Guest modehighFollowing graphHigh overlap
@ayi_ainotes

@ayi_ainotes · X · Global · OTHER

Opens the X profile. · Are you @ayi_ainotes? Claim this page

Free users are seeing the saved basic result as their first meaningful view, with deeper evidence still locked.
Account noteOpen

Real X profile and recent post evidence are attached for @ayi_ainotes; some items are still pending moderation.

Risk reading

High uncertainty

Evidence

Medium · 40

The record has unresolved uncertainty from risk tags, moderation, mixed signals, or thin context. Treat this as a review prompt.

Nothing on this page is an identity, safety, or transaction verdict — it is collected evidence plus interpretation.

Main basis

Collected profile snapshot is availableCollected posts are availableModeration items are still pending

Reading limits

Does not predict future behaviorSaved media may be staleCommunity feedback is still thinModeration is not final

Image inventory

Image inventory is stale

24

Saved media

Saved images exist, but the latest capture is past the freshness window. Refresh before relying on this inventory.

Post images

16

Unique images

16

Linked posts

16

Freshness

Needs refresh

Latest captured media: 2026-08-22 21:21

Refresh only adds new media. Saved media stays archived even after the creator deletes the originals.

Quick read

Key account facts

Open

Registration history

11 mo

Follower / following

61.1K / 481

Posting evidence

30 collected

Profile density

Checkable profile evidence

Community calibration

Does this score match what you see? Your signal helps calibrate the account.

Loading...

Radar score

78

Risk

high

Followers

61.1K

Following

481

Posts tracked

30

Data status

Evidence available

A real collected snapshot backs this page (updated 25 day(s) ago). Reviews and moderation may still revise this reading.

Source: Live collected (third-party API)Last collected: 2026-08-22 21:21Stored records: 1Collected posts: 30Media assets: 24Public reviews: 0
Result source: Saved + collected evidence

Saved server record plus real collected X evidence.

Latest collection run: Collected

Status: collected 30 recent post(s).

Started at 2026-08-22 21:21 · Finished at 2026-08-22 21:21

📖 Member preview

Preview

Full media inventory

Preview count only; full archive opens after member access

Linked posts

Members can inspect which posts support saved media

Interaction structure

Members see more context behind score movement, not just the score

Risk reasons

Members can read why the page asks for caution or review

History changes

Members can inspect selected historical archive and timeline context

Context only — not identity or safety conclusions.

Free: one full view per account. Member: full score logic + more reviews. Pro: deep archive + corrections.

Collected X evidence

Collected

Profile snapshot

Collected profile

Followers

61K

Following

481

Tweets

17K

Recorded at

2026-08-22 21:21

Collected media evidence

Library note

Previewing first 12 of 24. Open the archive tab for the full media library.

Open media library
Protected media

Sign in as a member to view this video or protected image.

2026-08-22

Followers

61K

Following

481

Tweets

17K

Source

third_party_api

@ayi_ainotes

AYi

Collected at: 2026-08-22 21:21

AI 实用主义,专注AI落地的组织和业务know how,关注商业,职场,投资,业务出海, 大厂组织发展专家 × 组织心理学,Prompt专家,分享有用的 AI 实践,也分享工具之外的深度认知 以术入道,以道御术,KEEP BUILDING 合作/交流:DM/TG @AYi_AInotes

Account created: 2025-10-05 01:50Next external refresh: 2026-08-25 21:21

Recent collected posts

tweet2026-08-22 08:29

RT @AYi_AInotes: Codex额度这事我看完全程,只能说OpenAI这波沟通烂到离谱,@OpenAI 这次把付费核心用户彻底得罪惨了, Codex额度突然大幅缩水的事炸了锅,官方负责人Tibo出来回应,说没偷偷调限额,是很多人在用sub2api把订阅转成API共…

@AYi_AInotes@OpenAIOpen on X
tweet2026-08-22 07:53

@dontbesilent 栋哥被纳瓦尔拉黑这个事,怎么说呢,既荒诞又治愈,就好像把创作者最深的一层精神内耗给戳穿了。 先说原因,估计是x自动翻译以后推到了纳瓦尔的for you页面,因为都是社会学,哲学,商业这些同类的话题,但纳瓦尔跟栋哥八竿子打不着,语言不通、从没提过他、对方关注列表甚至还是 0,居然都能被纳瓦尔拉黑,这大概是我见过最能说明互联网恶意毫无道理可言的真实案例了。 我们很多做内容的人只要一看到差评、被骂或者被取关拉黑,第一反应永远是往内审判自己:我是不是哪句话没说严谨?我是不是冒犯到谁了?我是不是水平不行? 我们总本能地以为别人的排斥是有逻辑、有理由的。 但现实往往残酷又好笑,互联网上绝大多数的恶意、喷和拉黑,纯粹是低成本的随机动作。 对方可能只是那两秒钟心情烦躁,可能只是算法推送了一条他看不懂的自动翻译,手指轻轻一滑,拉黑了。 他甚至连你是谁都不知道,你却在屏幕这头反思了三个晚上哈哈哈, 我账号做起来之后也是不是被人黑,被人怼,被人拉黑,一开始也会内耗, 后来想明白了,互联网从来不是一个讲道理的法庭,它本质上就是一个巨大的布朗运动场, 别人的情绪投射根本不是针对你的作品,而是针对他自己那一刻的混乱状态。 做内容最重要的心理建设,就是意识到你永远不可能靠讲道理去抚平没有道理的排斥。 把别人的随意当成对自己的审判,是创作者最不划算的自耗。你的真诚和表达欲,只该留给那些能接住和欣赏你的人。

@dontbesilentOpen on X
tweet2026-08-22 06:47

被骂了整整两个版本的 Gemini Flash,这次居然真的把场子救回来了。也算是把 Gemini 3.6 丢的脸全挣回来了哈哈, @OfficialLoganK Kilpatrick 和 @sundarpichai 昨天都在发推庆祝:Gemini 3.7 是 Google 史上增长最快的一次模型发布。 讲真这次还真不是官方自嗨, 回看前代,3.5 和 3.6 被开发者喷得有多惨,这一代的反弹就有多凶,之前很多人骂它为了省 token 把脑子省没了,写代码动不动就崩,前端生成稀烂,结果距离 3.6 发版仅仅 3 周,Google 就搞出了一场教科书级的救场。 官方直接把 3.7 Flash 定位成了"迄今最智能的工作马(workhorse)模型",最狠的提升全砍在实际干活的刀刃上: 第一是长周期工程与 Agent 容错。DeepSWE 评分从 48.6% 一口气拉到 65.3%,遇到编译报错不再瞎改,而是真会看测试反馈去自我纠错;WebDev Arena 直接干到 1588 分,单 Prompt 搓出能玩的网页终于不再是 PPT 忽悠了。 第二是价格和速度直接掀桌,入门价打到 0.75 刀输入加 3.75 刀输出每百万 token,直接砍半,同时还能保持 340+ tok/s 的极高吞吐和 1M 上下文。 这件事最值得琢磨的信号是:大模型竞争的胜负手,正在从比拼旗舰天花板,转向比拼牛马及格线。 以前做 Agent 系统大家只能咬牙上昂贵的前沿旗舰,因为小模型动不动跑偏死循环; 当一个几毛钱的 Flash 级别模型在代码闭环上能咬住顶级模型时,整个应用层的算力成本结构被瞬间重写了。 虽然顶级前沿模型负责探索能力的边界,但真正让成千上万个 Agent 在后台 24 小时跑出商业闭环的,永远是这种又便宜又扛造的重型工作马模型

@OfficialLoganK@sundarpichaiOpen on X
tweet2026-08-22 05:27

@AnthropicAI 内部工程师最近高频在用的一个提效神技,整个提示词居然只有一句话,而且我发现这里面还藏着一个特别狠的工程认知:HTML > Markdown。 Claude Code 团队的工程师 @trq212 发了他们团队内部常用的极简 Skill:/eli5。 在终端里敲一行 /eli5 <任何问题>,Claude 不会给你甩几千字的技术长文,而是强制生成一个大图少字的交互式 HTML 页面。 核心 Prompt 简单到令人发指:
Explain like I'm someone who knows nothing about this topic, using a HTML artifact with big pictures and few words. 没有长篇大论的 system prompt,全靠模型本身的指令遵循和 HTML Artifact 的可视化渲染。 他们在内部主要拿它干三件事:
1️⃣理解复杂模块(how does this module work)、 2️⃣复盘设计取舍(why did we make this tradeoff)、 3️⃣排查线上事故(what caused this incident)。 比如输入解释 Discord 机器人,它直接用卡通图和箭头画出 网站 ↔ 机器人 ↔ 数据库 ↔ 客户端 的 3 步链路,一眼看懂。 安装方式两行命令搞定: claude plugin marketplace add anthropics/claude-plugins-community claude plugin install eli5@claude-community 我发现这里藏着一个很狠的工程认知:HTML > Markdown。 因为纯文字解释不仅人类读得累,而且很容易用一堆术语掩盖逻辑断环。 如果强制生成大图,哪里链路不通一眼就能揪出来。 有人可能会说这纯属浪费 Token,先生成详细内容再简化,相当于多付了一倍算力。 但 Anthropic 团队算的是另一笔账:Token 消耗是机器的,人类工程师的认知带宽才是整个研发链路里最贵的东西。 真正顶级的工程能力不是堆砌黑话,而是能把最复杂的系统一秒降维给别人讲明白。 https://t.co/tbGyNs4rCv

@AnthropicAI@trq212LinkOpen on X
tweet2026-08-22 04:43

Dario 说 AGI 未来一两年内就到,必须提前做好准备; Sam 在旁边从容接招:我们既要追求 AGI,也要确保安全可控。 如果这两位真在 B 站开直播搞一场 AGI 巅峰辩论,大概就是这个场面了吧🤣 这张图整个画面是用最新的 GPT Image 2 搓出来的,十多条中文弹幕、双人辩论字幕、加上一整套完整的 B 站网页端 UI,GPT Image 2 居然一个错别字都没出, 咱们放大看几个细节,这代模型的进化有点吓人: 第一是中文排版与多行文本渲染。右侧十几条弹幕("AI安全太重要了"、"两位CEO都很克制")、底部双人同框字幕、顶部导航分类,全中文无一处乱码,字形结构和间距完全符合中文排版规范,彻底告别了以前 AI 生图那种鬼画符。 第二是对复杂 UI 拓扑结构的理解。B 站 Web 端直播间的层级被极其精准地还原了——从左上角的直播时间戳、播放器控制条、观看与点赞数据,到右侧高能榜、弹幕发送框、甚至两人的企业 Logo 徽标,全部严丝合缝落在该在的交互坐标上。 第三是人物神态与光影质感。Dario 眉头微蹙、神情紧绷的严肃感,对上 Sam Altman 标志性的松弛微笑,连镜片上的室内反射光和西装面料的自然折痕都抓得极准。 以前做一张这种级别的整活截图,找素材、扣像、对齐 UI 得折腾半天;现在原生多模态不仅能理解画面,还能把中文语境、产品逻辑和人物微表情一次性对齐。 我觉得文生图的竞争已经不是比谁画得好看了,关键看谁对真实世界复杂符号系统的理解更精准!

📊 Metric notes

6 items

Registration history

11 mo

Created at 2025-10-05 01:50

Audience scale

61.1K

Follower count at the last collection.

Follower / following

61.1K / 481

Followers >> following · Followers far outnumber following — a strong structural signal. Ratio: 127.

Posting evidence

30 collected

Latest collected posts become the basis for content and risk interpretation.

Profile density

Checkable profile evidence

Enough stored structure and content samples — no empty-account penalty.

Community sample

0

Too thin for a platform conclusion.

🗺️ Identity map

RegionGlobal
saved recordNo stronger location signal was found.
LanguageOther

🧩 Judgment stack

Primary identity

Image seller

Traffic funnel

Content themes

Threesome / 3PGroup / 4P+FootImage sellerVideo sellerPrivate chatPromotion onlyText / story

Monetization path

Paid contentCustom order

Trust signals

Stable identityLow mutual-promo risk

Risk signals

Traffic funnel riskUnclear pricingAgency risk
🧭 Why this type

Profile: Saved server profile for @ayi_ainotes. The visible score now blends the durable lookup result with collected X evidence, media inventory, audience structure, and interaction traces; risk remains separate.

Collected posts: 5 saved recent post(s)

Media evidence: 24 saved media asset(s)

Community: 0 public review signal(s)

Location is self-declared.

🪪 Credit signals4

Account nature

Traffic funnel

Strong

Based on saved profile state, 5 collected post(s), 24 media asset(s), and 0 public community signal(s).

30 source(s)collectionevidence

Positive signals

Stable identity / Low mutual-promo risk

Moderate

0 higher-weight review signal(s) and no public Photo Match attachment are currently available.

2 source(s)reviewcollection

Risk signals

Traffic funnel risk / Unclear pricing

Moderate

0 caution/risk review signal(s) and 1 pending governance item(s) are attached.

4 source(s)reviewmoderationevidence

Suitable for

Real X evidence is attached for @ayi_ainotes, so the profile no longer depends on a pure lookup-time identity guess.

Weak

Not enough reviews yet.

0 source(s)reviewcollection

🎬 Content experience

54/100

Visual appeal

6.5

Exposure level

Unknown from public sample

Stimulation

6.1

Originality

6.0

Update stability

5.0

Paid-match confidence

3.2

📡 Platform signals

P1

Following circle

thin

Analyzed

0

Priority candidates

0

Review queue

0

From accounts this handle follows; review-queue means lower priority only.

Open graph view

Site behavior

thin

7d views

2

Favorites

0

Reports

0

Photo Match

Verification records

No live record

No live public Photo Match result is attached yet.

Use Photo Match when someone needs to prove that a specific photo or teaser video matches the person who appeared at verification time.

Community stance

How the crowd currently leans

Signal strength: thin

No community signals yet — the first tracked review will appear here.

🪞 Identity signals

X account age11 mo
Identity confidencemedium
Archive freshnessfresh

🧮 Score detail

9 items
Base profile+19
Follower quality+16
Audience structure+10
Profile density+0
Media depth+14
Interaction trail+6
Positive feedback+0
Account age+8
Risk penalty-9

Featured media archive

Narrative fallback
Real X evidence is attached for @ayi_ainotes, so the profile no longer depends on a pure lookup-time identity guess.
1 moderation item is still pending, so score interpretation can still move.
Risk stays elevated, so this page should be read through caution notes and approved evidence rather than score alone.
Approved image and video archive will appear here.

🏛️ Evidence stack

0 archive cards · 2 timeline notes

0 archive card(s), 2 timeline note(s), and 0 approved review(s) currently back this score.