News for hackers. Stuff that matters.
Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by derangedHorse on Sunday April 12, 2026 @06:47AM (#47738203)
Pretty bad decision on his part. I've been telling other engineers within my company who felt threatened by AI that this would happen. That prices would rise and the marginal cost for changes to big codebases would start to exceed the cost of an engineer's salary. API credits are expensive, especially for huge contexts, and sometimes the model will use $200 in credits trying to solve a problem that could be fixed in an hour by a good engineer with enough context.

It kind of reminds me of the joke where a plumber charges $500 for a 5 minute visit. When the client complains the plumber says it's $50 for labor and $450 for knowing how to fix the problem.

[dead]
Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by airstrike on Sunday April 12, 2026 @08:53AM (#47739066)
Good story but not applicable at all
Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by wat10000 on Sunday April 12, 2026 @09:13PM (#47746369)
It actually matches up well with the current AI scene, except backwards. We use these model which cost ridiculous amounts of money to train, and all of that effort goes into producing the outputs we use, but we're paying something not too far above the marginal cost of inference when we use them.
Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by airstrike on Sunday April 12, 2026 @09:32PM (#47746510)
So not applicable at all
Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by wat10000 on Monday April 13, 2026 @09:26AM (#47751647)
Extremely applicable to illustrate the difference between people (time is precious, training and experience amortize across a relatively small amount of paid work) and software (can replicate infinitely, time is cheap, startup costs can amortize across billions of hours of paid work).
Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by kshacker on Monday April 13, 2026 @12:42AM (#47747682)
A good engineer and / or a tenured engineer could very well be compared to Picasso in this story. A tenured engineer did not just sit their entire career drawing that painting on the napkin, they delivered other results too. But at the end of it, they are able to deliver a Picasso at a moment's notice.
Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by breton on Sunday April 12, 2026 @11:28AM (#47740860)
> the model will use $200 in credits trying to solve a problem that could be fixed in an hour by a good engineer with enough context

So the price for fixing the problem is equal. Sounds like a great argument for AI.

Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by malinens on Sunday April 12, 2026 @11:50AM (#47741145)
99% of software developers earn less than 200 USD a hour
Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by d1sxeyes on Sunday April 12, 2026 @01:00PM (#47741946)
That “with enough context” is doing a lot of work here. If you take a great engineer, drop them in front of an unfamiliar codebase, it’ll take them more than an hour to do most non-trivial tasks.
Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by collingreen on Sunday April 12, 2026 @12:12PM (#47741411)
Equal sounds like a terrible argument given all the other problems with replacing engineering thought with ai. I don't know where the line is but I expect it's far beyond equal AND there needs to be a level of "this can debug effectively in production" before that makes any sense for a real business case.
Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by OlivOnTech on Sunday April 12, 2026 @11:50AM (#47741140)
Most good engineers are way cheaper than that. The world is bigger than the united states.
Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by charcircuit on Sunday April 12, 2026 @01:17PM (#47742121)
>That prices would rise

Competition will prevent that from happening. When anyone can host open models and there is giant demand for LLMs companies can not easily raise token prices without sending a lot of traffic to their competitors.

Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by paganel on Sunday April 12, 2026 @07:32PM (#47745606)
> When anyone can host open models

They'd still need to pay the actual power costs.

Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by charcircuit on Sunday April 12, 2026 @07:45PM (#47745748)
I didn't say that inference would be free, but that everything to do inference is a commodity which means that competition is easy to do.
Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by jrpear on Sunday April 12, 2026 @01:41PM (#47742354)
Even if you take it as true that prices have risen recently, and may continue to rise as the VC subsidies dry up, they will fall again long-term. Inference will get more power efficient with model-on-chip solutions like Taalas and God willing we will get cheaper and cheaper renewable energy.

Despite this I don't think engineers should feel threatened. As long as there is a need for a human in the loop, as today, there will still be engineering jobs. And if demand for engineering effort is elastic enough, there could easily be even more jobs tomorrow.

Rather than threatened, I think engineers should feel exposed. To danger, yes, but opportunity as well.

Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by travoc on Sunday April 12, 2026 @03:46PM (#47743650)
Increased demand will not drive down energy costs.
Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by lobf on Sunday April 12, 2026 @05:31PM (#47744736)
Why not?
Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by jrpear on Sunday April 12, 2026 @11:12PM (#47747137)
Of course not necessarily, but I keep seeing articles about how wind and especially solar power just keep getting cheaper.
Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by wat10000 on Sunday April 12, 2026 @09:11PM (#47746359)
It seems very unlikely that prices would rise in the long term. Yes, RAM and GPU prices are suddenly going up due to the demand spike and OpenAI's shenanigans, but I doubt it's going to last very long. Some combination of new capacity and reduced demand will most likely put things back on the usual course where this stuff gradually gets cheaper over time. And models are getting better, so next year you can probably get the same results for less compute. That $200 in credits becomes $150, then $100, then....