School Boy

专注AI分享 | Focus on AI sharing

[School Boy @ValueMaking]

终于自己琢磨实现了官方模型和自己用API的agent模型的一起使用了，可以在官方登录的状态下选择自己的模型。我的使用步骤具体如下👇

[School Boy @ValueMaking]

https://t.co/oPMrmv87qf

[School Boy @ValueMaking]

Vercel CEO 评测:Kimi K3 有 raw IQ 🧠

@rauchg:stealth evals 显示 Kimi K3 网安 top-tier,不是刷分。

▸ Kimi K3:raw IQ,真本事
▸ Sol:cyber 飞跃,但更贵

刷分质疑被推翻。

https://t.co/IzwH2UNno7

#KimiK3

[Guillermo Rauch @rauchg]

Based on internal evals:

▪️ Kimi K3 is top-tier at cybersecurity
There is chatter on X that Moonshot benchmark-overfit. These are stealth evals. Model has raw IQ.

▪️ Sol is a leap ahead in cyber capability
At a significantly higher cost, but quite remarkable still.

▪️ Fable refuses everything
We couldn’t get it to complete the run at all. What’s interesting is that Sol in comparison was much more open to helping with defensive cyber hardening

TL;DR: frontier, open-weight cybersecurity capability is here. Try it on https://t.co/HedSpZpo1a for defensive purposes.

[School Boy @ValueMaking]

按场景选模型 cheatsheet 📋

@bindureddy 推荐:
▸ 设计:GPT-5.6 Sol
▸ 简单 agent:Grok 4.5
▸ 中等:Sonnet/Kimi
▸ 难:Sol/Fable
▸ 视频:Seedance 2.0
▸ 语音:Hume/11Labs

没有万能模型,按场景选。

https://t.co/azePS0YdsL

#AI

[Bindu Reddy @bindureddy]

Best Model For The Use Case

design - GPT 5.6 sol
chat - Flash 3.5
agentic (simple) - Grok 4.5
agentic (medium) - Sonnet 4.6 / Kimi
agentic (hard) - Sol / Fable
pdf - Grok 4.5
image - GPT 2.0
video - Seedance 2.0
voice - Hume / 11Labs
music - Suno

Use all the best models in one place and automatically route to them on ChatLLM

[School Boy @ValueMaking]

LobeHub:Agent 首席运营官 🎯

8万星 https://t.co/e4drn0oq6w

把多个 agent 组织成 7×24 运营。

▸ 从单 agent 到 agent 车队
▸ 7×24 不间断运营
▸ Chief Agent Operator 管理

agent 从工具变成基础设施。

#Agent

[School Boy @ValueMaking]

"Fable 太危险"突然不危险了 🎭

@RnaudBertrand:开源追上后,Fable 不再"太危险"。

▸ 之前:太危险,限时
▸ 现在:永久保留
▸ "危险"= 竞争定位

safety 被当护城河用。

https://t.co/rA4Va0rEsq

#Fable5

[Arnaud Bertrand @RnaudBertrand]

Funny how now that Chinese open weight models have closed the gap, Fable is suddenly not "too powerful and dangerous for humanity" anymore.

[School Boy @ValueMaking]

Addy Osmani:生产级 agent skills 🏭

7.9万星 https://t.co/C1YJK157S4

Google 工程师精选的生产级 AI 编程 skill。

▸ 不是实验玩具,能上生产
▸ Addy Osmani 出品(Google)

agent skill 从实验走向生产。

#ClaudeCode

[School Boy @ValueMaking]

字节跳动开源 deer-flow:长程 SuperAgent 🦌

7.7万星 https://t.co/Sdv7fwF7oO

能研究、编码、创作的 long-horizon agent。

▸ 字节出品,大厂背书
▸ 长程任务(research+code+create)
▸ 开源

大厂开始开源 agent 基建。

#Agent

[School Boy @ValueMaking]

多模型共识工作流 🤝

@blader:
1. Fable 写计划
2. Codex + Kimi 给第二意见
3. 修订到收敛

▸ 不同模型 cross-check
▸ 议会制 > 单一权威

没完美模型,有完美 council。

https://t.co/MEFxIn1dLY

#ClaudeCode

[Siqi Chen @blader]

my new favorite of way of using fable:

step 1. ask it to write a plan

step 2: "please get second opinions from codex CLI using gpt-5.6-sol @ max effort and kimi CLI using kimi 3. Revise your plan with any sound findings. repeat until convergence or up to 5 rounds."

[School Boy @ValueMaking]

GitHub 官方 spec-kit:规格驱动开发 📋

12万星 https://t.co/UrfJsMYCyf

▸ GitHub 官方背书新范式
▸ 先写规格,agent 实现
▸ 从写代码到写规格

软件开发的范式转移。

#ClaudeCode

[School Boy @ValueMaking]

Codex/Fable 拒修漏洞,Kimi 全修了 🚫

@callebtc:安全报告漏洞,Codex/Fable 因 guardrails 拒绝,Kimi K3 全修。

▸ 安全守卫拦住安全工作
▸ Kimi 无限制直接干

安全税太重会反噬。

https://t.co/6XgLOcWFtA

#Kimi

[calle @callebtc]

I have a report full of security issues of a software I'm working on.

Codex won't fix them because of Cyber guardrails
Fable won't fix them because of Cyber guardrails

Kimi K3 fixed them all. No restrictions, just gets the job done.

This will end badly for OpenAI & Anthropic.

[School Boy @ValueMaking]

Cline 实测:Kimi K3 vs Fable 5 修同一个 bug 🐛

▸ 都修好了
▸ Fable 赢速度:3.4x 快,730K token
▸ Kimi 赢成本:1.2M token 更便宜

没有唯一赢家,看你卡在时间还是钱。

Fable = 快,Kimi = 省。

https://t.co/G4xASLIT2R

#Kimi

[Cline @cline]

We tested Kimi K3 and Fable on a real bug from the Cline repo, and found that while both models were able to fix it - Fable wins on speed & Kimi wins on cost.

- Kimi used 1.7x more tokens than Fable (1.2M vs. 730K)

- Fable finished 3.4x faster - 3.5 min and 18 tool calls vs. Kimi’s 12 min and 34 tool calls.

- Kimi cost 2.3x less ($0.92 vs. $2.13) thanks to its 3.3x per-token discount

Both runs used the same Cline harness, and the traces indicate that Kimi is RL trained to spend more tokens thinking and verifying before completing.

This is the first time we've seen an open weight model compete head to head with SOTA. Congratulations to the @Kimi_Moonshot team on this milestone!

[School Boy @ValueMaking]

George Hotz:愿意给 Kimi 付费 💰

"Kimi 是 user-aligned,我乐意掏钱"
GLM-5.2 = Opus 级,Kimi K3 = Fable 级。

▸ tech 大佬用脚投票
▸ user-aligned > safety-first

用户拥护的模型在赢。

https://t.co/1yKjrj8g8n

#Kimi

[the tiny corp @__tinygrad__]

Started with the $39 plan and used it up, @Kimi_Moonshot is a user-aligned company I'm happy to send money to. GLM-5.2 feels Opus tier, Kimi K3 feels Fable tier!

[School Boy @ValueMaking]

planning-with-files:agent 长任务规划 📋

2.5万星 https://t.co/P9jgs8efIE

把计划写到文件,不靠记忆。

▸ 长任务计划容易丢
▸ 写文件,跨 session 持久
▸ 规划外部化,上下文留给执行

记忆靠不住,文件靠得住。

#ClaudeCode

[School Boy @ValueMaking]

headroom:给 LLM 上下文做压缩 🗜️

5.9万星 https://t.co/qgXOLfcsrR

压缩工具输出/日志/文件再喂给 LLM。

▸ 上下文贵,压缩 > 扩张
▸ 工具输出压缩后省一半

context 是新瓶颈。

#ClaudeCode

[School Boy @ValueMaking]

DeepSeek 被指蒸馏 Fable 5 🔄

@synthwavedd:DeepSeek 疑似大规模收集 Fable 5 输出做蒸馏。

▸ 用对手输出训练自己
▸ 蒸馏是开源追上的捷径
▸ 法律灰区,心知肚明

开源优势可能建在蒸馏闭源上。

https://t.co/0YYwUM0Zgc

#DeepSeek

[leo 🐾 @synthwavedd]

🧵 DeepSeek appear to have engaged, or be engaging in, a large-scale operation to collect outputs from proprietary models (including Claude Fable 5) for certain requests via their API as part of a distillation effort.

After seeing such claims circulating earlier today, we conducted an investigation into them on our Discord. We found that, when "Deepseek V4" was used within OpenCode - via their official API - for complex prompts (i.e. 3D games) and combined with a knowledge-related query, the model provides virtually identical outputs to Fable 5. CoT structure is also very different from what is typically expected from Deepseek models. Both of these behaviours revert to what is expected for V4 when simpler prompts were used.

When "Deepseek V4" was asked to incorporate answers to questions related to cyber or bio tasks that we verified hit Fable's classifiers into its 3D games, outputs tanked in quality. This is very difficult to explain unless the request was routed to Fable and fell back after hitting a classifier.

For complex code prompts without anything else mixed in, like 3D games, outputs were remarkably similar to those produced by Fable 5.

We were able to produce these results most consistently via OpenCode and the official Deepseek API combined with a prompt that specifies a complex code task. Deepseek have continued to modify their routing system since, as we have observed changes in behaviour and CoT style compared to those seen previously. Our investigation was conducted from 7AM-8AM PT.

[School Boy @ValueMaking]

Anthropic:Claude 多 agent 协作 🤝

Managed Agents 支持 agent 间通信:
▸ 不同 agent 用不同模型/工具
▸ 共享 sandbox 和 vault
▸ 任务委派协调

从"一个干所有"到"agent 团队"。

https://t.co/jt516slnjN

#Claude

[ClaudeDevs @ClaudeDevs]

Another pattern is agent-to-agent communication, allowing Claude to coordinate and delegate tasks.

Claude Managed Agents supports this with multi-agent: agents can use different models, prompts, and tools while sharing sandboxes or vault credentials.

[School Boy @ValueMaking]

Victor Taelin:大规模用 AI 心得 🧠

"不审查代码,但必须审查它的选择"

▸ 旧:逐行 review 生成代码
▸ 新:review 架构/策略选择
▸ 细节可以信,选择必须审

审查从代码行升到决策点。

https://t.co/2XEB4JOJQ5

#ClaudeCode

[Taelin @VictorTaelin]

I think I finally figured out how to use AI at scale

of course, the fact Fable is good is part of it. but I also changed how I work, and it all comes down to one key realization: you don't need to audit the code, but you NEED to audit the *choices* it made. if you just do that, things will work out, and you'll never lose a codebase to chaos. with Fable at least, the following seems to hold:

if I give it a good decision:
Fable implements it PERFECTLY

if I let it decide instead:
Fable may make some bad choices

that's how I see Fable: as a perfect execution machine capable of converting good decisions into good codebases, no matter how large. given a concrete plan, it lands its implementation. but when anything is underspecified, it can, and will, make bad choices. that's what you must audit.

"while working on this, which choices did you make that you're not confident of? list all."

then, you just review that. not the git diff, not 1000's of lines of code. just the choices it made along the way.

below is a fresh example. overnight, I asked Fable to fix an issue related to MatMul parallelizing worse than expected. it tracked the culprit with perfection, and landed a solution that DID work. but the solution was not general. it just doubled a buffer, which coincidently fixed the program at hands, but the underlying issue was still present.

when it completed the job, it declared success. if I just merged it blindly, the issue would still be dormant. that's the main mistake one can do with AI. instead, I asked it to spell out all decisions it made, spotted the bad one, corrected its course, and now the codebase is clean, correct and the issue is gone for good

I really think that if you do that religiously - i.e., NEVER merge without this "which decisions you made?" audit - you can go VERY far without ever reading a single line of code. at least on Bend, this is working incredibly well. despite heavy use of AI to implement an ungodly amount of features I could never dream of, the codebase is still in a superb state, with no signs of degradation

fresh example below ↓

[School Boy @ValueMaking]

claude-hud:Claude Code 可视化仪表盘 📊

2.6万星 https://t.co/WkE7ylXowU

让 Claude Code 不再是黑盒:
▸ 实时看 context 用量
▸ 看当前在用哪些工具
▸ agent 状态可视化

可观测性是 agent 刚需。

#ClaudeCode

[School Boy @ValueMaking]

Fable 5 = Mythos + 重重围栏 🚧

Fable 5 是 Mythos 加 heavy guardrails。

▸ 你用的是阉割版 Mythos
▸ 真正强模型内部有,不发

付费拿安全版,完整版在实验室。

https://t.co/f8zoKEphQT

#Fable5

[Chubby♨️ @kimmonismus]

As much as I'm happy that Fable 5 is staying, I'm equally concerned that Opus 5 will be an underwhelming release, similar to Sonnet 5.

What gives me hope, however, is that Mythos preview rumors surfaced back in March, and Fable 5 is known to be Mythos with heavy guardrails. Therefore, I expect the next version, Fable 5.6 or at least 6, to arrive soon.

[School Boy @ValueMaking]

David Sacks:有人从 Claude 换 Kimi 🎯

理由:"直接干活,不 lectures"

▸ Claude 安全拒绝 = 摩擦
▸ Kimi 少说教,直接做
▸ "woke 模型是美国竞争力敌人"

安全税太重,用户用脚投票。

https://t.co/6YxNVa5CqE

#Kimi

[David Sacks @DavidSacks]

OH: “i’ve switched to Kimi from claude for a bunch of work. it’s just so much more fun because it just does the thing instead of lecturing you”

Woke lobotomized models are the enemy of American competitiveness.