Ash Stormer

ASD 患者，活着就行。已进入贤者时间。手残党, 在 dota2 浪费的大量时间.

开源

[Ash Stormer @_BrickRunner]

@SnowyWar36965 @materialinhai 因为用不了cc被封号了，考虑上车用opus5.5，但是怎么感觉额度已经砍了？

[the tiny corp @__tinygrad__]

Kind of tempted to try Opus 5.5, but then I remember that I have to install Claude Code spyware on my machine, it'll refuse half the things I want it to do, the money funds an aspiring bioterrorism org, and it'll use word vomit phrases like "pacing the frontier." I'll pass.

[Ash Stormer @_BrickRunner]

@MiaAI_lab AA should not be trusted any more.

[Ash Stormer @_BrickRunner]

@nihui hold my beer 2GB?!

[Ash Stormer @_BrickRunner]

Mimo v2.6 Pro 的性价比有点儿可怕啊

[Ash Stormer @_BrickRunner]

时间线上至少一半测试4.7退步，或者不如K3的，astra事件后，榜单已经失去意义了，可能早就失去了。平衡模型能力，让它稳步提升，可能真要点儿水平，OpenAI都不一定把握好。

[Ash Stormer @_BrickRunner]

@Bhavani_00007 it’s not a leap in version number. so i guess not a big model like k3. just more data and distillation. like 5.6 sol( not a big model too)

[Ash Stormer @_BrickRunner]

AGI should contains both system-1 and system-2.

2026 we are not going to see a model with automatical thinking budget? not a dummy router but from the model itself. we are not even close to AGI.

[Dan McAteer @daniel_mac8]

Opus 5.5 supposedly coming this week.

Rumor is it’s better and cheaper than Astra.

I just hope it talks like a normal human.

[Ash Stormer @_BrickRunner]

@yifanxu_ephai 可能luna max也能秒，astra 不好评价，还需要调教

[Ash Stormer @_BrickRunner]

只能说, 在LLM加持下, 剧本没问题了. 但是视频嘛, AI味还是太重了, 虽然未来可期, 不过什么时候真的能替代不好说, 就像现在LLM还是远远无法替代真正的写作者.

[Rainier @mtrainier2020]

归墟 1-9， 成本差不多是20K USD, 一个人的作品。
真人演员压力山大。

[Ash Stormer @_BrickRunner]

yas. system 1 is the same as next token prediction. just a smaller sampling space and less intelligence. A ultra flash llm or diffusion llm will do the same.

[Dmitriy Kovalenko @neogoose_btw]

If you didn't feel the ongoing revolution, here is my second project I built with jev.

Your software is drastically outdated. It contains a lot of instructions your CPU executes one by one. Why if we can simply ask Jev for the next instructions to execute

Introducing Jevassembler!

It is a new approach to execute something on CPU. You don't need to wait for code to be written, you don't need to compile anything and YOU DON'T NEED CODE.

You just give it a task and Jev predicts the next instruction to execute. In runtime.

who needs compilers if you have AI

[Ash Stormer @_BrickRunner]

@ILikeTouZheLe @riverleaf88 该升, 我的12 mini 之前从bug 17 升级到 26.6 突然很流畅变更耐用了😄 一下打消了我换新机的念头

[Ash Stormer @_BrickRunner]

@IFaRehJiFI @LothardPaladin @mylifcc 那 cursor 已经废了，老马是真的抠，估计是整体产品都在往grok迁移。

[Ash Stormer @_BrickRunner]

@ProbiusOfficial OpenAI也早就把 codex 和 chat 至少合并成一个app了, kimi的产品真的有问题. 还把 OK Computer 砍了换什么claw. 笑死了.

而且jev再一步说明, 生成速度+便宜就是一切啊, 他们还整个2.8.... 相反deepseek越想越牛逼, 抓住了重点

[Ash Stormer @_BrickRunner]

@ProbiusOfficial 然后, 他们所模仿的 A\ 已经合并了Work和Chat, 我不是产品, 但是稍微想想就知道, 不该开这个品类, 你只需要一个叫作AI的产品. code 也可以合并.

[Ash Stormer @_BrickRunner]

@huacnlee 是, 这个变成我的潜在使用心理障碍了, 每次想到这个, 我就让 AI 写 go 或者 c 🤣

[Ash Stormer @_BrickRunner]

本来还想着养号有朝一日订阅A/ 买好100分的家宽，聊个网页就给我封了🤡

[Ash Stormer @_BrickRunner]

@Fei2411 5.6 Luna Max 斩杀线以下的模型都不应该推出, 不过OpenAI产品策略问题不是一天两天了

[Diogo Almeida @CompleteSkeptic]

After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?

I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev

• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions

AFAICT the shortest path to AI-based economic revolution