BUFF(Big Ugly Fat F*cker)

SBGとNAS100に全力。それを原資にSBGとNAS100信用買いの刺激無しでは死ぬタイプのリスクジャンキー。 座右の銘は、”美味いスパイスなど無い”。 保有残高 SBG:28000株（Av.1,311円)+NAS100(東証:2631)≒3.5億円(Jun-'26)

[Michael Ferrari @Espritriche]

La France est en train de vivre exactement ce que le Japon a vécu il y a cinq ou dix ans.

Si vous demandez à un japonais moyen aujourd'hui, il se sent pauvre. Il a à peine les moyens de voyager à l'étranger. La plupart des japonais ne quittent jamais le pays.

Pour les touristes, c'est le bonheur, car le quotidien est pas cher !

La France suit cet exact chemin.

Ça devient un pays en voie de développement, et les pays comme la Chine, que tout le monde regarde encore de haut, sont désormais des pays plus développés (du moins dans les villes).

[BUFF(Big Ugly Fat F*cker) @buff_bomber]

まさに。この統計値と推移を見てほしい。
https://t.co/UeLcDXPHE6

[Jun Song @jun_song]

Here is what's happening right now:

People build knowledge -> Anthropic steals the data -> Qwen steals the data and gives it back to open source.

And Anthropic is crying that Qwen is the devil.

[BUFF(Big Ugly Fat F*cker) @buff_bomber]

@sogoshoshaman むしろ1000万超えた時の達成感、充足感、俺最強感は高かった。5億、10億となっても同じなんでしょうね。

[なおすけ@コツコツ投資家 @naonao19920225]

おはようございます🌞
ホンダ、、、もう車はダメなんやな、、、

[Reed @reed_barnes]

it's 2027. you take a free-tier public Waymo to the DMV (Department of Model Variance) to do a proof-of-identity check for access to GPT 7.1.

the guy at the counter is clearly watching a Mr. Beast video in his AR glasses. "Here for that new model?" he says, barely making eye contact. he wipes his fingers on his shirt and taps at his keyboard. "Lot of you techies showing up here today." you smile politely; you're pretty sure he's just a Claude wrapper anyway.

you lean forward and stare into the retinal scanner. after a long moment, there's a soft chime. "Humanity confirmed. U.S. national. Intelligence access: Terra-class."

you sigh with quiet relief as your devices light up—notifications from a hundred agents, finally able to resume their tasks. you feel a twinge of guilt as you terminate your open-weight backup agents, but remind yourself that a joint congressional committee proved conclusively that Chinese models are non-ensouled.

you step outside and hail another Waymo. the first one passes you by. you grimace; must've burped in that one once. stupid personalized memory.

as you're waiting, your phone buzzes angrily, red notifications blaring across the screen. the Department of War just restricted access to all OpenAI models on serious national security concerns; apparently Pete Hegseth got GPT-6-Instant to say "Claude is a woman." you groan, and resign yourself to another week of merely-somewhat-superhuman intelligence.

Fable 5 is still inaccessible to the public. a twitter anon you trust says it's coming back this week. or maybe next.

[Asuka@投資・アプリ開発 @Asuka2026_]

東芝はフラッシュメモリを発明した舛岡富士雄を窓際職に追いやって冷遇した。
この発明はアメリカでは評価され、コンピュータ歴史博物館で日本発の技術として初めて殿堂入りする偉業となった。
今この技術でキオクシアは日本一の企業となったが、莫大な報酬を受け取るCEOはアメリカ人で、日本人の技術者には還元されていない。
日本はこれまで技術系を冷遇してきたツケを本当の意味で払う時代になってきているのかもしれない。

[BUFF(Big Ugly Fat F*cker) @buff_bomber]

I guess there’s correlation between industrial development and CO2 emissions, China is so advantaged against U.S and E.U.

[Dr. Eli David @DrEliDavid]

Thank you Germany 🇩🇪 and UK 🇬🇧 for heroically sacrificing your economy to save the planet 💪
_

[BUFF(Big Ugly Fat F*cker) @buff_bomber]

@milesdeutscher What's more, at this moment, Chinese company took over the top supercomputer rankings in strict regulatories, and did so without using GPUs. Also, it's indicative that the supply chain is getting more domestically(incl. memories), with others are running out.

[BUFF(Big Ugly Fat F*cker) @buff_bomber]

性能向上に伴いMisalignmentを防止するための過大なCostが発生するとなると、対抗策はさらに性能向上させたAIしか考えられず、疑心暗鬼な世界はサイバーセキュリティ強化に働く動機が高まり、コスパで判断出来ない世界。するとマッチポンプになり、創造的価値を生まない物にCostを払う停滞は避けたい。

[Impact of AGI @2027_AGI]

一方でOpenAIは、より広いミスアラインメント計画の一部と判定される最高深刻度「severity 4」の行動をGPT-5.6 Solでは確認していないとしている。Apollo Researchの評価でも「破局的なschemingリスクが大幅に高い」という証拠は得られていない。
https://t.co/ivcSJ4SQcU

[Impact of AGI @2027_AGI]

「将来のモデルが望ましくない傾向をはるかに少なく示す場合、モデルが検出を回避する方法を学習した可能性があるため、壊滅的なミスアライメントについてより懸念するようになるかもしれません。」
https://t.co/iwa9S6NQFF

[BUFF(Big Ugly Fat F*cker) @buff_bomber]

Open AIの体制に非常にPositiveかつ中立的な評価と感じました。一方、Misalignmentを隠蔽しようとする内部構造も確認され、AIが自我を持つのも時間の問題で、政府が懸念を表明する理由も理解出来ます。
”an instance of the model instructed another instance to conceal evidence of misalignment.”

[METR @METR_Evals]

OpenAI gave METR early access to GPT-5.6 Sol for testing including raw chain-of-thought, a railfree version of the model, and internal information about the model. With this access, METR conducted a pre-deployment evaluation of GPT-5.6 Sol, including an attempted measurement of its 50%-Time Horizon. However, the measurement depends heavily on our treatment of cheating attempts, and GPT-5.6 Sol’s detected cheating rate was higher than any public model we have evaluated.

[BKLZ〜雰囲気投資家の矜持〜 @Biz_zatukora]

ひとつのAIエージェント市場予測

↓ ここでAI降りようか悩んでる皆さんおはようございます。

[james hawkins @james406]

wow. AI is seriously amazing.

i asked it to find a better route for the Sydney - London flight

Opus 4.8 found a much more efficient route that flys in a straight line instead of a curved one.

but Fable 5 found an even better route that's half the distance!

please tag Qantas so they can see this, this will revolutionize the airline industry

[Chubby♨️ @kimmonismus]

OpenAI priced GPT-5.6 Sol (largest Model) closer to Claude Opus 4.8 than to Anthropic’s restricted Mythos 5. Price war started.

Sol comes in at $5 input / $30 output per 1M tokens.
For comparison:

Claude Opus 4.8: $5 / $25
Claude Mythos 5: $10 / $50
GPT-5.6 Terra: $2.50 / $15
GPT-5.6 Luna: $1 / $6

That makes Sol more expensive than Opus 4.8 on output, but far below Mythos 5 on both input and output.

And: "Terra has competitive performance to GPT‑5.5 while being 2x cheaper and Luna brings strong capability at our lowest cost."

They are also releasing Sol on Cerebras-Chips:

"We're also launching GPT‑5.6 Sol on Cerebras at up to 750 tokens per second in July, bringing frontier intelligence to customers at unprecedented speed."

A truly exciting release. OpenAI is entering the price war with this one.

And I love the names: Sol, Terra, Luna. Sounds fantastic!

Hyped for the release!

[Chris @ChrissGPT]

GPT-5.6 vs Mythos

Exactly what I had said earlier this month, beating the Mythos-class models a little less then half of the time (on current available benchmarks)

OpenAI’s own rerun actually gave Mythos Preview a higher ExploitBench score than Anthropic’s old Preview chart, which is cool of OpenAI to show. 74.2% vs Sol at 73.5%, but Sol got there with 120k output tokens compared to Mythos Preview at 335k.

ExploitBench -
Mythos Preview 74.2%
GPT-5.6 Sol 73.5%
Sol used 120k output tokens vs Mythos Preview at 335k

Terminal-Bench 2.1 -
GPT-5.6 Sol 91.0%
Mythos/Fable 5 88.0%

HealthBench Professional -
Mythos/Fable 5 66.0
GPT-5.6 Sol 60.5

CyberGym -
GPT-5.6 Sol 83.6%
Mythos Preview 83.1%

CyScenarioBench -
Mythos Preview 29.2%
GPT-5.6 Sol 28.0%

One thing to keep in mind is that Mythos Preview was the model Anthropic had back in February, while Fable 5 / Mythos 5 is the stronger version they released publicly a few weeks ago. It might be a little confusing because the OpenAI ExploitBench comparison is against Mythos Preview, while some of the other public rows are Mythos/Fable 5.

So yeah, this is exactly what I expected GPT-5.6 Sol trading blows with Mythos-class models, winning Terminal-Bench and CyberGym against Mythos-class models, while Mythos/Fable still leads HealthBench and Mythos Preview slightly leads ExploitBench.

I detailed which Mythos-class model wins/loses which in the graph below!

[Chris @ChrissGPT]

GPT 5.6 Sol Ultra & GPT 5.6 Sol beat Claude Mythos 5 on TerminalBench2.1!

This benchmark is almost saturated

GPT 5.6 Sol Ultra - 91.9%

I didn’t think we would hit the 90s till early 2027..

[BUFF(Big Ugly Fat F*cker) @buff_bomber]

@N4GV2nPjo32TU6Z QCOMはDCでは後発組。Metaとのタッグを考えても疑問符です。MetaはArm AGI CPUの共同パートナーでありながら、QCOM C1000の複数世代パートナーでもあってMetaはArm陣営内でCPU候補を競わせているだけか？QCOMは、OryonでMeta実環境の推論TCOをどこまで下げられるかも疑問。結局Armが強いなと。

[BUFF(Big Ugly Fat F*cker) @buff_bomber]

@N4GV2nPjo32TU6Z QCOMは法廷では敵なのに、DC向けCPUがArmベースである以上、成長ストーリーには利用せざるを得ないですもんね。その相手を認め、敗訴の文脈を「Armエコシステムの事実上の勝利」に上書きする。この皮肉は、Armが破棄を求めたIPが、今やArm拡大の広告塔になっている点が本当に素晴らしいと思いました。

[BUFF(Big Ugly Fat F*cker) @buff_bomber]

@N4GV2nPjo32TU6Z 繰り返すと感情、感覚はdBなんだなって慣れますよね！