Typical starting point
AI pilots show promise but have no path to production; delivery is slow or unpredictable; review and testing cannot keep pace with faster implementation; a few people carry too much of the context.
Built in public Meshloop is Samuel's open-source runtime for running coding agents in parallel with isolated worktrees, compiler verification, and a human accept gate. Explore Meshloop
Focus area · AI in production
Get AI out of pilot and into daily operations.
Generating code or answers is the easy part. Enterprise AI maturity begins when AI becomes part of workflow design, with evaluation, controls, human approval, and named owners. In regulated environments, those guardrails are architecture rather than compliance paperwork.
What the work covers
The path from idea to release and from pilot to operation: architecture, integration, evaluation, automated testing, and clear responsibility for decisions. AI is one option among several, chosen where it improves the workflow.
How progress is judged
Measures that reflect the actual problem, such as waiting time, rework, review effort, defects, evaluation results, and operating cost.
AI in software delivery
Faster implementation makes the delivery system around it more important, not less. The work defines intent, acceptance criteria, review discipline, and evidence so that AI-assisted changes can be trusted and traced to an accountable owner.
Applied AI in operations
Each use case starts from a specific problem and a comparison of options. The work covers data and system integration, quality checks against real examples, operating costs, and how people will supervise, operate, and improve the solution.
Evidence
Built systems & related thinking.
Open-source implementations and articles on role-based delivery, review discipline, and explicit human judgement in AI-assisted engineering.
AgentFlow SDLC
A system for making responsibilities, review, and decisions explicit in AI-assisted software delivery. Separates implementation, gates, and verification into accountable human roles.
Agents Made the SDLC Conscious Again
The faster implementation becomes, the more deliberate the delivery system around it must be. Intent, evidence, review, and human authority in practice.
Start a conversation
Describe the AI or delivery decision in front of you, what has been tried, and what is at stake.