We've thought about the hard parts

How do you maintain accountability when there's no shared ledger?

We reframed this as commitment fulfilment, not balance tracking. The system tracks whether participants do what they say they'll do, not whether they give as much as they receive. Receiving more than giving isn't a problem; it's the system working. Surplus flows to where it's needed.

Our position

• Accountability signals focus on delivery: "Did they do what was agreed?" (Yes / Partially / No)
• Lightweight escalation: stuck flag, participant conversation, accept outcome, governance escalation
• Response is graduated and human-judged, with no automatic penalties
• Tax compliance is a participant responsibility, not a network function

The farmer analogy: the issue isn't "you took more corn than you gave." It's "you said you'd help with the harvest and didn't show up."

How do you stop large organisations from using the network as cheap procurement?

We reframed this as behaviour, not identity. The system doesn't define or exclude "enterprises." Instead, it uses structural constraints and behaviour monitoring to ensure all participants engage as peers.

Our position

• No size-based exclusion: any participant exchanging genuine surplus as a peer is welcome
• Structural prevention: matching concentration limits, relationship diversity preferences, no procurement features
• Behaviour monitoring: directional asymmetry, surplus scheduling patterns, dependency patterns
• Graduated governance response: conversation, review, remediation or removal

The key insight: the issue isn't "you're too big to be here." It's "you're treating peers as suppliers rather than exchanging as equals."

Should this be a managed service or a federated network? And how do you prevent "start centralised, federate later" from becoming "never federate"?

We designed a governed managed service with a federation escape hatch. The risk isn't centralisation itself. It's unconstrained centralisation.

Our position

• Phase 1 deploys as a managed service, proving the model before adding federation complexity
• The protocol is an open standard; the operator is one implementation
• Five binding operator commitments: data portability, transparent operation, protocol conformance, participant representation, non-interference with exit
• Federation-ready architecture: portable identity, separable components, exportable trust
• A participant advisory body with defined powers (approve algorithm changes, concentration limits, review transparency)

The governance commitments make the escape hatch credible. The escape hatch makes the governance commitments enforceable. They reinforce each other.

How do you maintain trust without creating insider/outsider dynamics?

This tension is real and irreducible. Unlike other questions, no reframing dissolves it. We chose to lean toward openness: constrain outcomes, not entry.

Our position

• Default to trust: a new Newcomer tier with bilateral-only exchanges provides Sybil resistance through structural limits, not social gatekeeping
• Vouching becomes an accelerator (skip ahead), not a gate (required to enter)
• Network position decays over time (180-day half-life) to prevent permanent first-mover advantage
• Bounded anchor privileges, so even the most established participants have limits
• Professional monitoring of newcomer health metrics catches problems early

We accept that some barrier is inherent to any trust system. But we choose structural limits over social gates.

How do you handle malicious or exploitative participants?

We mapped all bad actor types (free riders, quality cheats, data harvesters, reputation manipulators, collusion rings, Sybil attacks) against the trust and accountability designs. Most are already handled by three reinforcing layers.

Our position

• Structural prevention: exposure limits, bilateral-only newcomer tier, concentration limits
• Detection: fulfilment signals, behaviour monitoring, network health metrics
• Graduated human-judged response: conversation, review, remediation or removal
• Three genuine gaps addressed: data harvesting (algorithm-mediated discovery is the defence), in-flight exchange handling (graceful wind-down), response gaming (symmetric monitoring)

The surplus scope directly softens the worst-case impact. The baseline is zero, so affected participants aren't out of pocket in the traditional sense.

Should participants see how the matching algorithm works? Won't transparency enable gaming?

We chose full transparency and accept gaming. If gaming the algorithm means being more trustworthy, fulfilling commitments reliably, and building genuine relationships, those are aligned incentives, not a problem.

Our position

• Every participant sees their own scores, match factors, and reasons for matching decisions
• Anti-harvesting boundary: you see everything about yourself, nothing about other participants
• Systemic risk addressed through dual concentration defence: diminishing returns in scoring plus a governance-set hard cap
• Algorithm changelog: public, append-only record of all changes
• Major algorithm changes require advisory body approval; emergency changes expire after 30 days

Withholding scores that determine participant opportunities is hard to defend, ethically or regulatorily.

Could surplus exchange displace paid work? What happens to employees?

SEP operates at business level and cannot see inside participant organisations. Labour effects are a governance concern informed by design, not a protocol concern.

Our position

• Transparency norm: participants expected to be open with employees about SEP participation
• Surplus scheduling detection (from enterprise capture monitoring) gains a labour lens
• Aggregate sector monitoring: operator watches for patterns suggesting substitution for hiring
• Honest onboarding: surplus defined as genuinely idle capacity, labour effects named as worth reflecting on

The design is deliberately light because there's no real-world data yet. Governance adapts based on evidence: observe, research, then adjust.

How do AI agents participate without SEP becoming agent-infrastructure?

SEP is an exchange protocol that agents can participate in, not an agent protocol. The boundary is clear: SEP handles matching, trust, and exchange orchestration. External agents handle capability execution and decision-making.

Where we are

• Phase 1 is human-only; agents participate through delegation in Phase 2, autonomously in Phase 3
• Integration with existing agent protocols (MCP, A2A) rather than competing with them
• Trust profiles will need adaptation for agent-specific signals

What's unresolved

• How do you verify agent identity and accountability?
• What trust signals work for autonomous agents?
• How do you prevent agent-specific attack vectors?

How do you seed a viable initial network?

Historical systems show critical mass is essential: roughly 50+ members with diverse offerings before matching becomes reliable. But getting to 50 requires convincing early adopters to join a network that doesn't yet deliver value.

What we think

• Sector-specific seeding may help: start with professional services where offerings and needs naturally overlap
• Anchor participants who bring credibility can attract others
• Low-stakes first exchanges prove the mechanism before larger commitments

What we don't know

• Which sector has the best overlap of surplus and needs?
• How do you identify and recruit anchor participants?
• What's the minimum viable network composition?

How do exchanges get taxed when there's no shared valuation?

The surviving systems (WIR, Sardex, ITEX) are all fully tax-compliant. Exchanges are taxable as barter income in most jurisdictions.

The tension

• The network operates without requiring parties to agree on value
• Tax compliance requires assigning values to exchanges
• We need to satisfy external requirements without reintroducing currency internally

Possible approaches

• Fair market value estimation at point of exchange for tax purposes only
• Surplus disposal framing: could exchanges be classified differently?
• Jurisdiction-specific guidance rather than one-size-fits-all