Where we are, what we're proving, and where the project may go next.
SRR — SupRrrAi ToolKit is under active development.
This roadmap describes the direction of the project, not guaranteed release dates or product commitments.
Development may change as testing reveals new requirements, limitations, or opportunities.
STAGE 1 — BUILD & VERIFY
CURRENT STAGE
STATUS:
IN DEVELOPMENT
The first stage of SRR is about building the core agent infrastructure and proving that it can reliably interact with real systems.
The central question is:
Can an AI agent take an action and provide enough evidence to prove that the intended result actually occurred?
Stage 1 focuses on the foundation.
CORE AREAS
AI AGENT HARNESS
Build the infrastructure surrounding the model rather than relying on model intelligence alone.
The harness is responsible for connecting reasoning to controlled execution.
LINUX-FIRST EXECUTION
Develop SRR around Linux as a primary operating environment.
Areas being tested include:
terminal execution
POSIX shell operations
Bash workflows
files
processes
applications
local services
system tools
LOCAL AI
Test SRR with locally hosted AI models where practical.
Goals include:
model flexibility
local execution
reduced dependence on paid APIs
user-controlled infrastructure
CONTROLLED TOOL EXECUTION
Provide models with bounded tools rather than unrestricted computer control.
The system should understand:
what tool is being requested
what action is permitted
what action actually occurred
what evidence was produced
EVIDENCE COLLECTION
SRR should not rely solely on a model's description of what happened.
The system is being developed to collect evidence from the environment itself.
Examples may include:
file state
terminal output
application state
saved artifacts
system responses
structured results
VERIFICATION
Evidence should be evaluated against the requested result.
The basic idea:
MODEL:
I completed the task.
SRR:
Prove it.
Possible outcomes include:
VERIFIED
FAILED
BLOCKED
INCOMPLETE
NEEDS REPAIR
PERSISTENT STATE
AI workflows often require state to survive across multiple actions.
Current testing includes:
persistent shell state
working directories
environment variables
task state
handoff state
multi-step workflows
FAILURE DETECTION
Failures should become visible rather than silently passing.
SRR development currently emphasizes:
detecting failed actions
identifying missing evidence
finding incomplete outcomes
exposing incorrect assumptions
preventing false completion claims
REPAIR & RETEST
When failures occur:
identify the cause
repair the implementation
rerun the test
collect fresh evidence
verify again
This cycle is central to SRR development.
REAL-WORLD TESTING
Stage 1 includes practical tests involving areas such as:
terminal workflows
file handling
browser operations
application automation
email workflows
local AI
multi-agent communication
tool invocation
persistent shell execution
evidence verification
STAGE 1 SUCCESS CONDITION
Stage 1 is not complete simply because an AI can perform actions.
The goal is to establish a repeatable foundation where SRR can:
understand a requested action
route it to an appropriate tool
execute within defined boundaries
collect evidence
inspect the result
verify the resulting state
report failure when verification is not possible
STAGE 2 — EXPAND CAPABILITIES
PLANNED DIRECTION
Potential areas:
more tools
additional Linux workflows
browser automation
application integrations
more local-model support
multi-agent workflows
structured APIs
document workflows
communication tools
improved performance
expanded verification methods
Stage 2 scope will be determined by what is learned during Stage 1.
STAGE 3 — PUBLIC ACCESS PLANNING
FUTURE DECISION
Only after sufficient technical testing will decisions be made about how SRR may be distributed.
Possible questions include:
Will there be a public release?
Will there be a private release?
Will there be a GitHub repository?
Will some components become open source?
Will there be a hosted version?
Will there be a subscription model?
Will there be developer access?
No decisions are being promised yet.
STAGE 4 — RELEASE PATH
FUTURE
If and when SRR reaches an appropriate level of maturity, a release path can be defined based on:
technical readiness
testing evidence
hardware requirements
security
support requirements
community interest
development resources
distribution strategy
WHERE SRR IS TODAY
CURRENT STAGE:
STAGE 1 — BUILD & VERIFY
The focus remains:
Build it.
Test it.
Break it.
Understand why.
Repair it.
Run it again.
Collect evidence.
Verify the result.
FOLLOW DEVELOPMENT
Website:
Blog:
YouTube:
https://www.youtube.com/@TheSp_cemanChannel
SRR Waitlist:
IMPORTANT
This roadmap represents development direction.
It is not a promise of:
specific features
release dates
pricing
public GitHub availability
open-source licensing
beta access
commercial availability
Features, architecture, and priorities may change as SRR continues to be tested.