baljanak

# Class for managing multiple servers or anything with start() and stop() methods

class ServerRack(object):

    def __init__(self, servers):
        self.servers = servers

    def start(self):
        started = []
        try:

baljanak

import gevent
from gevent import monkey; monkey.patch_all()
from gevent_zeromq import zmq
import simplejson as json
import threading
import time

def server(context):
    """ server routine """

baljanak

import gevent
from gevent import monkey; monkey.patch_all()
from gevent_zeromq import zmq
import simplejson as json
import threading
import time

def server(context):

    poll = zmq.Poller()

baljanak

import os
import ast
import marshal
import uuid

import meta


def pyc2py(pyc_file, path='/tmp'):

baljanak

1	37	79	164	155	32	87	39	113	15	18	78	175	140	200	4	160	97	191	100	91	20	69	198	196	
2	123	134	10	141	13	12	43	47	3	177	101	179	77	182	117	116	36	103	51	154	162	128	30	
3	48	123	134	109	41	17	159	49	136	16	130	141	29	176	2	190	66	153	157	70	114	65	173	104	194	54	
4	91	171	118	125	158	76	107	18	73	140	42	193	127	100	84	121	60	81	99	80	150	55	1	35	23	93	
5	193	156	102	118	175	39	124	119	19	99	160	75	20	112	37	23	145	135	146	73	35	
6	155	56	52	120	131	160	124	119	14	196	144	25	75	76	166	35	87	26	20	32	23	
7	156	185	178	79	27	52	144	107	78	22	71	26	31	15	56	76	112	39	8	113	93	
8	185	155	171	178	108	64	164	53	140	25	100	133	9	52	191	46	20	150	144	39	62	131	42	119	127	31	7	
9	91	155	8	160	107	132	195	26	20	133	39	76	100	78	122	127	38	156	191	196	115	
10	190	184	154	49	2	182	173	170	161	47	189	101	153	50	30	109	177	148	179	16	163	116	13	90	185	

baljanak

#!/usr/env python


def merge(A, B, C):
    """ O(n)"""

    i = 0
    j = 0

    for k in range(len(A) + len(B)):

baljanak

3
9
8
4
6
10
2
5
7
1

baljanak

#!/usr/env python


def merge_and_count_split_inversions(A, B, C):
    """ O(n)"""

    i = 0
    j = 0

    inv = 0

baljanak

LLM Wiki: Adding a Learning Filter

Karpathy's LLM Wiki nailed something. LLMs are perfect for the bookkeeping that kills knowledge bases. Updating cross-references, flagging contradictions, keeping pages consistent - that's what they're good at. Humans curate, LLMs maintain. The wiki compounds because maintenance cost drops to zero.

I've been running a version of this for months. One thing kept breaking.

What happens at month 6

LLM Wiki has three layers: raw sources (immutable), the wiki (LLM-maintained), and a schema (configuration). The schema tells the LLM how to structure the wiki.

baljanak

LLM Wiki

A pattern for building personal knowledge bases using LLMs.

This is an idea file, it is designed to be copy pasted to your own LLM Agent (e.g. OpenAI Codex, Claude Code, OpenCode / Pi, or etc.). Its goal is to communicate the high level idea, but your agent will build out the specifics in collaboration with you.

The core idea

Most people's experience with LLMs and documents looks like RAG: you upload a collection of files, the LLM retrieves relevant chunks at query time, and generates an answer. This works, but the LLM is rediscovering knowledge from scratch on every question. There's no accumulation. Ask a subtle question that requires synthesizing five documents, and the LLM has to find and piece together the relevant fragments every time. Nothing is built up. NotebookLM, ChatGPT file uploads, and most RAG systems work this way.