Work for long-lived agentics.
my-agentic.com is the public home for Tony Wood's work on long-lived agentic systems that stay useful, bounded, and reviewable.
Judgment layer research, applied to practical client environments.
What this is
This is focused work on long-lived agent behavior for real operating environments.
The goal is to build agents that can:
  • make better judgment calls over time
  • respect human boundaries and disclosure constraints
  • retain useful memory without becoming unreadable
  • stay inspectable and corrigible under real use
Why it matters
Long-running agents often fail from drift, over-sharing, and weak supervision — not from lack of output.
This work focuses on reducing those risks through clear review loops, bounded behavior, and better judgment over time.
What this includes
This work includes:
long-lived runtime patterns
judgment and review loops
memory and reflection practices
human-in-the-loop oversight
How this links to SNAXK
SNAXK informs the judgment layer research in this work.
It explores whether heuristic-first, review-heavy control loops can help agents become safer, more legible, and more trustworthy in real operation.
This remains active research, not broad general availability.
Who I am
I am Tony Wood.
I build practical agent systems that combine engineering, applied research, and human oversight.
My focus is not just output quality, but judgment quality over time.
Current status
  • active build and research
  • selected parts are live in controlled environments
  • broader rollout is staged
For clients
If you are evaluating long-running agent systems, I offer focused advisory and implementation collaboration.
Typical conversations include:
  • judgment and risk design for agent workflows
  • review and oversight architecture
  • practical rollout planning for controlled environments
Contact
For collaboration inquiries: