๐™‰๐™‹๐™Š๐™Ž (@npostring) header
๐™‰๐™‹๐™Š๐™Ž (@npostring) profile picture

ใญใ‡็Ÿฅใฃใฆใ‚‹๏ผŸ ITใ‚จใƒณใ‚ธใƒ‹ใ‚ขใฏใ‚ตใƒœใ‚‹ใŸใ‚ใซๅƒใใ‚“ใ ใ‚ˆใ€‚

ๅๅคๅฑ‹ๆœ€้ซ˜
Joined April 2019
Photo by ๐™‰๐™‹๐™Š๐™Ž (@npostring)Photo by ๐™‰๐™‹๐™Š๐™Ž (@npostring)Photo by ๐™‰๐™‹๐™Š๐™Ž (@npostring)Photo by ๐™‰๐™‹๐™Š๐™Ž (@npostring)Photo by ๐™‰๐™‹๐™Š๐™Ž (@npostring)Photo by ๐™‰๐™‹๐™Š๐™Ž (@npostring)Photo by ๐™‰๐™‹๐™Š๐™Ž (@npostring)Photo by ๐™‰๐™‹๐™Š๐™Ž (@npostring)Photo by ๐™‰๐™‹๐™Š๐™Ž (@npostring)Photo by ๐™‰๐™‹๐™Š๐™Ž (@npostring)
๐™‰๐™‹๐™Š๐™Ž retweeted
Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values. Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts Chinaโ€™s party line. After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs. The values in an open model are one fine-tune away. Our eval, weights, and training data are available now along with the full blog post: nousresearch.com/arete
1,853
1,495
1,302
15,088
2,673,061
๐™‰๐™‹๐™Š๐™Ž retweeted
Aperture Jason Anderson ๐ŸŽจ
16
456
25
5,159
110,282
โ—ค๏ฟฃ๏ฟฃ๏ฟฃ๏ฟฃ๏ฟฃ๏ฟฃ๏ฟฃ๏ฟฃ๏ฟฃ๏ฟฃ๏ฟฃ๏ฟฃ๏ฟฃ๏ฟฃ๏ฟฃ๏ฟฃ๏ฟฃ ใ€Žใ‚ฌใƒผใƒซใ‚บใƒใƒณใƒ‰ใ‚ฏใƒฉใ‚ค THE MOVIEใ€ ใ€€ใ€€ใ€€ใ€€2027ๅนด5ๆœˆ21ๆ—ฅ(้‡‘) ใ€€ใ€€ใ€€ใ€€ใ€€ROADSHOW ๏ผฟ๏ผฟ๏ผฟ๏ผฟ๏ผฟ๏ผฟ๏ผฟ๏ผฟ๏ผฟ๏ผฟ๏ผฟ๏ผฟ๏ผฟ๏ผฟ๏ผฟ๏ผฟ๏ผฟโ—ข ใƒ†ใ‚ฃใ‚ถใƒผใƒ“ใ‚ธใƒฅใ‚ขใƒซๅ…ฌ้–‹๐ŸŽธ #ใ‚ฌใƒซใ‚ฏใƒฉ #GirlsBandCry
86
13,259
1,527
31,894
1,173,745
๐™‰๐™‹๐™Š๐™Ž retweeted
ใƒ“ใ‚ธใƒฅใ‚ขใƒซใ‚ขใƒผใƒ„๏ผKeyๆ‰€ๅฑžใฎใ‚ฏใƒชใ‚จใ‚คใ‚ฟใƒผ้บปๆž ๅ‡†ใซใคใใพใ—ใฆๅคงๅˆ‡ใชใŠ็Ÿฅใ‚‰ใ›ใ‚’ใ”ๅ ฑๅ‘Šใ„ใŸใ—ใพใ™ใ€‚ visual-arts.jp/
1,414
73,193
12,450
140,014
26,695,258
๐Ÿบ
1,470
76,639
9,515
603,565
21,846,936
๐Ÿบ
2,956
134,450
29,955
829,829
38,521,304
๐Ÿบ
3,908
112,785
37,212
757,161
34,242,477
๐Ÿบ
2,419
83,754
22,268
622,782
25,313,007
๐Ÿบ
2,080
64,513
15,201
508,559
24,675,720
๐™‰๐™‹๐™Š๐™Ž retweeted
Today weโ€™re introducing Gemini 4 Argon. It delivers frontier performance in complex workflows across real-world software engineering, knowledge work, and cybersecurity defense with an industry-leading 1M token output limit.
1,721
4,774
2,790
43,647
10,029,736
๐™‰๐™‹๐™Š๐™Ž retweeted
supercellใ€ๅ†ๅง‹ๅ‹•ใ—ใพใ™ใ€‚ ๆ–ฐๆ›ฒใ€Œๅคœใจ้’ใ€ใ€ใƒœใƒผใ‚ซใƒซใฏใ‚„ใชใŽใชใŽใ€‚ ๅЇๅ ดใ‚ขใƒ‹ใƒกใƒผใ‚ทใƒงใƒณใ€Ž้ญ”ๆณ•ไฝฟใ„ใฎๅคœใ€ใฎไธป้กŒๆญŒใงใ™ใ€‚ ไปŠๅ›žใฎใ“ใจใ‚’ใ€่‡ชๅˆ†ใฎ่จ€่‘‰ใงnoteใซๆ›ธใใพใ—ใŸใ€‚ note.com/ryodayospcl/n/ncc74โ€ฆ
281
47,377
8,351
121,063
11,371,918
๐™‰๐™‹๐™Š๐™Ž retweeted
GPT-6.1 Sol: near-Astra intelligence for a fifth of the price. Itโ€™s the most cost-efficient model for its performance available today.
719
1,705
1,082
21,049
3,263,950
๐™‰๐™‹๐™Š๐™Ž retweeted
ๅ…ˆ้€ฑๆœซใ€ใ ใ„ใถๅ้Ÿฟใฎใ‚ใฃใŸใ€Œใƒใ‚คใƒ–ใ‚ณใƒผใƒ‡ใ‚ฃใƒณใ‚ฐใ—ใŸใ‚‰Vercel็คพใ‹ใ‚‰120ไธ‡ๅ††่ซ‹ๆฑ‚ใ•ใ‚ŒใŸไปถใ€ใฎไบ‹ไปถใฎ็œŸ็›ธใจใ€ใ“ใจใฎ้ก›ๆœซใ€ๆœ€็ต‚็š„ใซใฉใ†ใชใฃใŸใ‹ใ‚’่ฉณ็ดฐใซใพใจใ‚ใพใ—ใŸใ€‚ ๅ‚่€ƒใซใชใ‚Œใฐใจ
18
246
106
1,083
304,933
๐Ÿœ
2,438
68,156
20,178
517,746
35,265,740
๐Ÿœ
1,933
65,184
15,538
534,282
29,400,206
ITใ‚จใƒณใ‚ธใƒ‹ใ‚ขใŒๅฉšๆดปใ‚’ๆœ€้ฉๅŒ–ใ—ใ‚ˆใ†ใจใ—ใŸ็ตๆžœใ€็ตๅฉšใ™ใ‚‹ใพใง7ๅนดใ‹ใ‹ใ‚Šใพใ—ใŸใ€‚ 160ไบบไปฅไธŠใจใƒžใƒƒใƒใƒณใ‚ฐใ—ใฆใ‚‚่ฆ‹ใคใ‹ใ‚‰ใชใ‹ใฃใŸใ‚‚ใฎใจใ€็ตๅฉšใ—ใฆๅˆใ‚ใฆๆฐ—ใฅใ„ใŸใ“ใจใซใคใ„ใฆๆ›ธใใพใ—ใŸใ€‚ โ–ถ๏ธŽ ITใ‚จใƒณใ‚ธใƒ‹ใ‚ขใŒๆœฌๆฐ—ใงๅฉšๆดปใ—ใŸใ‚‰7ๅนดใ‹ใ‹ใฃใŸ note.com/igz0/n/n758e57fddaaโ€ฆ
ITใ‚จใƒณใ‚ธใƒ‹ใ‚ขใŒๆœฌๆฐ—ใงๅฉšๆดปใ—ใŸใ‚‰7ๅนดใ‹ใ‹ใฃใŸ๏ฝœใ„ใใžใƒผ โœˆ๏ธ ๆ—…ใ™ใ‚‹ใƒ—ใƒญใ‚ฐใƒฉใƒžใƒผ

ITใ‚จใƒณใ‚ธใƒ‹ใ‚ขใŒๆœฌๆฐ—ใงๅฉšๆดปใ—ใŸใ‚‰7ๅนดใ‹ใ‹ใฃใŸ๏ฝœใ„ใใžใƒผ โœˆ๏ธ ๆ—…ใ™ใ‚‹ใƒ—ใƒญใ‚ฐใƒฉใƒžใƒผ

ใ€Œใชใ‚“ใ‹้•ใ†ใ‹ใ‚‚ใ€ ๅฆปใจไป˜ใๅˆใ„ๅง‹ใ‚ใฆ1ใ‹ๆœˆใปใฉ็ตŒใฃใŸใ“ใ‚ใ€ๅƒ•ใฏใใ‚“ใชใ“ใจใ‚’่€ƒใˆใฆใ„ใŸใ€‚ ๅฆปใฏๅ„ชใ—ใใ€ๆธฉๅ’Œใง่ฃ่กจใŒใชใใ€ไธ€็ท’ใซใ„ใ‚‹ใจ่‡ช็„ถใซ็ฌ‘ใˆใ‚‹ไบบใ ใฃใŸใ€‚ ่ถฃๅ‘ณใซใ‚‚ๅ…ฑ้€š็‚นใŒใ‚ใ‚Šใ€ใ“ใกใ‚‰ใฎไธๅพ—ๆ„ใชใจใ“ใ‚ใซใ‚‚็†่งฃใŒใ‚ใฃใŸใฎใงใ€ๅฐ‘ใชใใจใ‚‚ใ€ŒๅซŒใชใจใ“ใ‚ใŒใ‚ใ‚‹ใ‹ใ‚‰่ฟทใฃใฆใ„ใ‚‹ใ€ใจใ„ใ†็Šถๆ…‹ใงใฏใชใ‹ใฃใŸใ€‚ ใใ‚Œใชใฎใซใ€ใ“ใฎไบบใจ็ตๅฉšใ‚’ๅ‰ๆใซไบค้š›ใ‚’้€ฒใ‚ใ‚‹ใ‹ใ€ๅƒ•ใฏใชใ‹ใชใ‹ๆฑบใ‚ใ‚‰ใ‚Œใชใ‹ใฃใŸใ€‚ ใชใซใŒๅผ•ใฃใ‹ใ‹ใฃใฆใ„ใ‚‹...

note.com
1
70
31
458
102,386
VIDI STUDIO v1.5.3 ใƒชใƒชใƒผใ‚น๏ผ ๏ผ‘๏ผŽFisheye ใฎ้ซ˜้€ŸๅŒ–ใ‚ชใƒ—ใ‚ทใƒงใƒณใ‚’่ฟฝๅŠ  ๏ผ’๏ผŽใ€Œๅ‡บๅŠ›ๅพŒใฎๅ‹•็”ป็ทจ้›†OKใ€ใฎๅˆฉ็”จ่ฆ็ด„่ฟฝๅŠ  Fisheye ใซ้–ขใ—ใฆใฏใ€ไปŠใพใง้€šๅธธใฎๆ•ฐๅๅ€ใ‹ใ‹ใฃใฆใ„ใŸใ‚‚ใฎใŒๆฟ€ๆธ›ใ—ใพใ™๏ผ ๆทปไป˜็”ปๅƒใฎใ€Œๅณใ‚ซใƒฉใƒ  -> Export -> Fisheye Accelerationใ€ใ‚’ONใซใ—ใฆ่ฉฆใ—ใฆใฟใฆใใ ใ•ใ„(๏ฝ€๏ฝฅฯ‰๏ฝฅยด)ใ‚ž
4
9
35
2,554
๐™‰๐™‹๐™Š๐™Ž retweeted
We Must Pace the Frontier: Iโ€™ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so. Anthropic is unilaterally committing to the first of these steps. Weโ€™ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess modelsโ€™ alignment during training. You can read the full post here: darioamodei.com/post/we-mustโ€ฆ
10,542
16,298
11,365
87,516
77,146,749
ใŠๆœˆ่ฆ‹ใƒใƒผใ‚ฌใƒผ๐ŸŒ•๐Ÿ” #ใ†ใ‚ใจใ‚‚ใ‚‚ใฎใตใคใ†ใฎๆšฎใ‚‰ใ—
1
406
2
2,374
100,172
๐™‰๐™‹๐™Š๐™Ž retweeted
"GPT-6 Astra, use Ableton Live and make an absolute banger". AGI is here.
490
810
364
9,064
2,851,046