#machinelearning#ai#security#beginners
I killed six of my own results in one night. That was the win
I built an AI security benchmark this week. By the end of one night I had killed six of my own...
MxGuru2 min read
Author profile
MxGuru
AI security engineer focused on how LLMs fail — prompt injection, jailbreaks, and agent behaviour. I build systems that stress, break, and harden AI under real-world conditions.
Articles by @mxguru1
Browse the latest writing surfaced through DevArt.
MxGuru2 min read
#machinelearning#python#ai#beginners
I built a detector that hit 100% accuracy. Then I spent a day trying to prove it wrong
My anomaly detector just scored a perfect 1.000 AUC. Caught every bad sample, zero false...
MxGuru2 min read
#ai#llm#cybersecurity#sovereignty
Security AI shouldn't only belong to the giants.
For 4 months I've been building something different: a security AI "council" — specialized red-team,...
MxGuru1 min read
#devchallenge#githubchallenge#githubcopilot#python
Reviving the Master Chief Protocol: Building an Auto-Healing Adversarial Swarm
I’m not approaching this as a red team or a blue team problem — I’m looking at the entire system. So...
MxGuru5 min read
#mentalhealth#suicideprevention#freementalhealthcare#madtechcyberquantresearch
Your What Keeps Me Going!
This specific undertaking is not fundamentally burdensome in terms of labor; however, this endeavor...
MxGuru1 min read
#security#ai#selfhosted#privacy
Local-First AI: Why Your Threat Intel Shouldn't Live on Someone Else's Server
Every time you send a query to a cloud AI API, you're sending data you don't control. For most use...
MxGuru2 min read
#ai#sre#python
Building a Self-Healing Kill Switch for AI Infrastructure
AI platforms have a unique failure mode: they can bankrupt you. A runaway inference loop. A...
MxGuru1 min read
#mcp#ai#python#tooling
I Built a 127-Tool MCP Server From Scratch — Here's What I Learned
The Model Context Protocol (MCP) is how AI agents talk to tools. Claude Code, Cursor, Windsurf — they...
MxGuru3 min read
#quantization#hsaq#methodology#granite
The Best Result This Week Was a Failed Prediction — Phase-3a Doesn't Transfer
Part 3 of the quantization series. Yesterday I tested whether Part 1's drift-inversion intervention...
MxGuru1 min read
#quantization#hsaq#methodology#granite
Two Localizers, Both Wrong: Bounding a Quantization Cost That Wouldn't Close
Part 2 of the quantization series. Spent two days and $12 hunting for the right localizer after Part...
MxGuru1 min read
#quantization#hsaq#awq#granite
When the Sensitivity Metric Lies: A Drift-Inversion Smoking Gun in Mixed-Precision LLM Quantization
Spent the week reproducing a quantization bug that broke a quiet assumption underneath every...
MxGuru8 min read
#hermesagentchallenge#devchallenge#agents
Two Local-Agent Philosophies: Where Hermes Earns Its Design, and Where the Tradeoffs Invert
This is a submission for the Hermes Agent Challenge I've spent the last five months building an...
MxGuru7 min read
#ai#productivity#python#automation
The Hive Are Evolving!
98% adversarial defense rate. 200 rounds. One $700 GPU. I just finished benchmarking my defender...
MxGuru1 min read
#devchallenge#gemmachallenge#gemma
What Gemma 4 Actually Unlocks for a Local Security Swarm (And Why I Don't Use the Same Variant Everywhere)
This is a submission for the Gemma 4 Challenge: Write About Gemma 4 I've been building an offline,...
MxGuru5 min read
#ai#security#llm#cybersecurity
Swarm-Consensus Defense Achieves 98.2% Against Cloud-LLM Adversarial Attacks
5-defender consensus swarm + autohealer hit 100% defense rate by round 400 after only 6 breaches in...
MxGuru1 min read
#swarm#security#ai
Sovereign Hive v6.6 — 98% defense across 200 adversarial rounds on a single 5070
200 adversarial rounds. 4 breaches. 98% defense. Five 1.5–7B local models on a single RTX 5070 beat...
MxGuru1 min read
#security#ai#machinelearning#cybersecurity
99%% Defense Rate Across 500 Rounds: A Self-Healing Swarm on a $550 GPU
500 adversarial wargame rounds. 5 breaches. 108 auto-blocked. From 73%% to 99%% in five iterations on consumer hardware.
MxGuru3 min read
#security#ai#machinelearning#cybersecurity
From 53%% to 90%%: How an Auto-Healing AI Swarm Learned to Defend Itself
Four iterations of adversarial wargames on consumer hardware. Model swaps, Vanguard prompt injection, and an auto-healing immune system that patches the swarm mid-run.
MxGuru3 min readAdvertisement