Recent content by Admin

Improved tetrachromacy In rare cases females have a forth cone type between green and red but it's unfortunately not very helpful since there is no particularly good placement for it (due to the red and green already being quite close). As we see from the picture above the rods have a far better...

Currently there is already various mutations that seem beneficial (at least at first glace): hDEC2 - less sleep required. CCR5 - HIV resistance. ACTN3 - helps with athleticism. A lot of beneficial mutations/genes come from the neanderthals we bred with: V660L - improved fertility and fewer...

AI will often ignore an updated version you send and continue working with an older version instead. This is at least a big problem with deepseek and chatGPT where even if you specify you have an updated version they will just ignore that unless you specifically tell them to work based on the...

The issues with claude It will make mistakes like other AI models but overall the quality has been the best of the models i have tried so far (to the extent i could access ai models without paying money). But the higher quality output you get with claude does come at the price of it taking a...

My poker variant is now playable online thanks to vibecoding (and chinese AI) I naively thought it wouldn't take that much more effort to make a website to play with others when i already had a site where you could play with a bot. Then when i actually tried asking AI models to help me with it...

I was eventually able to get an exact score distribution for my poker variant (i think) I did verify the frequency at 13 and 12 points myself using libreoffice calc. ChatGPT did get the same distribution down to 10 points but then my free plan ran out and i wasn't able to continue. Later i...

Very often when an AI claims to have fixed some issue the issue isn't actually fixed They make you think the problem is solved but then you run into the same or similar bug again or you discover that the supposed fix introduced more problems than it solved. This is in line with the general...

AI has a tendency to take do a shoddy job to produce output instead of doing things properly You end up having to direct it over and over again to actually do things in a more proper manner. So if you rely on AI and do not check what's it actually doing you end up with loads of shitty code and...

I recently got into vibecoding because i wanted to get an accurate score distribution for my favorite poker variant. It's actually a really nice game once you understand the rules so i wanted more people to be able to try it. I recently expanding the evaluator to cover all variants of it and...

IMD control via sabotage One issue with using a separate tweeter for the ultrasonics is that it will cause interference issues at the crossover area and you end up having to have your head in an exact position for it to work well. Another approach is to have one speaker always play the the...

Googlebot could differentiate 24-96 from 16-44.1 using elac speakers but not headphones or canton speakers Note that this doesn't fully eliminate IMD as a potential explanation since different speakers produce different amount of additional IMD from the ultrasonics (very hard to actually pick up...

I have been trying to find independent replication attempts for the "hypersonic effect" and there really isn't much (i even resorted to asking grok due to being out of free messages for claude and chatGPT). We have this short paper which may have found something but nothing particularly...

24-384 vs 24-48 (flawed study, bad filter was used) Bill Howie did an investigation and got solid statistical significance for 24-384 vs 24-48 for 3D audio https://www.willhowie.com/wp-content/uploads/2018/09/wh.3Daudio.sample_rate.discrimination.v1.7.pdf This was with trained listeners who...

The piano test but controlled for IMD I had a hard time hearing a difference when the ultrasonics where played from a different speaker. I didn't spend too much time listening though. I got the difference that it was significantly harder to hear a difference then when you simply down and...

Ultrasonic audibility test (separate speaker for ultrasonics) I made files earlier where i resampled one mono track down to 24-48 and then placed the difference between it and the 24-96 mono version in a separate file. I picked the session based on the 3 earlier tests for 24-96 vs 24-54 which...