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.

Australia ,Queensland Joined on May 16, 2026 https://sovereignhive.com.au/

Education

Self Taught

Work

Sovereign Hive

MxGuru

Jun 14

I killed six of my own results in one night. That was the win

#machinelearning #ai #security #beginners

2 min read

Want to connect with MxGuru?

Create an account to connect with MxGuru. You can also sign in below to proceed if you already have an account.

Create Account

Already have an account? Sign in

MxGuru

Jun 13

I built a detector that hit 100% accuracy. Then I spent a day trying to prove it wrong

#machinelearning #python #ai #beginners

2 min read

MxGuru

Jun 9

Security AI shouldn't only belong to the giants.

#ai #llm #cybersecurity #sovereignty

1 min read

GitHub “Finish-Up-A-Thon” Challenge Submission

MxGuru

Jun 5

Reviving the Master Chief Protocol: Building an Auto-Healing Adversarial Swarm

#devchallenge #githubchallenge #githubcopilot #python

5 min read

MxGuru

Jun 5

Your What Keeps Me Going!

#mentalhealth #suicideprevention #freementalhealthcare #madtechcyberquantresearch

1 min read

MxGuru

May 20

Local-First AI: Why Your Threat Intel Shouldn't Live on Someone Else's Server

#security #ai #selfhosted #privacy

2 min read

MxGuru

May 20

Building a Self-Healing Kill Switch for AI Infrastructure

#ai #sre #python

1 min read

MxGuru

May 20

I Built a 127-Tool MCP Server From Scratch — Here's What I Learned

#mcp #ai #python #tooling

3 min read

MxGuru

May 20

The Best Result This Week Was a Failed Prediction — Phase-3a Doesn't Transfer

#quantization #hsaq #methodology #granite

1 min read

MxGuru

May 20

Two Localizers, Both Wrong: Bounding a Quantization Cost That Wouldn't Close

#quantization #hsaq #methodology #granite

1 min read

MxGuru

May 20

When the Sensitivity Metric Lies: A Drift-Inversion Smoking Gun in Mixed-Precision LLM Quantization

#quantization #hsaq #awq #granite

8 min read

Hermes Agent Challenge Submission

MxGuru

May 19

Two Local-Agent Philosophies: Where Hermes Earns Its Design, and Where the Tradeoffs Invert

#hermesagentchallenge #devchallenge #agents

7 min read

MxGuru

May 19

The Hive Are Evolving!

#ai #productivity #python #automation

1 min read

Gemma 4 Challenge: Write about Gemma 4 Submission

MxGuru

May 18

What Gemma 4 Actually Unlocks for a Local Security Swarm (And Why I Don't Use the Same Variant Everywhere)

#devchallenge #gemmachallenge #gemma

5 min read

MxGuru

May 17

Swarm-Consensus Defense Achieves 98.2% Against Cloud-LLM Adversarial Attacks

#ai #security #llm #cybersecurity

1 min read

MxGuru

May 17

Sovereign Hive v6.6 — 98% defense across 200 adversarial rounds on a single 5070

#swarm #security #ai

1 min read

MxGuru

May 16

99%% Defense Rate Across 500 Rounds: A Self-Healing Swarm on a $550 GPU

#security #ai #machinelearning #cybersecurity

3 min read

MxGuru

May 16

From 53%% to 90%%: How an Auto-Healing AI Swarm Learned to Defend Itself

#security #ai #machinelearning #cybersecurity

3 min read

DEV Community

MxGuru

Badges

Gemma 4 Challenge Completion

Hermes Agent Challenge Winner

1 Week Community Wellness Streak

Writing Debut

GitHub Repositories

Skills/Languages

Currently learning

Currently hacking on

Available for

I killed six of my own results in one night. That was the win

Want to connect with MxGuru?

I built a detector that hit 100% accuracy. Then I spent a day trying to prove it wrong

Security AI shouldn't only belong to the giants.

Reviving the Master Chief Protocol: Building an Auto-Healing Adversarial Swarm

Your What Keeps Me Going!

Local-First AI: Why Your Threat Intel Shouldn't Live on Someone Else's Server

Building a Self-Healing Kill Switch for AI Infrastructure

I Built a 127-Tool MCP Server From Scratch — Here's What I Learned

The Best Result This Week Was a Failed Prediction — Phase-3a Doesn't Transfer

Two Localizers, Both Wrong: Bounding a Quantization Cost That Wouldn't Close

When the Sensitivity Metric Lies: A Drift-Inversion Smoking Gun in Mixed-Precision LLM Quantization

Two Local-Agent Philosophies: Where Hermes Earns Its Design, and Where the Tradeoffs Invert

The Hive Are Evolving!

What Gemma 4 Actually Unlocks for a Local Security Swarm (And Why I Don't Use the Same Variant Everywhere)

Swarm-Consensus Defense Achieves 98.2% Against Cloud-LLM Adversarial Attacks

Sovereign Hive v6.6 — 98% defense across 200 adversarial rounds on a single 5070

99%% Defense Rate Across 500 Rounds: A Self-Healing Swarm on a $550 GPU

From 53%% to 90%%: How an Auto-Healing AI Swarm Learned to Defend Itself