Tsubasa Kato (@_stingraze) header
Tsubasa Kato (@_stingraze) profile picture

CEO at Inspire Search Corp. & Curioforce Corp. B20 India '23 Delegate. Building next gen search engine. Quantum web crawler designer https://nitter.cf/t.co/wo9M0cNDSL

Japan
Joined April 2020
Photo by Tsubasa Kato (@_stingraze)Photo by Tsubasa Kato (@_stingraze)Photo by Tsubasa Kato (@_stingraze)Photo by Tsubasa Kato (@_stingraze)Photo by Tsubasa Kato (@_stingraze)Photo by Tsubasa Kato (@_stingraze)Photo by Tsubasa Kato (@_stingraze)Photo by Tsubasa Kato (@_stingraze)Photo by Tsubasa Kato (@_stingraze)Photo by Tsubasa Kato (@_stingraze)
Went to a park today and used my DCR. I could have good reception there, and could hear some radio from Gotemba and more.
16
Tsubasa Kato retweeted
【超量子的拡散希望】SpringX超学校にて「量子コンピュータを作る、使う。」の講義をします 12月3日10時45分〜11時45分@梅田グランフロントThe Lab.内アクティブスタジオ 要事前申込(無料)※オンライン参加は申込不要 13時から梅田グラングリーンで開催される量子ジョブフェスタと併せて是非梅田に!
10
24
2,189
This is crazy! ☺️
They say size doesn't matter. BlackBerry walked so LongBerry could run. A ridiculously long display, a full keyboard, and absolutely no concept of pocketability. 🍇 Based on #zepir project
2
35
Tsubasa Kato retweeted
New NVIDIA paper finds that once an LLM has reasoned through a problem, the solution can run as CPU code for rephrased and re-numbered versions of it. that answers similar questions 16x to 291x cheaper. LLMs re-solve every problem from scratch, even when only wording or numbers change. Agents sending the same instructions with new data pay for that repeated reasoning each time. Their method has Claude Opus 4.5 turn a reasoning trace into small Python functions, with changeable inputs as parameters. At run time they execute on a CPU, calling a fine-tuned 1B model only for small text jobs like extracting values. The cache matched or beat the source model's accuracy on 7 of 8 math and science benchmarks, scoring 93.3% versus 83.3% on AIME '25. It handled at least 90% of rephrased or re-numbered variants. If your agent runs the same prompt on new inputs all day, cache the reasoning as code and use the LLM only on misses.
87
249
55
2,096
91,066
Tsubasa Kato retweeted
Independent mathematicians have now finished the full verification of the proof of the Riemann hypothesis announced yesterday by Anthropic mathematician Levent Alpöge. The results are real, they say. scientificamerican.com/artic…
107
767
142
3,428
394,734
Will try to read it. ☺️🧑‍🏫💡
Here is a human readable version of the Riemann Hypothesis proof: alexisdpc.github.io/riemann/
36
Tsubasa Kato retweeted
Quantum computers make errors constantly. A classical chip has to diagnose them from error signals, a million times a second, without ever falling behind. We built a neural network that does this on one FPGA, 240× more accurately than today's real-time decoders.
21
66
10
630
24,071
Today is a really exciting day for AI and other fields. Going to be studying and reading some books and designing new ideas.
30
Tsubasa Kato retweeted
Levent Alpöge claims to have solved the Riemann Hypothesis "It is natural to feel that this is an end ... but what I saw during the course of progress so amazed me that I believe it is one dawn across science of something far greater" alphaxiv.org/abs/2611.claime…
5
43
2
247
9,484
Outside today. A little chilly.
24
パシフィコ横浜にて開催中のEdge tech 2026に出展しております!! 今回は@Raspberry_Pi と@MechaTracks の2社と共同にて出展🎊 今日が最終日ではありますが、皆さんのご来場をお待ちしております!!!
6
28
2,907
Tsubasa Kato retweeted
AI runs on GPUs, so I wondered: how good would @ChatGPT be if it ran on the Super NES's best GPU, the Super FX 2 chip? With the help of GPT-6 Astra and gpt-oss, I present unofficial ChatGPT for SNES: 2.2M parameters on a real Super Nintendo. It can make mistakes, but it's fun
154
365
184
5,064
810,449
Tsubasa Kato retweeted
We shrank 1 GB of Wikipedia to 105.39 MB, losslessly. Introducing tufazip: the best entry on the Large Text Compression Benchmark using a neural network as its sole predictor. Everything above it mixes a transformer with hand-built models. Full post 👇 tufalabs.ai/research/tufazip
507
1,114
571
15,424
3,737,802
Bridge near Akihabara the other day.
1
1
863
Got a lot of coding and some PoC done today. Also researched with Perplexity Pro. Raining today in Kanagawa, and getting chilly. ☔️🧑‍💻⭐️
34
NVIDIAが小型AIスパコンDGX Sparkを4台つなげてメモリ容量512GBを実現する「DGX Spark Developer Kit」を発表、最大700Bのモデルを実行可能
10
177
36
900
85,400