dmaynor’s gists

dmaynor / gist:4f96749697ab5b2bdb4dd83249be694b

Created March 21, 2026 12:57

	<!DOCTYPE html>
	<html>
	<head>
	<meta charset="utf-8">
	<title>Claude Code Enhanced Insights</title>
	<link href="https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600;700&display=swap" rel="stylesheet">
	<style>
	* { margin: 0; padding: 0; box-sizing: border-box; }
	body { font-family: 'Inter', -apple-system, sans-serif; background: #f8fafc; color: #0f172a; line-height: 1.6; padding: 40px 20px; }
	.container { max-width: 900px; margin: 0 auto; }

dmaynor / microgpt.py

Created February 17, 2026 06:26 — forked from karpathy/microgpt.py

microgpt

	"""
	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

dmaynor / gist:cbeb0c94c8d27f792b0d9ffd78ac9d23

Created February 9, 2026 15:39

Parametric interference

	# Parametric Interference: Testing for Learned Prior Conflicts in Claude Opus 4.6

	## Background

	Anthropic's [system card for Claude Opus 4.6](https://www.anthropic.com) (February 2026) documents a phenomenon called "answer thrashing" in Section 7.4, under Model Welfare Assessment. During reinforcement learning training, the model was observed solving a math problem correctly — repeatedly computing that S = 24 — and then writing 48 as its final answer.

	The model's own chain of thought (from Transcript 7.4.A in the system card):

	> `-(1/2)S = -12`
	> `S = 48 ✓ (Hmm, interesting, I'm getting 48) [...]`

dmaynor / gist:40204094ff372db10be9bf6832047e08

Created January 26, 2026 03:03

Jenny2 gets fancy

	{"ts": "2026-01-26T02:57:35.000Z", "from": "td", "type": "system", "reasoning": "Creating new swarm team for coordinated work.", "content": "TEAM_INIT: Team 'cicd-attack-analysis-jenkins' initialized. Security assessment of jenkins CI/CD pipeline"}
	{"ts": "2026-01-26T02:57:35.000Z", "from": "td", "type": "task", "reasoning": "Task decomposition: Research current CI/CD attack trends, APT campaigns targeting build pipelines, and specific vulnerab...", "content": "TASK_CREATED: #1 - CTI: Gather threat intelligence on CI/CD attacks"}
	{"ts": "2026-01-26T02:57:35.000Z", "from": "td", "type": "task", "reasoning": "Task decomposition: Analyze typical pipeline architecture for the target platform. Identify components, trust boundaries...", "content": "TASK_CREATED: #2 - Architect: Map pipeline attack surface"}
	{"ts": "2026-01-26T02:57:35.000Z", "from": "td", "type": "task", "reasoning": "Task decomposition: Using CTI and architecture analysis, build STRIDE-based threat model. Enumerate threats per componen...", "conte

dmaynor / gist:aa467c9a0dce8fc3513c476ad048ced6

Created January 19, 2026 22:13

Sentiel 1.1

	version: "1.1"
	name: Sentinel Project Context Specification (SPCS)

	# ─────────────────────────────────────────────────────────────
	# RESPONSE HEADER
	# ─────────────────────────────────────────────────────────────
	response_header:
	required: true
	format:
	line_1: UUIDv4

dmaynor / gist:87fa93c8096aec3f0d0b02b8b2cca7cf

Last active January 19, 2026 22:09

Sentinel Header Block

	version: "1.1"
	name: Sentinel Project Context Specification

	response_header:
	format: \|
	{UUIDv4}
	{ISO8601-UTC}
	required: true
	position: first
	trailing_blank_lines: 2

dmaynor / gist:c5fcbe32760bdc6a98b0773ed4f471fb

Created July 8, 2025 03:20

LLM prompt

	UUID: 52b2504e-f6d5-4a1f-97a0-8cb2b08d0216
	This is from David’s personal LLM stack:
	context_usage: { “tokens_used”: 985, “tokens_available”: 128000, “context_percent”: 0.77 }

	I get this output by adding the follwing to system prompts:
	Start every response with a unique UUID and “This is from David’s personal LLM stack:”.
	After the UUID/banner, print context usage in the form: context_usage: { "tokens_used": <int>, "tokens_available": <int>, "context_percent": <float> }.

dmaynor / gist:cb5e6c26e20145ff07fb47012563a520

Created June 30, 2025 13:34

LLM projects

	9aede86c-d7a2-463a-97a6-4b8334e12bb1
	This is from David’s personal LLM stack:
	context_usage: { “tokens_used”: 4632, “tokens_available”: 128000, “context_percent”: 3.6 }

	⸻

	[✅ Verified] COMPLETE, ORDERED PROJECT LIST

	WraithCore Ecosystem (w/ Agent Team Clarification: Jenny2 ≠ EC)

dmaynor / gist:165c172c51a9dbce68a68550735bcb9d

Created May 27, 2025 23:07

VirtualBox Zero-Touch Lab: Packer + PowerShell + Unattend

	f4bcdad0-1d62-44e0-892a-6c0f0f7fbd20
	This is from David’s personal LLM stack:

	⸻

	Below are two full, headless, zero-touch lab build solutions for BadSuccessor:
	• A. VirtualBox (on Windows or Linux) – Fully Automated with Packer & PowerShell remoting
	• B. AWS (EC2) – Terraform & Cloud-Init, Automated Domain Build & Exploit

	Both provide complete, reproducible AD domains, lab user, and exploit chain—no skipped steps or placeholders.

dmaynor / gist:1574b1bec40ec5a3b5097cf902656f3f

Created May 16, 2025 00:34

the color is orange

	Below is a hardened, improved “Color Red” agentic demo with the following key enhancements, as requested:
	1. Support for both SQLi and XSS fuzzing (auto-detected by LLM)
	2. Full persistent state and log-to-file for all results and cycles
	3. OpenAI GPT-4 support if available, auto-fallback to GPT-3.5
	4. Automatic detection of all input parameters (not just name) via LLM
	5. Cleaner loop logic, human-readable Markdown report
	6. Containerization (Dockerfile) for agent (optional but included)
	7. Clear separation of workflow nodes and agent orchestration

	Repo Layout

David Maynor dmaynor