates its changes using OpenSpec and those that ignore it.
Hmm, I'm still undecided when it comes to the best approach: commit or not commit. Now I'm leaning toward the second one, as the implementation can diverge from the spec files, even though I store some impl logs in specs. As a part of feature delivery, I update the main repo docs based on the spec and code written. IMO it makes context for the LLM a bit cleaner.
I wish this had non-model comparisons. If Opus 5 is in the top ten, it’s clear that the entire benchmark is somewhere between “Tom Clancy” and “Dan Brown” and about 1,000 new model releases away from Hemingway.
When you see, “Wow, Fable is number one”, you might think it’s a good writer, but that’s not what the benchmark says.
Nearly 75% of Americans are overweight. (over 30% 'overweight', over 40% 'obese').
Obviously, 75% of Americans do not develop dementia. Do more who are overweight develop dementia than those who are not? Well...it's hard to say when 3/4ths of the population is overweight.
Subscriptions are a very small part of their overall revenue (estimates have been between 5% and 20% based on financial reporting). Enterprise users are charged per-token, and maximal input/output tokens nets them maximal revenue.
They still want you to hit the cache because their margin is higher on cache hits. That's actual compute they don't have to pay for and they don't have to have capacity for because they are supply limited on the compute side.
And the unit economics need to be there because there are competitors in the space. They can't just skin you on tokens or you'll jump ship.
Of course we are all guessing, but both things can be true: they don't want you to hit the cache, because cache writes are more profitable than cache reads, and they are supply limited on the compute. Pre-filling input tokens is very different from decoding, so when they claim they are supply limited on compute, do they mean mostly for decode or also for pre-fill (or only for pre-fill)?
I can imagine there are coding tasks where small edits to an existing huge codebase mostly consists of some small tool calls + processing a lot of input tokens, in e.g. a 20 to 1 ratio of input to output tokens.
This issue suggests that 74% of the charged input tokens could actually have been cache reads if claude code hadn't busted the cache. On the input side of things, this increased the cost (or count towards allowance) by ~3x (given that cost of cache write is 1.25 the unit price, and read is 0.1x the unit price).
OpenAI has a different cache pricing strategy, where cached reads are only 0.5x the cost, but cache writes do not cost extra.
Not sure how the unit economics play out claude vs openai, but it's safe to say that caching costs play a huuuge factor in this. It seems to be one of anthropic's USPs as a frontier-AI lab. They charge a premium for cache-writes (on top of the inflated token usage compared to OpenAI that was recently reported), but significantly discount the cache reads. The lack of care in tackling issues related to cache-busting is therefore really bad and suspicious.
We are solving problems, but we're also creating more problems with those solutions. Greenhouse gases, global warming, oil spills, wars over resources, straining power grids, species extinction, worsening mental health, worsening social cohesion, widening inequality. All are the result of technological progress within our systems of government.
Hedonic adaptation prevents us from asking "is this worth the downsides, or should we go back to how it was?"
It's possible to mistakenly think you've solved a problem when you've actually made the situation worse, of course, and that's just plainly a bad thing. It's important to be able to identify specific things which we thought were solutions but were actually new worse problems, and roll them back if possible. [0]
But that's not really what the hedonic treadmill refers to! The hedonic treadmill is referring to the fact that, when a problem is solved, the level of pleasure or happiness individuals experience reportedly quickly returns to the pre-solution baseline.
[0] Crucially, this must not be conflated with advocating a general slowdown or pause on attempts at progress. That's the civilizational death sentence.
> The hedonic treadmill is referring to the fact that, when a problem is solved, the level of pleasure or happiness individuals experience reportedly quickly returns to the pre-solution baseline.
That is not a fact or objective truth or even historical, that is a hypothesis and a theory.
And the distinction matters. It is not in fact true that all people in all setups and classes and historical periods and periods of their lives were equally happy. In fact, you can be unhappy, solve the problem and be more happy/content for the rest of your life. Overall peoples happiness go up and down.
Hedonistic treadmill is a theory. It is also a slight of hand argumentative trick to shoo away valid complains people have. It was worst in the past, but you are not completely happy accepting bullshit today, therefore you just dont know how good you have it, basically.
Global warming make quite a lot of people (like hundred thousands) miserable as they had to evacuate and then return back to non-existent houses. They literally lost everything except bank account and whatever fit into the car. The areas with high drug and alcohol use, high violence rates are in fact not full of people who have it too good. I could go on ... the inequality and accumulation of power is threatening to make most of us all worst off and it is not because we got spoiled by hedonistic treadmill.
Not by much. I seem to remember some research that compared lottery winners to double amputees. Within a short period both groups had reverted to similar levels of happiness.
>I could go on ... the inequality and accumulation of power is threatening to make most of us all worst off and it is not because we got spoiled by hedonistic treadmill.
There isn't much individuals can do about that outside of polling day. What they can do though is start comparing their lives to their grandparents rather than the Kardashians.
I feel that oftentimes our biggest problems come with the irresponsible scaling up of certain solutions, at which point it is usually too late to roll back because we have come to depend on them. There has to be some time between the moment we invent something great and the moment we make trillions of the thing and realize, oh, maybe that's a problem. But we don't pace ourselves. And if we keep plowing ahead heedlessly, we may very well end up burning everything and collapsing our entire civilization. It's a dangerous game to keep betting on the future without actually ever properly planning anything at all.
I don't think anything like a "general speed limit applied to experimentation" makes sense. In general we should move as fast as possible.
When there's a specific explanation why a specific experiment should move slowly (or even be prohibited), I respect that, but the explanation can't just be "we don't know all the bad things that might happen as a result of this experiment." That's the precautionary principle, which I staunchly oppose.
I'm describing a speed limit applied to production, not experimentation.
We can progress science, knowledge and technology all we want without necessarily committing every resource we can find to producing ten times more stuff than we need. Moving too quickly in terms of production can be detrimental to progress: first movers can and often do take over entire markets, leaving no room for the often better-designed second or third movers to shine. Moving more slowly allows for better maneuvering and making sure the best systems and technology can prevail. And it allows us to see problems coming before they hit us.
I think we could have made a utopia purely from fossil fuels if we had been more careful and restrained using them. In my view, the fact that we couldn't do that is a major indictment of our process.
Nah. Rapid, unconstrained growth in the early production and use of fossil fuels was one of the greatest factors ever in saving lives and improving quality of life worldwide. Stopping that would have condemned millions to poverty, darkness, and death.
And it's a moot point anyway because there was no way to slow down even if some people wanted to.
There is such a thing as borrowing against the future. If we fail to keep climate change in check and millions of people die as a consequence, that may just wipe out all the gains you mention.
And let's not mix everything up. Every innovation is different. Penicillin and the Green Revolution aren't in the same category as bottled water and fast fashion. We can certainly fast track technologies that have large, immediate positive effects while throttling frivolous crap.
Or maybe we can't. Maybe humanity can't plan ahead -- being made from thinking agents doesn't necessarily grant a system a capacity to think. But I find it's a strange act of faith to believe barreling ahead thoughtlessly is somehow always going to be fine.
https://jacobin.com/2026/07/ai-nationalization-sanders-liber...
reply