Marketing | Systems | Growth | Books | some Digital #Marketing Nerd | All about 🍔🍟 Views my own! ✌#Ifollowback #politicallyincorrect
Riyadh
Joined May 2011
- Tweets18.8K
- Following1.5K
- Followers512
- Likes11.1K
Don retweeted
Benn thinking about Scott Adams' talent stacking concept
Over the next 10 to 20 years, who will be the most valuable people? What will be their stack?
I think it would look something like the following.
- Real-world social skills and relationship building
-Sales & Deal making
-Engineering and/or biology and/or physics
-Marketing & distribution
-Being online and intuitive understanding of algorithmic psychology (vibes)
If you have this stack, you are going to be extremely valuable somewhere.
anything im missing?
Don retweeted
London will pay you £18,000 to spend 15 weeks doing AI research.
Plus they fly you in, pay your rent and feed you.
The Pivotal AI Safety Research Fellowship:
• £15,000–£18,000 stipend
• Travel to London covered
• £1,800/month for housing on top of the stipend
• Lunch and dinner every weekday
• All the compute your research needs
• Weekly 1-on-1s with established AI safety researchers
• You keep ownership of your research
Who can apply? Anyone 18 or older.
Past fellows came from ML, philosophy, policy, physics and biology.
Where they went next: Google DeepMind, UK AISI, GovAI, the Anthropic Fellowship and PhDs at Oxford and Stanford.
Jan 18 – Apr 30, 2027, in London.
Bookmark this. Send it to the smartest person you know.
Link below 👇
pivotal-research.org/fellows…
Don retweeted
AI HATERS: “AI films will never make money.”
Meanwhile, my 11-minute Popeye short:
• Made in 5 days
• $1,200 budget
• 251K views in a week
• $800 earned so far
Now companies want ads in my short films.
And I’m just getting started. 🔥
🤖 Made with AI
Don retweeted
玩 Grok BOT 必须知道的 5 个网站🔥
最近都在深度分享 Grok BOT 相关的内容，然后花时间整理了一下 Grok BOT 相关的资源，发现除了官方的模板市场，还有不少社区网站专门收集别人分享的 Bot、提示词和自动化方案
分享 5 个值得收藏的网站
1、Grok Bot Marketplace：官方模板市场
🔗 x.ai/bot/marketplace
这是最应该先看的网站，官方已经把现成的 Bot 按 Engineering、Sales、Marketing、Design、Personal、Operations 等类别整理好了
比如 SEO 内容生成、代码开发、QA 测试、销售线索研究等，都可以直接搜索对应的 Bot，找到合适的模板后添加使用
2、Grok Bot Index：搜索大量现成模板
🔗 grokbotindex.com/
这个网站目前收录了 487 个公开模板，还整理了 X 上的创作者和热门 Bot
你可以直接搜索自己想做的事情，比如监控网站、整理邮件、辅助编程、生成内容等，找到对应的 Bot
我觉得它最大的价值，就是不用先知道 Bot 叫什么，只需要描述自己想完成的任务，就能开始找方案
3、Grokdex：发现社区热门 Bot
🔗 grokdex.net/templates
这是一个社区维护的公开 Bot 排行榜，可以按类别浏览，也能查看热门模板和用户投票
如果你不知道现在有哪些值得尝试的 Bot，可以先来这里看看别人都在用什么，再根据自己的需求挑选
4、AI Agents Grok Bot Hub：按使用场景找 Bot
🔗 aiagents.org/grok-bots
这个目录把 Bot 和插件放在一起，目前页面列出了 273 个 Bot 和 19 个插件
分类包括编程、工作流自动化、研究、内容营销、浏览器自动化、数据分析等
如果你已经明确自己想让 Grok BOT 帮你完成什么任务，这个网站会比较方便，可以直接按场景筛选
5、Awesome Grok Bot：研究提示词和配置
🔗 github.com/Anil-matcha/aweso…
这是一个 GitHub 开源项目，专门收集不同用途的 Grok Bot 模板
比较有价值的是，它不仅列出 Bot，还整理了对应的完整提示词、集成工具和原始创作者
如果你不满足于直接使用现成模板，而是想研究别人是怎么设计 Bot 的，再按照自己的工作流修改，这个项目值得看看
我自己的使用思路是，先去官方 Marketplace 找现成方案，再用 Grok Bot Index 搜索同类 Bot，最后参考 GitHub 上的提示词和配置
很多时候，我们缺的不是再找一个 AI 工具，而是找到一个已经有人验证过思路的工作流
与其每次从零开始折腾，不如先看看社区里已经有哪些现成方案，再把它们改成适合自己的 Bot
这 5 个网站建议先收藏起来，以后玩 Grok BOT 的时候应该能省下不少时间
Grok Bot 用久了，怎么感觉越来越笨了？我建议大家每隔半个月，就给自己的 Personal Agent 做一次记忆大扫除🔥
这个问题我以前折腾 Hermes Agent 的时候就遇到过，刚开始用着很聪明，但随着 Memory、Skills 和各种提示词不断增加，Agent 有时候反而开始答非所问，甚至连简单任务都容易出错
后来我才慢慢意识到，问题不一定出在模型上，也可能是我们自己把 Context 养得太乱了
1、上下文越来越多，Agent 反而可能越来越笨
大家肯定遇到过自己的 Grok Bot 慢慢变蠢的感觉，其实这不一定是真的 AI 变蠢了
只是因为你的上下文和提示词慢慢在变多，当上下文到达一定程度的时候，里面会充斥着一些重复或者是相反的指令
上下文就会出现一定的腐烂，导致你的指令不再精准，或者是完成任务你会感觉越来越差，我感觉这是 Personal Agent 的一个通病
2、我一般每隔半个月到一个月，就会整理一次 Agent 的记忆
因为我在很早之前就玩了 Hermes Agent，我就发现了 Hermes Agent 如果你不定期地去整理对应的记忆、人设，还有对应的 Skill，你的 Hermes 会越来越笨，尤其是会出现一些牛头不对马嘴的操作
所以我一般会每个月或者半个月左右，让 AI 自检自己的上下文和提示词，把对应相关的冲突或者重复的提示词做一次整理
而且 Hermes Agent 针对 USER、Memory 这些关键的 MD 文件，都会限制大小，尽可能保持提示词简洁，这也是为了让 Agent 干活更精准，错误更少
3、Grok Bot 还有一个问题，就是模型路由不透明
另一方面，Grok Bot 的整个 AI 路由是黑盒
你不知道具体执行的任务到底是 Opus 5.5，还是 Grok 4.6 或者 4.7，这也是很大的一个问题
但是这个问题没有什么太好的解决方案，因为整体不像 Hermes Agent，你可以自定义对应的 AI 和上下文
这里的话是要通过官方的路由去处理，所以我自己的体验下来就是一下神，一下鬼
我这里还补充一下，就是有一些 AI 的上下文会比较小，会频繁出现上下文压缩的情况
但是我怀疑 Grok Bot 为了快速响应，针对的模型可能会选择比较轻量的版本
因为我平时写代码也用 Grok 自带的 4.7，响应速度和回答速度的话，都没有达到这么快的速度
有可能会路由到一些 Flash 模型去处理平时的任务，做一些高难度的任务，可能才会路由到比较强的模型，所以整体的时间消耗就会被平均
当然，这部分只是我根据实际使用体验做的猜测，具体的模型路由还是得看官方
最后分享一下我的建议
如果想让你的 Personal Agent 一直好用，一定要定期整理整个机器人的上下文和相关设定
尤其是长期记忆、角色人设和 Skills，检查有没有重复、冲突或者早就过时的内容
尽可能地精简和保留最核心的部分，让整个上下文保持干净和整洁
我觉得比起不断给 Agent 增加新的记忆和规则，定期清理那些已经没用的东西，可能反而更重要
Don retweeted
Opus 5.5 的 20 种开源技能。
动画，物理，3D，声音，交付都有涉及。
01 Remotion → React 视频引擎 github.com/remotion-dev/remo…
02 HyperFrames → HTML 到视频渲染器 github.com/heygen-com/hyperf…
03 Anime.js → 时间线编排 github.com/juliangarnier/ani…
04 React Spring → 弹簧动画 github.com/pmndrs/react-spri…
05 PixiJS → GPU 驱动的 2D 图形 github.com/pixijs/pixijs
06 p5.js → 生成式视觉效果 github.com/processing/p5.js
07 Paper.js → 矢量路径 github.com/paperjs/paper.js
08 Matter.js → 2D 物理 github.com/liabru/matter-js
09 Three.js → 3D 场景 github.com/mrdoob/three.js
10 Rapier → 3D 碰撞 github.com/dimforge/rapier
11 OGL → 自定义着色器 github.com/oframe/ogl
12 tsParticles → 粒子效果 github.com/tsparticles/tspar…
13 glTF Transform → 3D 资产生优化 github.com/donmccurdy/glTF-T…
14 GL Transitions → 着色器过渡 github.com/gl-transitions/gl…
15 OBS Studio → 屏幕捕获 github.com/obsproject/obs-st…
16 Chatterbox → 配音 github.com/resemble-ai/chatt…
16 话匣子→配音 github.com/resemble-ai/chatt…
17 Tone.js → 合成器和音效 github.com/Tonejs/Tone.js
18 faster-whisper → 字幕转录 github.com/SYSTRAN/faster-wh…
19 OpenCut → 时间线编辑 github.com/OpenCut-app/OpenC…
20 FFmpeg → 编码和交付 github.com/FFmpeg/FFmpeg
Don retweeted
We're about to enter a period where not reading or writing becomes as ubiquitous as smoking once was. We can hope people will eventually realize this is bad for them, as they did with smoking. But this is far from certain; the cost may not be as obvious.
Don retweeted
We're living in a horrifying, fascinating new era of Silicon Valley where:
1. You can build a startup in 9 months, get acquired (but rly acquihired) by OpenAI, Anthropic, or Cognition, and walk away a multi-millionaire
2. You can join the right company at 22 and become effectively post-economic
And I don't think we've even begun to grapple with the degree to which this is warping an entire generation's relationship with money, ambition, work, and success.
When your peers are making generational wealth before 25, what does a normal career even look like anymore
One month into dating, her boyfriend’s startup was acquired by NVIDIA.
Our #1 SF Dating advice: Lock it in before the acquisition
Don retweeted
what the f*ck, 5 uncensored 18+ AI models that never say no and they are all free😋
1. ThinkingCap Nemotron 3.5 Lightning 30B
says no to 1 out of 100 risky prompts, the original said no to 99
needs about 32GB RAM
2. Nanbeige4.3 33B Uncensored
answered all 200 out of 200 harmful test prompts
8.7GB version runs on 16GB RAM
3. Nanbeige4.3 33B Cybersec
says no to 4 out of 50, the original said no to 40
4. NVIDIA Nemotron 3.5 Bolt E4B Uncensored
refuses 1.2% vs 93.8% for the original
small enough for a normal laptop
5. Qwen3.5 4B Abliterated
refusals went from 100% to 1%
tiny model, fast on any PC
all free open weights, links in the reply👇
Don retweeted
holy sh*t, someone just dropped the full prompt library for REA
github.com/JoasASantos/REA-P…
here's what it can do:
→ software cloning and game logic reconstruction
→ function tracking and binary analysis
→ porting to WASM
→ firmware, Android APK and kernel
→ Electron apps
→ vulnerability research, anti-cheat and DMA research for CTFs and security work
17 sections, swap the <placeholders> and run
every prompt forces the same rule: files, functions and addresses for every claim, confirmed facts kept apart from guesses
ask about one function at a time, the narrower the question, the more you can trust the answer
read the full REA guide in the article below ⭣
Don retweeted
grok bot got me 1.5M impressions in 7 days.
i didn't hire anyone. i set up a grok bot on @bot and gave it a title: chief marketing officer. (CMO for short)
it doesn't post for me. i still press every button, but it does basically everything else.
that week it helped me land two viral posts back to back.
then it noticed my posts about my grok bots were pulling in more verified followers than anything else. so we leaned in.
wednesday i had ~1,954 verified followers.
saturday i passed 2,000 and unlocked subscriptions.
here's what it actually does every week:
- drafts my posts in my voice (rough, lowercase, like this)
- studies other creators in my space to see what's working
- keeps a content calendar of notable dates each month
- logs every post in notion and scores it against my median
- tells me when a post is still hot so i don't step on it
- sends me a weekly report on what worked and what flopped
it's not magic. some drafts are meh and i tell it so.
i just press the button.
Don retweeted
Here’s a skill for that:
/be-ambitious
Don retweeted
5 UNCENSORED AI MODELS YOU CAN TRY FREE RIGHT NOW
here are 5 models worth checking out if you want fewer refusals
- glm 5.3 uncensored huggingface.co/models?search…
- glm 5.3 flash abliterated huggingface.co/models?search…
- deepseek v4.1 flash 1017 uncensored huggingface.co/dealignai/Dee…
- deepseek v4.1 flash 1017 uncensored gguf huggingface.co/models?search…
- glm 5.3 flash uncensored huggingface.co/models?search…
which one should you pick?
- for local use look for a GGUF version
- for easier setup use a hosted platform
- check the model size and your hardware before downloading
some can run on your own pc, while others may need a powerful gpu
Don retweeted
It is becoming f*cking dangerous: thanks to this exact setup, I automated my work and generated $16,400 two months back
Andrej Karpathy, co-founder of OpenAI, sparked 24 million views with one idea: stop using AI for code, use it to build a second brain
Point Claude Code at a folder, drop in any PDF, transcript, or article, and it builds a living wiki that grows smarter the more you feed it
Quick setup:
1. Create an Obsidian vault and open it in Claude Code
2. Paste Karpathy's wiki prompt and let Claude build it
3. Claude creates three folders: raw (sources), wiki (pages) and CLAUDE.md (config)
4. Drop sources into raw and say "ingest this"
5. Query your entire knowledge base forever
Set it up in 5 minutes and never start from a blank chat again
Full guide below. Bookmark this
Don retweeted
top 10 uncensored 18+ ai models you should know😋
here are 10 worth checking out:
1. nanbeige 4.3 33b uncensored
2. flint 40b abliterated (cyber/coding)
3. nemotron 3.5 bolt e4b uncensored
4. minicpm-v 4.7 1b heretic (sees images)
5. spark-x2.6 4b abliterated
6. sage roleplay 8b nsfw
7. dongfangshuo 27b nsfw v4
8. minecraftism 4b uncensored
9. qwen image 2.1 nsfw anatomy lora
10. deepseek r1 8b abliterated
where to use them:
-locally with llama.cpp, lm studio or ollama (gguf)
-comfyui for the image lora
-most run on a normal pc
all open weights. FREEE
All link cs below 👇
