adi
@adonis_singh
Followers
14K
Following
76K
Media
2K
Statuses
10K
17 • contributor @_mcbench • ex @lmstudio
✝️
Joined November 2022
I thought he was joking but no lmao this is just a little embarrassing
15
2
181
I thought that 3 models at 61% seemed a bit suspicious, so I asked gpt-5.2 in codex. A pretty good explanation from it: Not really. With n=100 you only get 1% resolution, and if several models are clustered around ~0.6 true accuracy, landing on exactly 61/100 is common (binomial
1
0
8
North American PEA Shows After-Tax $1.25B NPV at US$3,650 Gold
0
9
86
gpt-5.2-high still gets ~61%, no improvement over medium reasoning on IBench
3
0
24
Mini-to-Micro Contract Converter = instant product fungibility. Swap E-minis and Micro E-minis on the fly with a clean 10:1 ratio—no margin math, no friction. One unified risk limit, seamless contract sizing, and your max position stays tied to the primary contract. Trade up.
49
28
222
gpt-5.2 is a substantial improvement in essentially every aspect i can think of - long context - visual reasoning - math/science/stem still need to chat with it to test sychophancy/creativity stuff, but seems really good on paper
4
1
49
gpt-5.2 ties gemini 3.0 pro on IBench! this is huge, openai was always lagging behind on visual tasks but seems to have caught up
2
14
168
Looking to invest in the Enablers and Adopters of AI? Consider an actively managed fund investing in companies actively involved in developing and implementing AI technologies.
0
36
307
this in reference to them actually naming it 5.2, benchmarks are alright too i guess
1
0
17
i hope they dont actually call it 5.2 it'll be too much for me to handle gpt-5.2-codex-max-extra-high-fast
8
0
91
if this were anthropic, they would've just released a groundbreaking model without the vaguepoasting but sure
22
2
644
Keep your outfit clean anywhere with this portable lint roller. It removes dust, pet hair, and lint in seconds, helping your clothes stay fresh all day. Compact, reusable, and easy to carry—your quick way to stay neat on the go. 👉 Ready to buy? Just click the video or the link.
0
5
32
wow haha so subtle
1
0
55
this looks very cool, cant wait to run the small variant locally!
Our next-generation coding model family Devstral 2 is available in two sizes: Devstral 2 (123B) under a modified MIT license, and Devstral Small (24B) under Apache 2.0. Both SOTA, open-source, free to use, and available now via our API.
0
0
5
this is such a baseline understanding of "AI", it baffles me how it has 70k likes
as a reminder: AI cannot generate knowledge. It cannot create knowledge. It cannot find new information. It can only mix information that has already been found and written and input into computers by humans.
111
9
661