I don't know if I'm doing something wrong, but if I set Capacity Retained to 100% then no matter what I set the other values to a bunch of countries deplete their reserves.
This feels wrong; but I'm inclined to think I'm missing something.
I was surprised at how little improvement they could eek out of tuning it; but it is a non-trivial improvement which is much less likely to wastefully spin if your validation is more expensive than you realise.
Makes sense that its in the major harnesses and not the self-built ones.
Commands like /goal and similar are the more complex version of this; you write a prompt like it was a singular iteration, it runs one iteration, then runs an "evaluator" to determine if the goal has been reached, then runs the prompt again with a little extra to make it go again - and so on. The evaluator is just an LLM with most of the result or context looking at the original goal and the state and answering the question "has the goal been met".
The interesting part of this: while some leading implementations use the same LLM and context for the evaluator, some call out to a different context, some to a tuned LLM and different context; so which is better? many blog-scale benchmarks are calling it a toss-up that is highly dependent on the primary model.
A graph is much more than one conclusion; in fact, almost the entire point of graphing is to allow the comparison of "shapes" and to easily hypothesise about associations across datasets.
This graph misrepresents the rate the price declines and the length of time it has been stable for, which throws off nearly all non-trivial conclusions.
What on earth is that H100 price trend graph. The equally spaced x-axis points are 2x6 months, 6x3 months, 9x1 month. The whole visualisation of the trend is ruined on the back of that.
There are liars, damned liars, and people who play silly buggers with scales.
IIRC Apple have been allowed to remove Fornite from the Apple Store, they just fall foul of the EU Digital Markets Act / DMA (?) when also blocking the Epic Games Store as a route to add/sideload it.
Removing ChatGPT due to ToS violations seems like it would be ok.
Built to work with lmstudio, one of the leading easy to use local model servers. LMStudio is the closest to plug-and-play without sacrificing play that I've seen; a harness that works well with it is nothing to sniff at. Its not earth shattering either.
This is genuinely hilarious. Are you able to elaborate? Which banks? Which B2B? There is probably a shared product stack here that is making some hilariously poor decisions.
I think that "Africans should not have access to high-speed satellite Internet" is something you've just made up; the article you link talks about African people turning to Starlink because of local infrastructure issues, and the original article notes that the current satellite count is currently around 14k satellites. 100k is more than enough satellites to provide high-speed satellite internet globally.
The article makes mention of specific endeavours, like the night-time mirror satellites, which are particularly disruptive to astronomy, and the general risks of high numbers of satellites.
The ability to do Earth based astronomy is something that is of value to all the peoples of Earth, and is mainly funded by the western nations because of their current position as the people with more money.
When you didn't learn to type properly, relearning to type can be a very difficult task; re-learning on a split keyboard is particularly unforgiving. Around three weeks into re-learning I was convinced I would never learn properly and that I'd wasted a lot of time and money (I was freelancing at the time) on something that wouldn't help me eat, never mind sleep.
Two weeks later I was back up to normal typing speeds, a month after that I was faster than ever. Two months or so after that, my back pain was gone.
Of course, my back pain was caused by sitting lopsided - something an overdominant hand on a standard keyboard pushes you towards. No amount of exercise and posture correction was solving it - but when the true cause was resolved it cleared up (with exercise) very quickly.
I'm not sure if it is that the desktop is being taken more seriously, or that its easier to write code that works on many distributions and configurations, greatly reducing the cost and increasing the value of the existing 'market'.
I think your critique is actually the hollowest thing here.
The article is presenting an idea, not a solution. You've failed to see this and have constructed several strawman arguments in order to critique it.
Most importantly the article does not present itself in a definite sense - it is written with care to say "I think", "Next thought:", "Probably", "So yeah.". This article is a person sharing what they are thinking, and unlike many of my thoughts - it is a fairly complete thought which is clearly sparking many other reasonable people to think along similar lines.
The author doesn't present the solution - but there is no reason they should have to. What an odd and unreasonable bar you set.
I also don't find your attacks on the authors site particularly endearing. The taste gap is well known, and punishing someone for their conceptual contribution outstripping their practical skill is quite... distasteful.
A more competently written critique would have been more charitable and in the spirit of this community.
Ha. Yes I haven't calibrated myself well for that form of users. It would seem that telling them to tell their agent to read cooldowns might be the best advice there; they sure as he'll won't read it.
This feels wrong; but I'm inclined to think I'm missing something.