Build. Test. Break. Repair. Verify.
A development newsletter from SRR / SupRrrAi ToolKit
By Issiah Hill
This issue
SRR is being built around one simple idea:
An AI saying, “I did it,” is not the same as proving the task was actually completed.
SRR — SupRrrAi ToolKit is a Linux-first, local-first AI agent harness being developed around controlled execution, evidence, verification, and modular AI tooling.
This newsletter is where I share what I am actually building, what worked, what failed, what changed, and where the project goes next.
No hype.
No fake launch promises.
Just the work.
What I built
This section covers the latest additions or improvements to SRR.
Examples:
- terminal execution
- persistent shell state
- evidence verification
- tool adapters
- local model integration
- multi-agent communication
- application automation
- browser workflows
- file handling
- email workflows
- safety and execution controls
What I tested
Every SRR development cycle includes real testing.
I use tests to answer questions such as:
Did the agent actually perform the requested action?
Did the resulting state match the intended result?
Can SRR prove what happened?
Can a failed action be detected instead of silently accepted?
Can the same task be repeated reliably?
What failed
Failures are part of the development process.
When something breaks, I do not want SRR to hide it.
I want to know:
What failed?
Why did it fail?
What evidence exposed the problem?
Can the architecture be improved so the failure does not repeat?
Some of the most important improvements in SRR have come directly from failed tests.
What changed
This section explains what was repaired or redesigned after testing.
The development loop is:
BUILD IT
TEST IT
BREAK IT
UNDERSTAND WHY
REPAIR IT
RUN IT AGAIN
COLLECT EVIDENCE
VERIFY THE RESULT
What I learned
AI models are becoming increasingly capable.
But model intelligence alone does not create a reliable AI agent.
The surrounding system matters:
tools
permissions
execution
state
evidence
verification
failure handling
The better the model becomes, the more important that infrastructure may become.
What comes next
SRR is still being developed.
There is currently no promised:
release date
pricing model
public GitHub release
open-source commitment
beta date
commercial launch date
The goal right now is to continue building, testing, verifying, and understanding what SRR can actually become.
Watch
Latest videos and SRR testing:
Read
SRR development articles and project updates:
Join the SRR waitlist
Want to follow SRR as it develops?
Join the SRR Waitlist:
The waitlist does not guarantee access, pricing, beta participation, GitHub access, or a release date.
It simply means you want to hear what happens next.
Until next time
Build it.
Test it.
Prove it.
Issiah Hill
Creator — SRR / SupRrrAi ToolKit