here you goThings I’ve made, read and saved — shared here.

Archive

9,565 entries: 46 from X, 8,923 from Pinboard, saved 1998 to 2026.

Browse by tag, site, author, folder or language

Tags

Sites

Authors

Folders

Languages

Tags and folders

Showing 1–50 of 545Folder notes ×

September 2026

english-expressions

  • that's rich coming from you
  • ok no more mr nice guy
  • let's skedaddle
  • let me be frank
  • now listen hear pal
  • cup of joe
  • you're barking up the wrong tree
  • how you like them apples
  • newsflash buddy
  • what on gods green earth
  • quit horsing around
  • no shit sherlock
  • Get a load of this guy
  • I don't give a rat's ass
  • let's blow this popsicle stand
  • you’re skating on thin ice buddy
  • good grief
  • well whoop-dee-doo!
  • heavens to Betsy
  • I’m Joshin’
  • cut me some slack
  • frick
  • give it a rest
  • that really irks me
  • pet peeve
  • douche-wagon
  • blabber mouth
  • ding-a-ling
  • doodad
  • thingamajig
  • eureka
  • Let me be frank
  • speak of the devil
  • hold yer horses there bud
  • quit pokin' the bear
  • oh fiddlesticks
  • look for the silver lining
  • she looks like death chewin' on a cracker
  • Woah there, bucko
  • That’s rich coming from you
  • geronimo
  • kerfuffle
  • humdinger
  • there’s more where that came from
  • that's a bunch of baloney
  • you're darn tootin'
  • don't bet the farm
  • you're cruisn' for a bruisin'
  • holy smokes
  • scared the bejesus outta me
  • no more mr nice guy
  • see you in the funny papers
  • screwed the pooch
  • skedaddle
  • your mouth is writing checks that your ass can't cash
  • living the dream
  • what can I do you for?
  • well, what do ya know?
  • well I'll be a monkey's uncle
  • you're barking up to the wrong tree
  • how d'ya like them apples
  • spoiler alert
  • holy tamale

notes#english

July 2026

QR Code

https://qr.jim.sh/ -- This is a real TrueType/OpenType font that turns bracketed text into QR codes during text shaping. There is no separate image generation or preprocessing step: type text like [hello], apply the font, and the font's built-in OpenType rules render the QR code. Because the QR code is still text, you can copy and paste the rendered QR block as ordinary characters, store it in plain text, or mix it inline with regular Latin text. Text outside brackets remains readable.

notes

June 2026

Sandbox for Coding Agents

Container Use -- Sandboxed dev environments for coding agents.

GitHub - abshkbh/arrakis: A fully customizable and self-hosted sandboxing solution for AI agent code execution and computer use. It features out-of-the-box support for backtracking, a simple REST API and Python SDK, automatic port forwarding, and secure MicroVM isolation. Perfect for safely running, testing, and backtracking multi-step agent workflows. See Arrakis: How To Build An AI Sandbox From Scratch - Abhishek Bhardwaj, OpenAI talk at 2026 AI Engineer Conference.

GitHub - tizkovatereza/awesome-ai-sandboxes: A list of cloud sandbox providers for AI agents. Information sourced exclusively from official docs and landing pages. · GitHub

See also Tools for Agents

notes

Jetbrains

Junie "The coding agent that works with any model you choose."

curl -fsSL https://junie.jetbrains.com/install.sh | bash

Mellum by JetBrains: Fast language models for real-world AI workloads. "The open-source LLM by JetBrains, optimized for real-world development workflows where latency and performance matter most."

Mellum2 -- Best for low-latency, high-performance inference

A 12B-parameter open-source mixture-of-experts model, designed for real-time workflows, combining strong coding and language capabilities with exceptional efficiency.

Mellum 2 on huggingface

notes

pi.dev

Pi Coding Agent.

I use it with GitHub - mitsuhiko/pi-ds4: Run deepseek4 locally on metal right from within Pi · GitHub, which allows me to call GitHub - antirez/ds4: DeepSeek 4 Flash local inference engine for Metal and CUDA · GitHub directly from within pi, without having to start a separate server.

Running Cohere North Mini 1.0

via

uv tool install git+http://github.com/Blaizzy/mlx-vlm
uv run mlx_vlm.server --model CohereLabs/North-Mini-Code-1.0

Add this following line to ~/.pi/agent/models.json under providers key:

"mlx-lm": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "CohereLabs/North-Mini-Code-1.0"
        }
      ]
    },

Pi will use the above server running on port :8080:

pi --provider mlx-vlm --model CohereLabs/North-Mini-Code-1.0

The above will download the weights.

notes

May 2026

Attention

How's Your Attention Span? - by Alessandra Codinha

I’ll give you a program to follow every night, a very simple program. For the next thousand nights, before you go to bed every night, read one short story. That’ll take you ten minutes, 15 minutes. Okay, then read one poem a night from the vast history of poetry... Read the great poets, go back and read Shakespeare, read Alexander Pope, read Robert Frost. But one poem a night, one short story a night, one essay a night, for the next 1,000 nights. From various fields: archaeology, zoology, biology, all the great philosophers of time, comparing them. I want you to read essays in every field. On politics, analyzing literature, pick your own. But that means that every night then, before you go to bed, you’re stuffing your head with one poem, one short story, one essay—at the end of a thousand nights...

notes

Github Actions

Alternatives to Github Actions:

  • CodeRabbit
  • Pullfrog -- open source CodeRabbit alternative that runs in GitHub Actions. see

Configure Pullfrog to review new PRs, auto-address new human reviews, triage/label issues, make plans and PRDs, autofix CI, autofix merge conflicts, or anything else—just mention @pullfrog to assign tasks.

billing:

  • first 30 runs free, then pay as you go: 7¢/run plus model costs. for model costs, either:
  • use Pullfrog Router to access any model at cost (no markup)
  • bring your own key or Claude Code subscription

notes#github

Speech To Text

Handy Is a speech to text application that runs on Mac OS. And you trigger that by pressing the default key - option+space.

notes#stt

ds4

ds4.c is Salvatore Sanfilippo’s deliberately narrow inference engine for DeepSeek V4 Flash on Macs with 128GB+ of RAM only. It is not a generic GGUF runner and it is not trying to be a framework. It is a model-specific native engine with a Metal path, model-specific loading, prompt rendering, KV handling, server API glue, and tests. (May 2026).

notes#ds4#inference-engine

April 2026

OS Sandboxing

Agent Sandboxes

notes

Browser Automation

What can Browser Run do?? a demonstration of the capabilities of cloudflare Browser Run

Browser Use - The way AI uses the internet

remote browser infra for ai agents and web scrapers

agent-browser | Browser Automation for AI CLI designed for AI agents. Compact text output minimizes context usage. 100% native Rust.

# Navigate and get snapshot
agent-browser open example.com
agent-browser snapshot -i

# Output:
# - heading "Example Domain" [ref=e1]
# - link "More information..." [ref=e2]

# Interact using refs
agent-browser click @e2
agent-browser screenshot page.png
agent-browser close

--

Autobrowse -- "TL;DR Browser agents have an amnesia problem. They re-discover every site from scratch on every run, paying the full discovery tax forever. Autobrowse fixes that by letting an agent iterate on a real task until it converges, then graduating the winning approach into a durable, reusable skill. That skill is memory the next agent, teammate, or customer can pick up and run without “re-learning” what's already been learned."

notes

Agent Skills

<<<<<<< HEAD Every Company’s First AI Strategy Should Be a Skill Library

GitHub - fayazara/macos-app-skills: AI coding agent skills for building, shipping, and maintaining native macOS apps · GitHub

Every Company’s First AI Strategy Should Be a Skill Library by Hiten Shah; Jun 2026.

7deacaffba7df66dad346cc7fdc10742fd4aa1b0

GitHub - microsoft/waza: CLI / Framework for Agent Skills - create, test, measure and improve skill quality and effectiveness · GitHub

GitHub - Yeshwanthyk/agent-skills: Agent Skills for Claude Code, Codex, and Pi: stateful-systems and interactive-system-explainer · GitHub

--

Skill to write a will.

--

Drop your screenshots and this Claude skill turns them into animated onboarding. Github location.

--

tufte-viz Claude Code skill -- using Edward Tufte data visualization principles.

notes

March 2026

AI Cloud

AI Cloud computers

Agent Computer -- "Cloud computers for AI agents. Virtual sandboxes that spin up in ~0.5s with shared and swappable credentials and filesystems. Your agents delegate work via SSH with zero friction in a peristed agent ecosystem."

Perplexity Computer "is a general-purpose digital worker that operates the same interfaces you do. Computer creates and executes entire workflows, capable of running for hours or even months. Chat answers. Agents do tasks.Computer works."

See also: exe.dev

notes

AI PRD

How to Write Product Requirement Docs (PRDs) in the AI Era

a few of the things a vibe coded prototype doesn’t specify:

1. What is the **hypothesis** of the change?
2. How does this **fit within the overall strategy**?
3. Is it going to be **rolled out to everyone or A/B tested**?
4. What **constitutes passing metrics** to graduate the experiment? 
5. What are **non-goals and potential side effects** that are worth accepting?

notes

Luck Surface Area

How to increase your luck surface area - Codus Operandi

L = D * T, where L is luck, D is doing and T is telling. This demonstrates clearly that the more you do and the more people you tell about it, the larger your Luck Surface Area will become.

LSA

Some Signals:

  • Founder DNA — you’ve built something from scratch, or you will
  • Extraordinarily high agency and product taste, obsess over craft and design details

notes

LiquidAI

LFM2.5-230M: our smallest model yet, built to run fast anywhere (CPUs, NPUs, and GPUs) to enable agentic tasks on phones, robots, home and network automation devices. X/liquidai; Jun 26, 2026.

--

Set Up LiquidAI LFM2.5-8B-A1B + Hermes Agent locally in 10 minutes via X/noctus91. ^db45f5

--

Liquid Demo | Tool-Calling Agents on Consumer Hardware with LFM2-24B-A2B - YouTube

"LocalCoWork"

No Cloud, No Waiting: Tool-Calling Agents on Consumer Hardware with LFM2-24B-A2B | Liquid AI; Mar 2026.

--

maximecb/bebelm: CPU-only, pure-Rust implementation of LiquidAI's LFM2.5-8B-A1B · GitHub; Jun 2026.

This model is very capable and has only 1B active parameters, making it possible for the model to run at interactive speeds without a GPU.

This package intentionally has very few dependencies and requires no extra system packages to run, making it easy to build and run. This is a library crate which can be imported into your Rust projects

notes

February 2026

microvm

MicroVMs

GitHub - mishushakov/libkrun-go: Go bindings for libkrun -- Go bindings for libkrun, a dynamic library for creating lightweight microVMs using KVM (Linux) or HVF (macOS/ARM64).

Machinen — Hand off a running coding agent between machines

A MicroVM that runs on hardware you already own. Close your laptop and it hands off to another host.

Start work on your laptop. If you need to leave, snapshot the VM and restore it on another machine. The work continues from the same point — same process, same memory, same open files — without starting over.

--

hyperlight: Hyperlight is a lightweight Virtual Machine Manager (VMM) designed to be embedded within applications. It enables safe execution of untrusted code within micro virtual machines with very low latency and minimal overhead, as seen on Harness-Driven Agents: Secure Podcast Pipeline in Hyperlight MicroVM Sandbox | Microsoft Community Hub

Freestyle VMs - The Most Powerful VMs for AI Agents "Full Linux VMs with up to 32 vCPUs, 32 GB RAM, instant startup, and live forking."

InstaVM | Production Sandbox for AI Agents -- "Beyond sandboxes. A cloud built for agents, complete with runtime, storage, networking, secrets, and policy."

Unix Shells -- "Your AI agent needs a machine that's always on. Boot a full Linux VM in seconds — 200+ tools pre-installed, sessions that never die, accessible from anywhere. From $6/mo."

--

How We Built Secure, Scalable Agent Sandbox Infrastructure "From AWS Lambda to Unikraft micro-VMs with a control plane architecture."

See also: exe.dev, smol machines

notes

OpenAI Codex

notes#coding-agents

January 2026