Yes but my point is
- Resource limits are a "recommendation" and are not strictly enforced
- Significantly boosting resources up to 3 did not statistically shift performance results
Sure for old tasks you could argue that now its not required to boost because infra errors are alleviated with better default limits. My point more so is that its a strange thing to index on because if you wanted to cheat on the benchmark, it does not particularly seem like something that shifts results? Once the API is out maybe I'll eat my words, but I don't really believe that if you manually tried to reproduce the results with lower limits you'd see significantly different results
It seems to rank at around the same as nano banana (slightly higher) in blind A/B test benchmark but of course gpt image is a step above both right now
I'd be very cautious how matching works. For some markets like sports it's trivial, but many politics or economics markets have minute rule differences that dramatically change what the actual market is betting on. Many markets have identical titles but are actually totally different markets.
None of those are standard and open except for Decent and UVI which does have third party developers albiet not many. Anecdotally the only dev I know on UVI is acoustic samples? Since Virharmonic moved off of UVI onto their own powered by gorilla
The rest you listed, Soundpaint, Engine, etc. are company specific proprietary samplers. Their only market share is the market share of their own company
A lot of these aren't even samplers?
SINEfactory is just OT's free collection of instruments. Labs is not even a free collection either anymore, Spitfire has done some strange subscription or monetization wrapper around it (and fyi labs used to be on Kontakt).
Kontakt has no competitors really. There are no standard and open competitors, the closest is maybe something like HISE but it's not even close from a marketshare perspective.
Is this post not just an ad for a vibe coded site / product? It adds no new info on the mercor breach and advertises something which I presume has even worse safety practices
Polymarket isn't doing anything about it. It's the US government because obviously while I suppose this info made a more accurate "prediction" it also yk, leaked confidential state military secrets which is something the government can prosecute. They're not being prosecuted for insider trading on Polymarket
Huh, that's news to me. I had to do some googling and the only one I could find was by Phalicity which seems to be from a Meta employee who left and made it into a company (a la Statsig, or to some extent Graphite)
Seems that company shutdown though a while ago and it got forked into Phorge, but either way I assume there's some divergence from what's internal at this point.
Still I'll look into it, it does look neat and might suit my preferences still. Thanks for the headsup :)
What are with these comments? Did no one even read the article. Not once is composer 2, or kimi, or anything mentioned... because the article is about Composer 1.5.
I mean sure the techniques are probably the same in 2 but its not like they're exactly advertising composer 2 here lol
Sure for old tasks you could argue that now its not required to boost because infra errors are alleviated with better default limits. My point more so is that its a strange thing to index on because if you wanted to cheat on the benchmark, it does not particularly seem like something that shifts results? Once the API is out maybe I'll eat my words, but I don't really believe that if you manually tried to reproduce the results with lower limits you'd see significantly different results