Thread: BBFS Providers Manhattan
+
Add Report
Results 646 to 660 of 27113
-
03-08-26 10:39 #26468Senior Member

Posts: 4751May have been fixed as of Thursday night if I recall. All good there now. Pussy can be flushed after some nice CIP.
I see a few pussies we need to DIP still.
📍NYC (21 babes):
⭐ Kathy PPC - best Vietnamese little girl action is finally back after the she was sugared for 2 months?
⭐ Sana MK - best pussy
⭐ Mai PPC
⭐ Kawaeyua PPC
⭐ Audrey Love MK
⭐ Yumekawa Honoka MK
⭐ Kiho Osaka MK
⭐ Mio MK
⭐ Katie MK
⭐ Mina SE
⭐ Rino SE
⭐ Nazu SE - pervert special
⭐ Mayumi SE
⭐ Nuko R8 - petite pussy
⭐ Momoka R8
⭐ Natsu R8
⭐ Bonnie PPC - Tits!
⭐ Emikosumi PPC
⭐ Keiko NYAS
⭐ Yuki R8
⭐ Mizuki NYAS
Ah I see, I assumed you were on the AMD unified memory. It is strange the speed is that slow on the 9 B. I don't like AMD as much since it's large memory but speed will lag after MacBook and nVidia chips (lower VRAM though) and if I am running Windows, it will have nvidia. The Macbook 128 GB will be a lot more expensive so I am just trolling myself on needing one. It's a great system for price and memory for larger language models > 30B.
Thanks!
I have GLM 4. 7 which is about as good as 2nd best model from Anthropic and I have a sub for another year, it was just around $250 total (avg $20 a month but the usage limits are super generous and I never get a warning where with Anthropic I had to wait a few hours as I prompted too quickly and hits the limit fast. I never run out of tokens or prompt turns with GLM. It also can be used for OpenClaw so it's an all-arounder best price for a SOTA form China. Moonshot and Minimax are too expensive when compared to GLM so I am still on GLM for primary dev while I have Anthropic if I was bored at work.
GLM is also why I never need a local model when local lags too far behind still on performance but Qwen is getting better but since I still have GLM, I just don't see a need for local models.
Audio / image / video generation is slow as fuck even on the best Macbook so nVidia is still good, just low on VRAM (24 max on mobile and 32 GB on 5090 desktop and I'm not spending $9 k on a 96 GB VRAM RTX 6000 and I am not going to get a 4 K DGX spark. I rather have a $4 k mabcook if I am spending that kind of money for more utility.
Originally Posted by AlmostVirgin
[View Original Post]
-
03-08-26 10:09 #26467Senior Member

Posts: 4090High quality H2 O
The water has been fine since Friday.
Originally Posted by AlmostVirgin
[View Original Post]
-
03-08-26 09:59 #26466Senior Member

Posts: 608It'sa me, Mario
Plumbing issues SE fixed yet?
PB, I'm running a AMD Ryzen 395+ which is a similar unified memory setup to the Mac's actually. The pool of 128 GB can be configured in BIOS to be a 96/32 or 64/64 split. I found 64/64 better for holding as much context as possible for the latest models (roughly 250 k). SOTA are advancing quickly (everyone is distilling from Anthropic and OpenAI) but in short after having spent the $$ it's not worth the hardware cost at the moment when you can easily get 1 million tokens commercially for cheaper. Fine for a hobbyist I guess. I think real GPU VRAM would be superior in most cases except if trying to run models larger than you can fit in memory.
Here's my tests with your prompt.
Qwen 3. 5 35 B Q8 (37 GB).
44 tok / sec.
Qwen 3. 5 9 B Q8 (10 GB).
22 tok / sec.
Strange that the 9 B model performed worse on this prompt.
-
03-08-26 08:31 #26465Senior Member

Posts: 4751Tsukiji
Bonnie big tits PPC is back https://www.pleasure-palace.club/bonnie.
Yuki R8 has tits too https://risingsun8.com/view/29895.
Kiho MK https://misskyoto.com/ is on the mission. I have to DIP her today.
Virgin - can you just run 9B Qwen 3.5 and also on the 30B so I can see how fast your inference is? I think Nvidia is still king but I can't tell for sure until I get my hands on the M5 Max -- was leaning toward 40 GPUs but I think it will still nbe much slower than nvidian 5090 laptop (24GB VRAM)
"write me 500 words on AI"
my 5090 mobile (24G VRAM on Qweb 3.5 9B was) 80 toks/s. I didn't try the 30B. don't think it will fit.
I think I will get a new Macbook instead of a desktop 5090. I can tunnel my Mac to my 5090 to run inference (but it will be limited to 24G VRAM). I was just gonna get a faster mac with ram "in case" I have asspirations to do more on my macbook.... but it's also $5000 when I checked ther actual price lmao so I don't know if I need a 5090 laptop + a $5000 mac. I did order a $499 Macbook Neo too. 20 GPUs is kind of slow (what I have now and it does 22 toks/sec which is dog slow for me. I am used to 150+/toks from online models.
I am guessing the M5 Max 40GPUS will ne 44 toks or maybe 65 which is same as 5090 mobile (but I can run larger model on the Mac since I have more RAM but it's still gonna be slow on raw power.
Originally Posted by PokerBrah
[View Original Post]
-
03-07-26 11:21 #26464Regular Member

Posts: 17Happy to hear you had a similar. If not even better, experience LOL. Maybe next time I'll try for the DFK FIV and CIP, but it really may be a YMMV with her. Which is cool, and really goes along with her GFE / casual personality. I'm guessing no further upcharge for the CIP?
Originally Posted by BBPornoGeek
[View Original Post]
-
03-07-26 03:21 #26463Regular Member

Posts: 7Mina AU Link
Here's the link I posted, https://auescort.com.au/en-AU/post/239. Hopefully we get so see her again some day.
Originally Posted by NCGold
[View Original Post]
-
03-07-26 02:14 #26462Senior Member

Posts: 4751Tsukiji PPC
Emiko PPC https://www.pleasure-palace.club/emikosumi.
Yuki R8 https://risingsun8.com/view/29895
Kiho MK is busy as fuck. God damn it.
64 VRAM and 64RAM for system? Weird split unless you got 2x5090 RTX. Quite an upgrade from your old PC running in your parent' log cabin.
what's the system build? I still don't run anything for real local because SOTA models online are still best by Qwen 3.5 may be ok. I need to just see how fast it really is, if it's too slow I won't want to use it. I probably just get macbook with 48GB but I was thinking I will do more if I have highest spec but it's usually just asspirational purchase.
-
03-06-26 20:01 #26461Senior Member

Posts: 608
Originally Posted by PokerBrah
[View Original Post]
Yeah. It's only inferred data from the early morning periods 12 am-6 pm most girls get a nice spread of views which I assume partially are other crawls.
Originally Posted by Ayasaki
[View Original Post]
I have 128 GB memory currently running Qwen3. 5 35 B on a 64 GB system memory and 64 GB VRAM split. Paying $20/ mo for something online would be cheaper Tbh LOL.
-
03-06-26 19:34 #26460Senior Member

Posts: 1401Yeah she was pretty cute. Had a good time chatting with her about Europe. I asked if she does the German FKK and she laughed saying no because the Germans are lecherous bastards then turned around asking if I do. I mentioned the girls there and the clubs I visited which unsurprisingly she knows about. She does realize she's not like the other hot eastern europeans at the FKKs but delivers a different experience. She was full GFE with me, DFK, FIV, etc. No greek but filled her up twice with no complaints. TLDR; she's more a cute GND type of eastern european girl vs the hot pros out to get every dime you have on you.
Originally Posted by OzyMandias2
[View Original Post]
-
03-06-26 19:33 #26459Senior Member

Posts: 823Only the RisingSun8 style sites provide that info. But for the ones that do, it's on the girls profiles. In the info section to the right. But a site like PPC doesn't display that information.
Originally Posted by PokerBrah
[View Original Post]
-
03-06-26 19:26 #26458Senior Member

Posts: 4751Where are you getting the view counts.
Originally Posted by AlmostVirgin
[View Original Post]
I only hit the details for new ones. I was optimized first. I still only take a small storage. Even though I have about 6000 girls and their pics. I history crawled the J girls that can find
I feel renewed energy to add some features for myself. I need to test some new tools.
I was gonna get a 128GB MacBook for the hell of it but I dont think I will really run very large local models.
Always looking for lame excuses to buy new gear
I actually ordered the small MacBook Neo cuz I liked the color. I don’t really need a low powered MacBook either
-
03-06-26 17:50 #26457Senior Member

Posts: 608I should also clarify in my specific setup 15 minutes polling rate is only hitting the homepage to find changes and not crawling down to the girls ads. I'm still generally only incrementing view counters on the listing themselves once or a sometimes a few if revisiting to checking for new pics or something.
The highest traffic Manhattan sites are no surprise SE and R8. Within a 15 minutes span there's usually 50+ views easily. How many are horny mongers scrolling their phones, how many are like us hahaha.
-
03-06-26 16:27 #26456Senior Member

Posts: 823Not exact timing but it shifts for me, however, I'd say most of the times it is 4 times max. Although if format changes, debugging will cause that to go up for that individual one. But they all got a one button "go away" option if they want, so I know when I've stepped on toes and haven't. But it's also sometimes entirely unrelated, I believe, like raids in California.
Originally Posted by PokerBrah
[View Original Post]
But this was a subtle change so that's why I took it as a personal no. Plus, it was Xiao-dimsum which you had apparently realized was similar to privateparadise, so if I moved targets, I didn't want to take the risk of messing that one up for yours, so another reason to leave it alone. I did like them though because they had a decent shot at putting a realistic photo in the profile. But again there's no point if they never come to my spot haha.
I even recognize one, Megu / Tomie etc https://xiao-dimsum.com/sinyeo-%e7%b...8%8c%e9%87%8e/ https://mimi10086.com/view/24560 Never saw her but I remember one of the sites edited her face into a smile with teeth and thinking that was odd.
-
03-06-26 14:42 #26455Senior Member

Posts: 4751I only crawl a few times a day and I have about 240 active girls across I think 25 sites. Including couple of Canadian ones. It was easy to add but I didn't add a lot that people gave me since there is no guarantee the girl will come and if girl moves to a town I don't care about then I don't need to track them.
Originally Posted by AlmostVirgin
[View Original Post]
-
03-06-26 09:28 #26454Senior Member

Posts: 608Pussy proximity is real
Similar sentiment. Didn't feel highly motivated to expand past Manhattan where it's more than enough to sift through and take your pick.
But the cat is out of the bag since the barrier to entry became way lower 2026. Based on measuring some of the site views at the start of this project months ago vs. Now there are definitely a few others silently doing this.
My polling rate is once every 15 minutes which is conservative based on this differential I've noticed over the last 6 months.









Reply With Quote





