If you send the full 200k tokens on every request you will get very few requests before you hit the token limit. Caching reduces the number sent but I don't know how much they can cache?
ReasonML is just OCaml with a different syntax. It sort of very loosely resembles JS syntactically in that it's got curly braces, but that's about as far as the similarities go - about as much as Rust resembles C. The type system and language are very different from TypeScript or JS, and much more rigid.
"In the polar Southern Ocean, cold, fresh surface waters overlay warmer, saltier deep waters (Fig. 2A). During winter, surface cooling and sea ice formation reduce stratification, allowing vertical mixing to transport heat upward, either melting sea ice from below or limiting its growth (8). However, decades of surface freshening strengthened stratification, trapping subsurface heat at depth, sustaining expanded sea ice coverage (7, 9) and limiting deep convection along with open-ocean polynyas (10). Here, we show that since 2015, these conditions have reversed: Surface salinity in the polar Southern Ocean has increased, upper-ocean stratification has weakened, sea ice has reached multiple record lows, and open-ocean polynyas have reemerged."
That trapped warmth doesn't mean the deep water is "hot" in an absolute sense, just that it's saltier and denser and relatively warmer than the surface
In general at high enough depth the ocean temperature is constant[0].
At high latitude like the southern ocean it's constant at whatever depth.
I think the surface ambient temperature is below(cooler) the temperature where the water density is the highest around 4 degrees for pure water(water has negative thermal expansion which causes it to expand and float!), for southern ocean salinity is between 33-34 and maximum density is below 0[1] but still ambient might be lower which means the colder water is lighter.
Being relaxed is important in boxing too, if you don't stay loose and relaxed (other than snapping explosively when throwing a punch), it is easy to get fatigued quickly and also harder to slip or roll.
Yeah - I imagine that’s true for a ton of sports, but especially for any combat sport. I mean, even in shooting pistol (I shot sport pistol in college), breathing is the main thing you can control, and the main thing that will screw you.
The point is that staying relaxed in boxing during sparring or working the heavy bag feels counter-intuitive in the beginning. It seems like tensing up the core and arms would help punch harder, but you quickly learn that conserving stamina is more important than hitting as hard as you can (especially when starting) because otherwise you run out of gas pretty quickly.
I switched to Kagi in June last year. I just realized I tried it initially because I wanted to try out blocking sites in search results, and I have only ever needed to block three domains.
Kagi is kind of like Google in 2009, seriously good coverage, good ranking
... but also:
- more modern
- more features (summarizer, bangs like in DDG, FastGPT and probably a few I forgot)
- blocklists for websites (and also options to pin, raise and lower)
- with actual support: report a bug and you get an answer from a real engineer, a follow up when it is fixed and a shout out in the relevant release notes
I use FastGPT quite often although I’m not a subscriber to Kagi itself. For me it’s everything an AI search engine should be. Here is your answer, and here are a bunch of links to research further. Something that works without making the web obsolete. Not like the walled off garden of OpenAi which often hallucinates links, or Google’s “I through everything at the wall to find what sticks” effort.
I like Kagi a lot (just look at my comment history), but I'm letting my subscription lapse when it comes time to renew. I've found myself going to Google a lot more often, and I'm finding more and more transparently spammy sites in the Kagin index. Some, for example, are clearly Gen AI created.
If I were a rich man, I would probably keep my subscription just to support a Google competitor. Alas, I'm not, and so I'll be going back to Google.
I wouldn't necessarily call them fishy, but I am very tired of them. They have a very evangelizing tone. But I think they're ultimately just people excited about the tool they're using and wanting to share it with others.
I took a screenshot years ago where 10/14 of the viewable top headlines on my screen where positive Google discussion. From an advertising perspective it was all earned marketing (satisfied customers speaking highly).
While these situations could be a pg-style astroturf submarine, or they could be satisfied customers (the best kind of advertising), I wouldn't necessarily say fishy (you can look at the satisfied users' previous contributions to make that judgment yourself! :)).
Personally, I've not used Kagi, but I hear positive things from people I trust that use it. So I'll likely try it in the future.
Did we not all evangelize Google in it's early days?
Also, none of these accounts saying nice things appear to be bots or kagi-focused in any way, so I think it's safe to assume they do actually just like it.
I never say it but here it is: for the price of 2 packs of cookies, I went from being a 1x programmer to a 1.5x programmer without doing anything. If the results are good, it’s good for me and my job which brings me way more money and satisfaction than $10.
The alternative is Searx and I may try it sometimes, but so far Kagi is cheap and very efficient for me (C++ coding and other languages).
Google Search being a bit rubbish has been in the zeitgeist for a while, it's not surprising that people then talk about an alternative they've found that is much better in their experience
