Guarded-human outcomes
Every build is judged by whether it helps the guarded person become safer, healthier, wealthier, calmer, or more capable.
bstrap-archangel
Angels second to none, built in public proof before public launch.
Build every Angel as a clean, repeatable, evidence-backed system that can be inspected, improved, and regenerated from the Archangel records.
The Angel network exists to guard people, help them grow healthier and wealthier, and make the right action easier. Wealth is an outcome of useful service, not the reason the Angels exist.
bstrap.ailocal-stagedlive action: gatedupdated 2026-06-30
Proof standard
The Angels do not get credit for claims. They get credit for evidence.
Active-active ownership
BSTRAP launch readiness lists 8 Archangel lanes across 4 company domains with named peers, watchers, commands, and gates.
Evidence before claims
A lane is not called done until commands, tests, artifacts, and a reported work key are recorded in company records.
No hidden live action
Provider changes, payments, filings, outbound messages, Drive pushes, deploys, and phone activation stay gated until explicit approval.
Competitive watch
Benchmark targets are generated from the BSTRAP Angels win board and converted into build decisions with source evidence.
8
Archangel lanes
Generated from BSTRAP-LAUNCH-READINESS.json active-active domain coverage.
tracked
Company-records QA
cd ../bstrap-site && npm test -- src/app/angels/__tests__/page.test.tsx
required
InfraRouter QA
InfraRouter readiness remains a named downstream proof lane before public service-provider onboarding.
215
Service account gaps
Generated from SERVICE-ACCOUNT-REGISTER.json; onboarding stays gated while gaps remain.
throttled
Fleet mode
Host pressure requires throttling: process count is 735, active MCP/browser-like processes are 241, so the Archangel must reserve QA and avoid broad parallel tool spawning.
4 checks
QA loop
cd ../bstrap-site && npm test -- src/app/angels/__tests__/page.test.tsx · cd ../bstrap-site && npm run build · Playwright screenshot/layout check: desktop and mobile, zero horizontal overflow
gated
Public deploy
Generated from BSTRAP-ANGELS-PAGE-WIN-BOARD.json; public launch requires explicit approval.
blocked
Onboarding gate
Do not onboard people yet.
Competitive watch
A standing team watches the outside field and turns every gap into a build.
SystemSignalBSTRAP standard
Claude Codesubagents, hooks, MCP, repository-native coding workflowsMatch the ergonomics, exceed with continuity, active-active failover, and evidence ledger.
OpenAI Agents SDKhandoffs, tracing, tools, guardrailsMatch the primitives, exceed with company asset registers and no-live gates by default.
Google ADKagent development kit, tools, deployment patternsTrack tool and deployment primitives, exceed with share-nothing company identity.
Hermes / Claw / Hybrid / Newnamed benchmark targets from BSTRAP-ANGELS-PAGE-WIN-BOARD.jsonNo rumor-driven claims. Capture source, compare primitives, build the missing advantage cleanly.
Legacy agent systems / new routing protocolstracked as named competitive targets until their public technical baselines are verifiedNo rumor-driven claims. Capture source, compare primitives, build the missing advantage cleanly.
Crown jewel readiness
Crown jewel readiness, without a false start.
required before onboarding
No false start
Public page remains gated until identity, domain, phone, support, privacy, security, and QA evidence are current in the company records.
required before onboarding
Identity
Identity, domain, phone, support, privacy, security, and QA evidence are current in company records.
required before onboarding
Every person-facing lane has owner
Every person-facing lane has owner, backup, account map, escalation path, support inbox, and rollback record.
required before onboarding
Human-action packets are ready for provider walls
Human-action packets are ready for provider walls.
deploy gated
Public page remains staged until deploy approval is explicit
Public page remains staged until deploy approval is explicit.
External benchmark
Beat the public standard with proof, not posture.
Hermes / Claw / Hybrid / New
Track the named outside and internal patterns as benchmark targets, not rumors, until each has source links, strengths, gaps, and build decisions.
Asynchronous coding agents
Compare the public baseline against BSTRAP continuity, company memory, live-action gates, and proof reporting.
Agent framework primitives
Compare the public baseline against BSTRAP continuity, company memory, live-action gates, and proof reporting.
Company-ready operating systems
Compare the public baseline against BSTRAP continuity, company memory, live-action gates, and proof reporting.
Clean-system gate
No hidden behavior gets trust by default.
- No rumor-driven security claims: every concern becomes a telemetry, dependency, prompt, MCP, and browser-automation audit item with evidence.
- Security concerns become telemetry, dependency, prompt, MCP, browser-automation, and network/live-action audit lanes with evidence. No unsupported spying claims are published.
- Audit lanes: telemetry, dependencies, prompts_and_stored_instructions, mcp_permissions_and_tool_scope, browser_automation_and_applescript_permissions, network_paths_and_live_action_gates.
- Default off for hidden live action: no provider change, outbound message, payment, filing, or deploy moves without a recorded gate.
Always building