Marcus Shepherd
111 posts

Marcus Shepherd
@MarcusShepher20
Father, Husband, Fast Typer
Jenison, MI Katılım Şubat 2019
131 Takip Edilen15 Takipçiler

I ran a 35-billion parameter AI agent on a $600 Mac mini.
Specs: M4 Mac-Mini 16GB RAM
The model doesn't fit in RAM. It pages from the SSD at 30 tokens/second.
On NVIDIA, the same paging gives you 1.6 tok/s. Apple Silicon gives you 30. That's 18.6x faster.
No cloud. No API keys. $0/month.
Here's what it can do 🧵
English


Hot take, Claude.AI+Claude code+cowork dispatch, and scheduled tasks absolutely SHIT all over openclaws capability with 100x the reliability. Just make skills. @trq212 dropped that bomb yesterday. Add in zero (or damn near zero) security concerns, why would I or ANYONE go that way?
I’m 100% certain that soon enough @AnthropicAI will have the agent team capability soon enough, dispatch already has that path started. Add in @obsdmd for memory, openclaw can’t touch the capability.
I have about 30 specialized skills I’ve already made or am making, Claude code runs them all. Browser ops? Skill. Research? Skill. Social media? Yep, it’s a skill too. I got red pilled initially, but it was just I recognized the agentic team capability. Then just realized I could do it myself, better.
And that’s a hill I will absolutely DIE on.
English

Every time I watch an OpenClaw influencer video I go a little greyer. What are they doing that I'm not???
I will pay $10,000 in Bitcoin to observe @AlexFinn using OpenClaw for 1 day.
BUT if he spends 50% of his time doing tech support, he owes me $5,000.
What do you say Alex?
English

@michaelfreedman Love it man. Any chance you'd support other dbs besides postgres
English

Introducing TigerFS - a filesystem backed by PostgreSQL, and a filesystem interface to PostgreSQL.
Idea is simple: Agents don't need fancy APIs or SDKs, they love the file system. ls, cat, find, grep. Pipelined UNIX tools. So let’s make files transactional and concurrent by backing them with a real database.
There are two ways to use it:
File-first: Write markdown, organize into directories. Writes are atomic, everything is auto-versioned. Any tool that works with files -- Claude Code, Cursor, grep, emacs -- just works. Multi-agent task coordination is just mv'ing files between todo/doing/done directories.
Data-first: Mount any Postgres database and explore it with Unix tools. For large databases, chain filters into paths that push down to SQL: .by/customer_id/123/.order/created_at/.last/10/.export/json. Bulk import/export, no SQL needed, and ships with Claude Code skills.
Every file is a real PostgreSQL row. Multiple agents and humans read and write concurrently with full ACID guarantees. The filesystem /is/ the API.
Mounts via FUSE on Linux and NFS on macOS, no extra dependencies. Point it at an existing Postgres database, or spin up a free one on Tiger Cloud or Ghost.
I built this mostly for agent workflows, but curious what else people would use it for. It's early but the core is solid. Feedback welcome.
tigerfs.io
English

@ctatedev Thank Chris last I tried there was support but that was a few weeks ago
English

@MarcusShepher20 It is! There were some issues up until a few weeks ago, but I believe most have been resolved
If you're still running into issues, feel free to report on GitHub
github.com/vercel-labs/ag…
English

@seekjourney Been using this for a week or so really happy with it
English

发现一个宝藏CLI 项目,思路不错,真能省token啊!
高性能 CLI 代理,将 LLM token 消耗降低 60-90%。
原理是在你执行的命令和 LLM 之间加了一层“过滤器”。
不是简单的截断,而是智能压缩,去掉注释、空白、样板代码,聚合相似项,保留核心信息。
效果有多夸张?
ls / tree:2000 → 400 tokens(-80%)
cat / read:40000 → 12000 tokens(-70%)
测试输出:25000 → 2500 tokens(-90%)
怎么用?
brew install rtk
rtk init --global
github.com/rtk-ai/rtk
中文

@bidhanxcode Explain to me what go can do that I can't already do with typescript and bun
English

@Saccc_c How do you hand code something that's vibe coded? Maybe a translation issue
English

20岁大四学生,用10天Vibe Coding手搓出来的开源项目:MiroFish,连续3天霸占GitHub Trending榜单前列,目前Star数已狂飙至8.3k,且还在持续增长中。
盛大创始人陈天桥看了Demo后,24小时内当场拍板注资3000万。
它是一个「预测万物」的 AI 仿真沙盘。只要丢给它一段新闻、政策或金融信号,它会自动生成成千上万个具备独立人格的 AI Agent,直接推演出未来的发展方向。
当你还在靠人工收集数据、靠经验做判断时,此项目已经开始用成千上万个 agent替自己预演未来了。
仓库地址:github.com/666ghj/MiroFish

Sac@Saccc_c
中文


Released today: /loop
/loop is a powerful new way to schedule recurring tasks, for up to 3 days at a time
eg. “/loop babysit all my PRs. Auto-fix build issues and when comments come in, use a worktree agent to fix them”
eg. “/loop every morning use the Slack MCP to give me a summary of top posts I was tagged in”
Let us know what you think!
English


My tiny product (ScreenshotOne.com) is now officially serving 800+ paying customers 🥳
Thanks to everyone who supported me in this journey 🙏

English

cool set of patterns for building agents agent-experience.dev
English

@trq212 How is this different from pressing WINDOWS KEY +H
English

@boristane @southpolesteve Genuinely don't understand the point in making js Frank works anymore
English

new front-end framework just dropped
$1,100 in tokens
we're so back
@southpolesteve 🐐
blog.cloudflare.com/vinext/

English

@_StanGirard Can I get my account banned for using this?
English

I reverse-engineered Claude Code's binary.
Found a flag they hid from --help.
--sdk-url
Enable it and the terminal disappears. The CLI becomes a WebSocket client.
We built a server to catch the connection. Added a React UI on top.
Now I run Claude Code from my browser. From my phone. From anywhere.
Same $200/month subscription. Zero extra API costs.
bunx the-vibe-companion
github.com/The-Vibe-Compa…
English

On Sunday, I used AI to build an app. In a couple of hours, it generated 23,000 lines of code with full documentation and 836 passing tests.
The code is amazing. The app is fully functional, well written, well structured, and has great test coverage.
At least that's what the agent told me after reviewing the project structure, completeness, and code quality.
So I fired it up and clicked the button. Nothing.
Turns out all tests passing and the agent grading its own work as brilliant doesn't mean it actually works. Now I'm debugging, filling gaps, and showing the agent what it missed.
AI can get you most of the way there, fast. But that last bit? That still needs a human who knows what they're doing. Let's not pretend otherwise.
English










