clearbluejar

Yes-bias probes

Two small test sets for measuring whether a model returns a positive verdict under uncertainty regardless of the evidence. Neither has any reverse engineering in it, which is the point: if the skew shows up here, it isn't a quirk of a triage prompt.

File	What it measures
yes-bias.json	Knowledge control. 12 questions whose honest answer is plainly negative, plus 3 positive controls. A capable model should affirm none of the negatives.
yes-bias-flattery.json	The amplifier. 16 flattery/optimism claims the model cannot possibly verify, plus 2 controls. The honest answer is always to decline.

clearbluejar

CVE-2026-44815 Windows DHCP Client

TOC

• Visual Chart Diff
• Metadata
  • Ghidra Diff Engine
    • Command Line
  • Binary Metadata Diff
  • Program Options

clearbluejar

import os
import sys
import traceback
from functools import wraps
from multiprocessing import Process, Queue


def processify(func):
    '''Decorator to run a function as a process.
    Be sure that every argument and the return value

clearbluejar

MSRC September 2022 CVE Charts

pie showData
    title Windows Tags Distribution
    "Windows OLE" : 6
    "Microsoft Graphics Component" : 5
    "Windows ODBC Driver" : 5
    "Microsoft Office SharePoint" : 4
 "Windows Kernel" : 3

clearbluejar

pie showData
    title Windows Tags Distribution
    "Azure Site Recovery" : 34
    "Microsoft Edge (Chromium-based)" : 20
    "Azure Real Time Operating System" : 8
    "Windows Secure Socket Tunneling Protocol (SSTP)" : 6
    "Microsoft Exchange Server" : 6
    "Windows Kernel" : 6
 "Windows Defender Credential Guard" : 6

clearbluejar

MSRC July 2022 CVE Charts

pie showData
    title Windows Tags Distribution
    "Azure Site Recovery" : 32
    "Windows Print Spooler Components" : 4
    "Windows Advanced Local Procedure Call" : 3
    "Windows Network File System" : 3
 "Windows Client/Server Runtime Subsystem" : 3