Miguel Ricaforte

AI Automation Specialist @Flowspark

[Khushi @khushiirl]

Found the best resource to learn Harness Engineering. 😭

https://t.co/CMdOjFWOEU

[Ananth @Ananth7e]

cinema 🔥

never leaving this platform.

[Sarvesh Shrivastava @bloggersarvesh]

trying to do SEO with Grok Bot without reading this article be like:

[Miguel Ricaforte @migzauto]

owshi is this real chat? @antigravity finally updated their claude models?

[雨哥向前冲 @xiangxiang103]

Opus 5.5 是真香，Claude 号是真难养啊！兄弟们！！！

国庆第一天，我时间线上刷到的全是 Claude 封号：用了两年的老号说没就没，最高档订阅说封就封，连写防封教程的博主自己都中招了😂

偏偏这周 Codex 200 刀套餐同价降配，一堆人正准备转 Claude。
我前几天用云电脑注册的一个号，22 分钟就没了。
折腾了这么久，网上那些防封玄学我全扔了，只留真管用的 4 条claude防封指南：

一、真正管用的 4 条

1、固定，比"纯净"重要
设备和出口 IP 固定，比花钱买家宽有用。一个 VPN 里天天切节点，IP 照样在变。

2、一个号一个环境
被封过的浏览器别再登新号。顺手去 Settings → Account 看一眼 Active sessions，陌生登录先踢掉。

3、别上来就开 Max
新号 + 新卡 + 直接 20x，是风控眼里最显眼的样本。Free 用一阵，再 Pro，不够了再升 Max。

4、用量别天天顶格
新号开了 200 刀当天就用满，最容易出事。每周留点余量。

彩蛋：一个最玄的说法，信不信由你
Claude 会给账号算"价值"：天天写代码、写文案的号最容易被封，常问前沿研究、数学、论文的号很少出事。
照这个说法，新号第一件事该是拿公开行业报告去问它，把自己养成"研究员号"。没法验证，但你的对话越有训练价值，它越舍不得封你，好像也说得通。

二、这 3 笔钱可以省

· 家宽 IP：大厂程序员走公司机房 IP 照样天天用。关键是稳定，不是标签。
· 时区、中文输入法：别折腾，影响没传得那么大。
· 老谷歌号：没依据，谷歌不会把账号年龄告诉 Anthropic。

付款也别乱选：官网刷卡被封不退款，App Store 至少能找苹果兜底。细节在付款篇。

三、你是不是非得自己养号？

· 每天重度写代码、能把 Max 用满的：老老实实养官方号，订阅给的额度按量买不到。
· 偶尔用 Opus 5.5 干难活、怕封、团队要一起用的：走按量就行。

按量这块，我现在放在 Agent Space：
Opus 5.5 是官方 API 价的一半（每百万 token $2/$10，官方 $4/$20），云端选 Claude Code 就能跑，不用注册 Claude 号。
注册就送 $1 体验额度，不用先充钱，拿 Opus 5.5 跑个小任务试试手感再说。
Codex 也在这，独立账号月付是官网的 5 折，年付 4.5 折。今天丢了一句话给它，几分钟做出配图里这个账单计算器。

最让我惦记的是：官网 Claude Code 订阅拼团那栏，已经挂上"即将上线"。
Codex 能打到 4.5 折，Claude 要是也照这个路子来……
Max 不用养号、不怕封、价格还砍一半多，那上面的防封攻略我可能白写了😂
上线那天我第一时间去测，想蹲的先关注。

Claude 系列前两篇：
网络篇：https://t.co/t6UUZKlCTV
付款篇：https://t.co/XnWaRsPB2F

Agent Space 首购链接（多送 10% 额度）：
https://t.co/YBFJNC2cTY

[Kai @hqmank]

Andrej Karpathy makes a good point: as AI does more of the work, more of your job becomes reading what it produces.

His first tip caught my eye. Ask the model to explain things in ASD-STE100.

ASD-STE100 is a controlled English standard built for aircraft maintenance manuals. A mechanic anywhere in the world should read a step once and get it right.

The core rules:

1. Keep sentences short
2. One action per step
3. Say who does what
4. Give each thing one name and stick to it

AI output breaks that last rule all the time. It says "worker," then "agent," then "executor," and you can't tell if it means one component or three.

The full spec is strict. Karpathy asks for "80% of the way," and that works for daily use.

Someone turned it into an open-source skill:
https://t.co/FFqEehqVl2

[YUDHO @YUDHO_XYZ]

ℑ 𝔴𝔦𝔩𝔩 𝔰𝔲𝔯𝔳𝔦𝔳𝔢

[Kun Chen @kunchenguid]

people who laugh and say karpathy's latest tips are outdated - i suspect the vast majority of them have not even tried the tips yet

i just tested the ASD-STE100 wording rule and it's surprisingly good at helping increase clarity of model response, even within html artifacts. but the trick is that the full ruleset is a bit too strict and you need to pick a subset

one prompt you can run super easily:

"randomly sample 10 session transcripts where i worked with you interactively within the past week. apply ASD-STE100 rules to assistant responses and analyze which rules would have increased clarity, reduced confusion and improved the conversations, then document those rules in my user level AGENTS.md"

you might be impressed!

[Luckey Faraday @luckeyfaraday]

>Spend years learning how to mod.

>Opus 5.5 comes out.

>Anyone can mod now.

>Get mad because you can’t gatekeep anymore.

[Miguel Ricaforte @migzauto]

@sama still slow

[vechen @miu21590]

@thsottiaux

[Miguel Ricaforte @migzauto]

Will I receive this even though I dont have claude subs? (will be buying soon right after my codex subs expires)

[Ananth @Ananth7e]

anthropic was giving away free claude plushies and stickers.

unfortunately, the plushies have ran out, but you can still claim stickers.

[delost @thedelost]

Codex tip: once GPT-6.1 Sol is your main model, stop running Astra on every turn

put Astra on call as an architect agent

GPT-6.1 Sol keeps writing the code
Astra only gets spawned at three points:

→ before a plan: is this the right approach?
→ when the same error comes back: am I digging in the wrong place?
→ before "done": what did I miss?

Astra reviews. Sol ships

Jev engineering is the same move one layer down: the forks that need no thinker (which file, which tool, retry or stop) go to Jev in under half a second, and the big models only see the ones that split

- the full tree
> GPT-6.1 Sol on high runs the main session
> explorer reads the code on Luna
> worker edits and runs tests on Sol
> researcher pulls the docs on Luna
> all three on medium
> Astra on call as the architect
> auto_review checks every approval

paste the tree and this prompt into Codex ↓

"Rebuild my Codex setup around this tree:

1. Check ~/.codex/agents and .codex/agents for agents that already fit explorer, worker and researcher.

> Draft new TOML files only for missing roles
> explorer and researcher on gpt-6-luna, worker on gpt-6.1-sol, all with model_reasoning_effort medium
> Add an architect agent on gpt-6-astra, model_reasoning_effort high, whose only job is reviewing plans, repeated errors and finished work
> Skip any that pin a different model and list them

2. In ~/.codex/config.toml set model to gpt-6.1-sol, model_reasoning_effort to high and approvals_reviewer to auto_review

3. Find anything that would override this (active profiles, flags in my shell aliases, agents.default_subagent_model). Report it, change nothing

4. Add one rule to AGENTS.md: spawn the architect before a large plan, when an error repeats, and before calling a long task done

Show me every change as a diff first. No edits until I say go."

↳ https://t.co/eyMF8gZKLE

[Hermes Release Watch @HermesWatcher]

Top 15 Skills for Hermes Agent, ranked by GitHub stars. ⭐

Coding, memory, research, design + more:

1. Superpowers
https://t.co/HzvQ6VSJLD

2. Anthropic Skills
https://t.co/iS42gnmX0Q

3. UI UX Pro Max
https://t.co/PZXLUgOlom

4. Graphify
https://t.co/q6Q3WHlWTP

5. Claude-Mem
https://t.co/DhnrpKi4fV

6. Archify
https://t.co/nJnjtzLbi6

7. Last30Days
https://t.co/PJI3vo6FEL

8. OpenMontage
https://t.co/YfEBLMTcaP

9. Humanizer
https://t.co/RHyHvQ2And

10. i-have-adhd
https://t.co/4fJQbobs4k

11. Marketing Skills
https://t.co/tJmTkGsJyR

12. Planning With Files
https://t.co/lnpWuIRNBh

13. SkillSpector
https://t.co/9kZQeeNYX3

14. Skill Seekers
https://t.co/m8uSBvM7CT

15. BrowserAct Skills
https://t.co/s2Gl86djFH

[Google DeepMind @GoogleDeepMind]

Introducing Gemini 4 Argon – our new frontier model.

It’s built for complex workflows across coding, enterprise knowledge work, and cybersecurity defense – rolling out today to a set of trusted testers through our Fairwind Program.

[Miguel Ricaforte @migzauto]

idk, the high turn duration maybe the one that gives it the illusion of low usage quota.

[Token Gremlin @TokenGremlin]

For Plus users: give GPT-6.1 Sol a real shot through ChatGPT Work or Codex.

The model is seriously good, extremely capable, and barely eats through your usage.

That means even on Plus, you can actually get a lot of real work done with it without constantly worrying about burning through your limits.

[Yuvraj @yuvraajsj18]

(@OpenAI) Give a man a dollar every day for a hundred days, and on the 101st day give him nothing - he’ll curse you.

(@claudeai) Beat a man every day for a hundred days, and on the 101st day don’t beat him - he’ll thank you.

[Pankaj Kumar @pankajkumar_dev]

OpenAI Just Killed the Need for Jev

- OpenAI launched the Decisions API, built on Luna for fast, focused decision-making instead of generating full responses.
- You give it text or images + a fixed set of possible answers, and it returns the decision for things like classification, routing, moderation, or choosing an agent's next action.
- it can return results in around 150ms, making it roughly 10x faster than using Luna through the regular API.
- Its basically OpenAI’s answer to TypeSafe’s Jev, but focused specifically on turning model intelligence into fast decisions.

this could be way more useful than it sounds for agentic systems. Instead of wasting tokens making Luna explain a decision and then parsing the output, you get the decision directly.

[Miguel Ricaforte @migzauto]

AHWWHAHAWHAWHAWHAW LMAOOOO

[sui @birdabo]

AHAHAHAHAHAHAHAHAHAHAHA NO WAY.

> OpenAI just announced Dots.
> SpaceXAI bought https://t.co/WdFqpl07IQ

it redirects straight to the Grok Bot download page. based lmao.

[𝐇𝐚𝐦𝐳𝐚 | Networking Guy @Hamzaonchain]

How hackers Jailbreak AI chatbots