karpathy

"""
The most atomic way to train and run inference for a GPT in pure, dependency-free Python.
This file is the complete algorithm.
Everything else is just efficiency.

@karpathy
"""

import os       # os.path.exists
import math     # math.log, math.exp

uxderrick

Web Animation Best Practices & Guidelines

Document Purpose

This is a comprehensive reference guide for creating high-quality web animations. Use this as a knowledge base for implementing animations in web applications. All principles, timing values, and easing functions provided here are production-tested and ready to use.

---

Core Principles

Principle 1: Natural Motion

Seas0

Running multiple tailscaled instances with netfilter integration disabled

---

This guide explains how to spawn multiple instances of tailscaled on a single system using a systemd.service(5) template and, optionally, customized configurations. By setting "netfilterMode": "off" in all configurations (or by manually configuring them using tailscale up --netfilter-mode off), you can connect to multiple Tailnets simultaneously without resorting to SOCKS proxy–based userspace networking, while preserving functionalities like Magic DNS integration with systemd-resolved (i.e. you can simutaneously have direct access to other machines in BOTH tailnets via their hostnames).

DISCLAIMER: This method completely disables Tailscale’s automatic (iptables/nftables) netfilter firewall rule creation and management. As firewall rules from an earlier tailscaled instance would be wiped out by a new one, potentially locking you out.

transitive-bullshit

// Claude Code is a Beta product per Anthropic's Commercial Terms of Service.
// By using Claude Code, you agree that all code acceptance or rejection decisions you make,
// and the associated conversations in context, constitute Feedback under Anthropic's Commercial Terms,
// and may be used to improve Anthropic's products, including training models.
// You are responsible for reviewing any code suggestions before use.

// (c) Anthropic PBC. All rights reserved. Use is subject to Anthropic's Commercial Terms of Service (https://www.anthropic.com/legal/commercial-terms).

// Version: 0.2.9

hensg

Ghostty Keyboard Shortcuts

Default keyboard shortcuts for Ghostty terminal emulator. Platform-specific differences are noted where applicable.

Window Management

Action	Windows/Linux	macOS
New window	Ctrl+Shift+N	Cmd+N
Close window	Alt+F4	Cmd+Shift+W

Nabrok

Notes on migrating from RouterProvider to framework

Official documentation: https://reactrouter.com/upgrading/router-provider

Issues

1. Vite config. If the vite config contains a root property it needs to be removed. If you are running vitest coverage reports you may want to move it to test.root or it will show coverage for the entire project. Any paths relative to root need to be adjusted as well, e.g. if you have publicDir set to ../public it should be set to just public (or removed as that is the default).

2. Vitest. You don't want the reactRouter plugin running for tests, so setup as

hamelsmu

def follow_user_follows(client, target_user):
    "Follow everyone the target_user is following."
    
    cursor = None
    total_followed = 0

    while True:
        # Step 1: Fetch a batch of accounts the target user is following
        # https://docs.bsky.app/docs/api/app-bsky-graph-get-follows
        response = client.app.bsky.graph.get_follows({

ianmcook

{
  "statement": "SELECT * FROM MYTABLE",
  "resultSetMetaData": {
    "format": "arrowv1"
  },
  "timeout": 60,
  "database": "MYDATABASE",
  "schema": "MYSCHEMA",
  "warehouse": "MYWAREHOUSE",
  "role": "MYROLE"

z3z1ma

#!/usr/bin/env python
"""A Language Server Protocol (LSP) server for SQL with SQLMesh integration."""

import asyncio
import gc
import io
import logging
import re
import typing as t
import weakref

Artefact2

Which GGUF is right for me? (Opinionated)

Good question! I am collecting human data on how quantization affects outputs. See here for more information: ggml-org/llama.cpp#5962

In the meantime, use the largest that fully fits in your GPU. If you can comfortably fit Q4_K_S, try using a model with more parameters.

llama.cpp feature matrix

See the wiki upstream: https://github.com/ggerganov/llama.cpp/wiki/Feature-matrix