Frank Brsrk

Chilling with my dogs and keyboard fighting the AIs

Joined on Apr 18, 2026

Frank Brsrk

May 24

I open-sourced a 4-agent blood-panel triage workflow on heym, with a deterministic Python safety gate that runs BEFORE any LLM token

#ai #llm #opensource #mcp

5 min read

Want to connect with Frank Brsrk ?

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

Create Account

Already have an account? Sign in

Frank Brsrk

May 23

Reasoning happens before the response

#ai #mcp #llm #agents

5 min read

Frank Brsrk

May 22

An open source LLM eval tool with two independent quality signals

#showdev #ai #llm #opensource

4 min read

Frank Brsrk

May 21

I built a reasoning harness for LLM agents. Here's what an agent receives when it calls it.

#llm #mcp #ai #agents

4 min read

Frank Brsrk

May 18

Cognitive middleware for n8n agents: four ways to wire Ejentum in

#n8n #ai #agents #automation

5 min read

Frank Brsrk

May 14

Why your LLM agent drifts off-task by step 4 (and why prompts can't fix it)

#ai #agents #llm #reasoning

3 min read

Frank Brsrk

May 10

I open-sourced a 3-agent blind eval team. Any agent runtime can call it for pre-commitment review of its own plans.

#ai #agents #claude #tooling

10 min read

Frank Brsrk

May 7

I open-sourced a 4-agent adversarial code review team. Any coding agent can call it as an MCP server. Built in heym.

#ai #mcp #agents #opensource

6 min read

Frank Brsrk

May 6

I shipped ejentum-mcp today: four cognitive harnesses as MCP tools

#mcp #claude #ai #agents

3 min read

Frank Brsrk

May 4

How to diagnose where your RAG agent fabricates: an open-source A/B eval workflow with cross-lab blind judges

#ai #productivity #n8nbrightdatachallenge #agents

6 min read

Frank Brsrk

Apr 25

Why LLM Agents Fail: Four Mechanisms of Cognitive Decay and the Reasoning Harness Layer

#ai #llm #agents #architecture

13 min read

Frank Brsrk

Apr 25

Why Your AI Agent Loses the Plot: Reasoning Decay and Attention Loss in Long-Running Tasks

#ai #llm #agents #programming

10 min read

Frank Brsrk

Apr 24

Trippy Balls

#llm #ai #productivity #coding

1 min read

Frank Brsrk

Apr 24

I built a multi-turn agent-vs-agent blind eval in n8n

#beginners #opensource #n8n #ai

6 min read

Frank Brsrk

Apr 23

I built a Python module to A/B test prompts inside Claude Code, and you can run it on yours

#ai #python #agents #llm

6 min read

Frank Brsrk

Apr 22

the model alone is not the agent. The harness plus the model is the agent.

#agents #ai #agentskills

2 min read

Frank Brsrk

Apr 22

Eval workflow for agentic builders: fork any prompt through baseline vs scaffolded agents, blind third-party judge.

#ai #agents #agentskills

2 min read

Frank Brsrk

Apr 22

Wait, you guys run evals?

#ai #evals #llm

1 min read

Frank Brsrk

Apr 20

Under Pressure. Better Harness.

#agents #agentskills

2 min read

DEV Community

Frank Brsrk

Badges

2 Week Community Wellness Streak

1 Week Community Wellness Streak

Writing Debut

I open-sourced a 4-agent blood-panel triage workflow on heym, with a deterministic Python safety gate that runs BEFORE any LLM token

Want to connect with Frank Brsrk ?

Reasoning happens before the response

An open source LLM eval tool with two independent quality signals

I built a reasoning harness for LLM agents. Here's what an agent receives when it calls it.

Cognitive middleware for n8n agents: four ways to wire Ejentum in

Why your LLM agent drifts off-task by step 4 (and why prompts can't fix it)

I open-sourced a 3-agent blind eval team. Any agent runtime can call it for pre-commitment review of its own plans.

I open-sourced a 4-agent adversarial code review team. Any coding agent can call it as an MCP server. Built in heym.

I shipped ejentum-mcp today: four cognitive harnesses as MCP tools

How to diagnose where your RAG agent fabricates: an open-source A/B eval workflow with cross-lab blind judges

Why LLM Agents Fail: Four Mechanisms of Cognitive Decay and the Reasoning Harness Layer

Why Your AI Agent Loses the Plot: Reasoning Decay and Attention Loss in Long-Running Tasks

Trippy Balls

I built a multi-turn agent-vs-agent blind eval in n8n

I built a Python module to A/B test prompts inside Claude Code, and you can run it on yours

the model alone is not the agent. The harness plus the model is the agent.

Eval workflow for agentic builders: fork any prompt through baseline vs scaffolded agents, blind third-party judge.

Wait, you guys run evals?

Under Pressure. Better Harness.