Who decides, and who builds
Keep engineering judgment and release sign-off human-owned while running parallel coding agents in isolated sandboxes.
Move the Needle · Samuel Mota
Open-source systems
Samuel builds the AI-delivery and observability systems he writes about, as open source, so the ideas and the choices behind them are available to inspect.
How the ecosystem fits together
Six independent systems. One coherent philosophy.
Each project solves one human problem end to end—with zero mandatory coupling—and connects through open contracts when combined across your engineering lifecycle or personal workflow.
01 · Plan & GovernAgentFlow SDLC
02 · Build in ParallelMeshloop
03 · Watch Liveagent-otel-bridge
04 · Learn & OptimizeDendro
What is happening now, and where time went
See live agent behavior inside your existing monitoring stack, then read the growth rings of completed sessions to find rework and stalls.
Who you are, and what you decided
Carry your background and voice across every AI tool with a local context passport, and keep career decisions in a record you can trust.
New
Dendro
See where time, handoffs, rework, capabilities, and token budget actually go across your multi-agent AI coding sessions—using verifiable local DuckDB evidence instead of vanity productivity claims.
- 4-Layer Session Dial & Causal X-Ray
- 5-Layer Capability Studio (FUNNEL I06)
- 25 DuckDB Insights & 96.9% Cache Yield
Why it matters & how it works
- Why care
- Six open terminals look busy even when four agents are stalled on a handoff and two are looping. Dendro turns raw local session history into a 4-layer causal X-ray of where time, handoffs, and budget actually go.
- How it works
- Reconstructs local session logs out-of-band across Claude Code, Codex, AGY, Grok CLI, and Pi (plus optional multi-agent orchestration adapters) into 4 architectural layers, 5 capability funnels, and 25 deterministic DuckDB insights.
New
agent-otel-bridge
See what your AI coding agents are doing in real time—adoption, tool behavior, and rate-limit headroom—inside the monitoring stack you already run.
- Zero-Lag Live Telemetry
- Intent & Risk Classification
- Multi-Agent Trace Links
Why it matters & how it works
- Why care
- AI agents are now part of software delivery, so they deserve the same live visibility as production services—inside Grafana, Datadog, or Honeycomb rather than another siloed vendor console.
- How it works
- Captures live OpenTelemetry traces and metrics across Claude Code, Codex, and Gemini CLI with zero developer slowdown, linking parent and delegated sub-agents into one timeline.
Better together with Dendro →
New
Meshloop
Run multiple coding agents in parallel while keeping your codebase and main branch completely clean: isolated sandboxes, compiler-verified self-repair, and a human merge gate.
- Isolated Sandbox Worktrees
- Compiler-Verified Repair
- Human Merge Gate
Why it matters & how it works
- Why care
- Running multiple agents on one repo fails without physical boundaries. Meshloop isolates parallel work until it compiles, passes tests, and earns human acceptance.
- How it works
- Gives every parallel agent task its own isolated Git sandbox, verifies every change against real compiler and test exit codes, and merges only after explicit human approval.
Better together with AgentFlow SDLC →
New
Nextstep
Let AI accelerate your research and drafting while keeping every career decision, application, and outreach thread in a record you can trust.
- Auditable Career Ledger
- Portfolio Decision Thesis
- Structured Outreach Playbooks
Why it matters & how it works
- Why care
- Agents are great at research and drafting but terrible at remembering what you decided three weeks ago. Nextstep keeps your career search coherent across months and across AI tools.
- How it works
- Separates AI research and drafting from the factual record of your search—keeping target companies, pursue-or-decline decisions, tailored submissions, and outreach playbooks in an auditable local ledger.
Better together with Holoself →
AgentFlow SDLC
Let AI write the code while keeping engineering judgment, risk boundaries, and release acceptance explicitly human-owned.
- Separation of Powers
- Proportional Risk Gates
- Compounding Team Rules
Why it matters & how it works
- Why care
- When code generation becomes cheap, decision quality and clear ownership become the bottleneck. AgentFlow keeps roles, risk boundaries, and release sign-off in your team's hands.
- How it works
- Separates steering, building, and independent acceptance across risk-calibrated slices, turning recurring review friction into human-approved engineering rules.
Better together with Meshloop →
Holoself
Stop re-explaining who you are, how you write, and what you value every time you open a new AI tool—without handing your identity to a cloud vendor vault.
- Local Markdown Ownership
- Task-Specific Lenses
- Reviewable Memory Proposals
Why it matters & how it works
- Why care
- Useful AI depends on the right personal context, not an opaque cloud memory profile. Holoself keeps your identity local, portable across every AI tool, and under your explicit approval.
- How it works
- Stores your background, voice, and working principles in plain local files, shares only the slice each project needs through task-specific lenses, and turns new learnings into reviewable proposals.
Better together with Nextstep →
MetaSkills
Turn hard-won agent practices into portable, installable skills, and verify whether the guardrails around a coding agent can be trusted.
- Portable Agent Skills
- Read-Only Harness Audits
- Cross-Tool Adapters
Why it matters & how it works
- Why care
- Agent reliability comes from the harness around the model. Keeping practices as readable files in Git makes them inspectable, portable across AI tools, and improvable in the open.
- How it works
- Packages reusable engineering practices as portable skills in Git and audits whether the guardrails around your coding agent are actually enforced.
Also in the workshop
Better together. Complete on their own.See how Holoself + Nextstep, AgentFlow 2.0 + Meshloop, and agent-otel-bridge + Dendro complement each other through public contracts—and why every pairing stays optional.→