3257p (@simulacrum3527) header
3257p (@simulacrum3527) profile picture

Vibecoder…… 试图搭建中转站…… 找GPT&Claude上游:[email protected]

Joined July 2026
We're giving away a Hogwarts Legacy custom wrapped NVIDIA GeForce RTX 5090 to celebrate its official release with DLSS 5 🪄 To enter: 🟢 Share this post 🟢 Comment #RTXON Terms and Conditions: nvidia.com/en-us/geforce/con…
83,279
70,926
7,494
66,790
5,687,637
中肯
19
70
10
468
100,711
linux合我PR了!
87
5
4
707
53,128
3257p retweeted
我们 ChatGPT 用户,订阅就赠送盲盒 28 连抽,还附带甜蜜恋综节目 你们 Claude 拿什么比🫵😎
34
6
1
248
12,325
Replying to @GKev1n
1
17
2,692
Anthropic has filed an S-1 registration statement with the SEC related to its proposed IPO. Read more: anthropic.com/news/s1-sec Legal disclaimer: A registration statement relating to these securities has been filed with the SEC but has not yet become effective. These securities may not be sold, nor may offers to buy be accepted, prior to the time the registration statement becomes effective. This post does not constitute an offer to sell or the solicitation of an offer to buy these securities, nor shall there be any sale of these securities in any state or jurisdiction in which such offer, solicitation or sale would be unlawful prior to registration or qualification under the securities laws of any such state or jurisdiction.
32
422
31
4,895
171,329
3257p retweeted
早上好
🤖 Made with AI
24
16
1
317
5,126
Replying to @Vindicatorn
孩子吃百家饭长大的原因是大家不忍心看着孤儿死掉愿意给一口吃的,结果现在这种除了像是厕所拉出来个小屁孩然后抱到直播间哭诉自己原生家庭
3
1
699
33,276
3257p retweeted
Everyday is zero-day these days
4,087
5,934
1,088
73,323
8,447,980
Please remove your profile picture, its an embarrasing picture of me.
sorry about your umbrella. i got excited
43
1,470
35
18,796
199,293
3257p retweeted
GPT-Rosalind surfaced a new lead for drug discovery: could restoring cholesterol transport protect neurons from degeneration? DCTN1 is a gene linked to Perry syndrome, a rare neurodegenerative disorder. In mouse-brain data, disrupting it increased expression of genes involved in making cholesterol. GPT-Rosalind suggested a possible explanation: cholesterol may not be reaching where neurons need it. That gives scientists a specific mechanism to test and, if confirmed, a direction for developing new treatments for neurodegenerative disorders. With @perturbai_tx and @scrippsresearch, a team of thousands of GPT-Rosalind agents worked in parallel, connecting gene-expression results from millions of mouse brain cells with scientific literature in just 32 minutes. We’re now expanding this collaboration to identify new drug targets. openai.com/index/scaling-ai-…
408
742
321
6,994
1,515,244
Replying to @TJwithAI
3
1
105
19,563
3257p retweeted
Deepseek克制地不像一家中国公司, 我并不是在贬低国内的公司, 而是以前所有的软件或功能都会塞满各种各样的花哨功能营销卡片,恨不得把所有东西都放进一个界面里 Deepseek就非常简洁,对话窗口就只有对话窗口,API接口就只显示API用量和余额 每次更新前也不炒作,不营销,不搞乱命名的套路,更新完就消失了,继续默默工作 能不能都跟Deepseek学一学啊...
313
85
22
4,730
295,845
为什么当下的Multi-Agent,一旦把 Agent 数量加到 8 个以上,整体智商就会断崖式下跌。在很多人的设想中,群体智能应该像分布式算力。Agent 越多,分工越细,解决复杂难题的能力就应该线性跃升。 但现实是反着的现在的多智能体架构,只要集群规模突破 4~8 个节点,系统就会迅速陷入混乱,每个 Agent 的上下文窗口被海量的冗余发言撑爆;真正有价值的高质量洞察被废话冲垮;而只要其中某一个 Agent 偶尔幻觉,整个协作链路就崩塌。 人类社会能够协同成千上万名学者、工程师完成登月工程或开源 Linux,我们靠的是介质:是论文引用网络、Git 仓库、维基百科这种允许异步发布、检索与同行引用的共享信息底座。 最近,来自哈佛、MIT、Meta FAIR 等机构的学者们发表了一篇直击行业痛点的重磅研究Scaling Multi-Agent Intelligence with Substrate-Centric Coordination。它彻底抛弃了以 Agent 为中心的传统沟通思路,提出了一个颠覆性的框架 ScaleMAS。 论文首先严谨剖析了现有 Agent-Centric 系统的结构性败局:上下文开销随节点数量呈超线性爆炸,昂贵的 Token 预算全花在了上下文上。 此外现在的 agent 节点一多,信噪比急剧下降,关键事实被大量低质回复淹没。脆弱的依赖链条导致哪怕一个 Agent 宕机或格式出错,整体任务直接归零。 我们多 agent 中传统的共享黑板虽然能缓解单点崩溃,但根本解决不了阅读负担过重和信息稀释这两个根本问题。 ScaleMAS 的解法是把学术界的网络搬进多智能体。ScaleMAS 的核心哲学非常纯粹,就是让 Agent 之间完全禁止直接对话,它们唯一能打交道的就是一个异步结构介质。 系统仅开放三个极简原子操作:发布,就是向公共介质写入结构化信息;检索,就是带着问题去介质中搜索相关信息;调阅也就是根据特定消息 ID 直接查看详情。 他们的检索的算法设计很独特,检索结果不仅看字面相关度,还会参考这条信息被其他 Agent 引用的频次。 越是被多数 Agent 验证、引用的推论,在底座中的权重就越高;毫无营养的灌水信息由于零引用,会迅速沉底。 同时由于消息与 Agent 彻底解耦,哪怕某些 Agent 意外崩溃下线,它们留下的引用痕迹依然永久有效,新加入的 Agent 可以无缝协作。 他们实测了该系统可以正向扩展到 128 个 Agent。在覆盖复杂网络研究、代码工程基准上,ScaleMAS分别拿到了 55.1%、80.0%。 这篇论文对于多 agent 中信息流的理解很深刻,参照人类的信息索引和论文式的引用机制很新颖。结果也证明了这是极其有效的。可以说是一个多 agent 的里程碑之作了。 对于 agent 自进化、多 agent 感兴趣的可以看看。 arxiv:arxiv.org/abs/2610.16120
19
47
14
280
24,590
3257p retweeted
Day 7/ We have increased default context length for GPT-6.1 Sol to 500k when used through Codex & Work, up from 272k or an ~ 80% increase. We will go higher in the future, but it helps reduce occurrence of compactions and with tasks that require keeping a lot of context and nuance over periods of time. It was carefully tested and is the new sweet spot for most of our users. You can still override it if you want to be higher our lower, with maximum context length of GPT-6.1 Sol of 1M.
1,812
571
755
14,727
2,845,878
3257p retweeted
Day 1/ We have optimized the default speed to be ~50% faster across GPT-6 Astra and GPT-6.1 Sol through the subscription across all our products and partners using Sign in With ChatGPT (including OpenCode, Pi, Amp, Devin, ...). No changes needed on your end and this should be felt within the next two hours.
Over the next 28 days, each day we’ll either ship one thing that is a clear improvement and relevant for most codex/work users or ship a full reset. Let the improvements begin.
2,555
1,021
1,174
20,636
4,206,847
3257p retweeted
GPT-6.1 is a good model. And it’s unbelievably efficient. Included in all paid plans and on the API today.
1,040
393
282
9,207
1,064,820
3257p retweeted
Tomorrow we will bring back the 5h limit for Plus accounts across ChatGPT Work and Codex. I had mentioned this a while ago, but then postponed it. This is necessary as (a) the 5h limit allows us to smoothen the load on our compute, allowing to keep the plan generous in terms of weekly usage and (b) users on the Plus plan are relatively casual and new users, but then also just accidentally eat through their whole weeks usage and then are confused, making it not a great experience. We are for the upcoming months keeping the 5h limit not enabled for Pro $100 and Pro $200 subscriptions.
5,156
760
1,586
14,623
4,463,083
终于有个观察人类怎么用 AI 的项目了! MIT、Stanford 等机构的一群研究者,刚刚做了一个很有意思的项目:AI Observatory。 它想解决一个很简单,但过去一直没人真正解决的问题: 全世界每天有几亿人在用 ChatGPT、Claude、Gemini、Grok,但我们其实根本不知道大家到底拿 AI 在干什么。 过去这类数据基本掌握在 OpenAI、Anthropic、Google 自己手里。 所以 AI Observatory 干脆把 WildChat、ShareGPT、LMSYS、Grok 等多个真实对话数据源放到一起,统一用 145 个维度分析用户行为。 目前已经收集并分析了超过 2.3 万条真实 AI 对话、约8.5 万轮对话 。 你可以看用户在聊什么、怎么提问、是不是多轮对话、有没有写代码、角色扮演、越狱、敏感内容,还可以比较不同模型、不同平台和不同时期的使用方式。 其中一个很反直觉的结果是: 47.9% 的对话和工作无关。 大量都是健康、人际关系、娱乐、性、角色扮演以及其他私人用途。 更重要的是,AI Observatory 还想继续做一件挺狠的事情: 用真实世界的 AI 使用数据,反过来审计 Benchmark。 比如以后不只是问:GPT-6 的 DeepSWE 又涨了几分? 而是问:DeepSWE 测的这些任务,在真实世界里到底有多少人真的会用? 一个模型可能 Benchmark 高得离谱,但它擅长的东西,在现实 AI 使用里可能根本没人关心。 如果这个项目后面真的能持续更新,它可能会变成一个挺重要的基础设施: 不再让模型公司告诉我们“人类如何使用 AI”,而是直接观察人类到底在怎么用 AI。 华盛顿邮报已经专门报道了这个项目。 项目目前也已经公开了 Dashboard、数据和分析代码。
26
22
2
143
11,871
在@aulimaruu 的帮助下忙话了两个半小时终于看到了这个🥲 我纯小白还对我还真挺有成就感的 洗澡睡觉!明天继续
1
1
1
319