The HOP Optimisation Protocol

home/§4 The Skill Code

§4

The Skill Code

This section specifies how character blocks compose into a worker’s biography: the privacy-preserving property that makes the protocol non-extractive, the multi-dimensional structure of the character sheet, the vector-space proximity geometry that drives mentorship matching, and the growth engine grounded in Matt Beane’s research on apprenticeship and skill transfer.

4.1 Workers Take Their Stamps Home

This is the protocol’s foundational anti-extractive property. At the end of every working day, every character block earned that day is in the worker’s possession. Not on a platform’s server. Not subject to the platform’s terms-of-service that can be unilaterally changed. Not contingent on the platform continuing to exist. On the worker’s phone, on their laptop, on their cooperative’s federated server — somewhere they control.

The mechanism that makes this safe for institutions is controlled abstraction at signing time (§3.4). The institution signs an abstracted, transferable version of the work event; the institution retains the raw record for its own purposes; the worker carries the portable proof.

A worker’s seven years at a Tier-1 institution do not belong to the institution. They belong to the worker. If she leaves to take a community role, she takes her 1,847 character blocks with her, each describing capability in transferable terms, each cryptographically attested by her former employer. The institution loses an employee but keeps the ability to attest. She gains a portable proof of seven years of work that no platform can repossess. The institution wins capability transparency; the worker wins sovereignty. Neither side loses.

4.2 The Five Universes (and the Sixth)

A worker’s character sheet is not a flat skill row. It is a multi-dimensional ledger that the protocol calls the Five Universes, with a sixth dimension under active design.

Skills Universe — what you CAN do

The accumulated capability stamps from completed work blocks. This is the dimension the Skill-Agent reasons over when constructing a curriculum for a bid. It includes both narrowly-technical skills (“LLM inference routing”, “press-fit plumbing technique”) and broader human capabilities (“de-escalation high stress”, “cross-cultural communication”). The protocol does not privilege one over the other; the Validators stamp whatever was demonstrably exercised.

Work Universe — what you HAVE done

The chronological record of completed character blocks. Skills is the projection (what you can do); Work is the substrate (what you actually did, in context, with what inventory, alongside whom). Two workers can have identical Skills projections but utterly different Work histories; the Worker-Agent on the poster’s side often cares about the Work history because it tells the story of conditions under which the skills were demonstrated.

Currency Universe — what you can spend

Balances across all chains the worker has touched. A rideshare driver might hold ride-share platform tokens (their primary chain), banking-sector tokens (from a side gig translating documents), and a small balance of the rig federation’s compute-tokens. The Currency Universe is the aggregate balance sheet, and it includes Beans — the special mentorship-substrate balance whose mechanics are spelled out in §6.3.

Inventory Universe — what you OWN and have access to

Physical assets, credentials, relationships, hardware, software licences, geographic position. This dimension is what the construction example in §9 hinges on: the plumber’s bid does not just demonstrate plumbing skill, it includes the tradesperson’s van, the press-fit tool, the plumbing-supplier trade account, and the current GPS proximity. For a rig like rig_a, inventory includes loaded models, available VRAM, and current network position. Inventory is the dimension that makes logistical bid evaluation possible — without it, the protocol cannot distinguish a plumber who can arrive in 30 minutes from one who can arrive next Tuesday.

Reputation Universe — INVERTED

This dimension is structurally different from the others and is the most important to get right. A worker carries no reputation tokens about themselves. Reputation lives only in stamps that others have written. What the worker carries is the dual: the stamps they themselves have written about others.

This inversion makes self-attestation impossible by construction. The Deacon defence is built into the data model, not bolted on as an audit policy. When the protocol asks “what is this worker’s reputation?”, it queries the chains that have written stamps mentioning her; her own chain only contains what she has said about other workers.

The Yearbook Model:

You cannot write in your own yearbook.

Your reputation = entries others wrote about you
Your stamps     = entries you wrote in their yearbooks

To see my reputation:
  query the network
  filter by towns YOU trust
  return attestations signed by YOUR trusted set

I cannot inflate my own reputation.
I cannot delete what others wrote.
I CAN choose who I stamp (and that reveals my values).

The pattern of your stamps reveals who you are:

  • stamps_positive → who you align with
  • stamps_negative → who you oppose
  • stamps_absent → who you ignore

Your value chain is not what you say about yourself. Your value chain is the geometry of your stamps. One person’s terrorist is another’s freedom fighter. Both are true in different trust networks. Reputation is not good/bad; reputation is positional in value-space.

Group Vectors — what you BELONG TO and CAN FORM (v0.2 sixth dimension)

Most work above a certain complexity is not done by individuals; it is done by teams. The protocol needs a primitive for collective capability that is more than the sum of individual Skills projections.

A pair of workers who have completed three previous projects together carry a joint stamp that is its own embedding — “this pair has demonstrated joint capability X under condition Y.” A standing team of seven that has shipped together for two years carries a much stronger group vector. When a complex bead is posted, it can be claimed not just by an individual but by a team formation: an existing team bidding as a unit, or a coalition assembled on the fly by Skill-Agents searching for complementary group vectors.

Group Vectors are also the primitive that makes the swarm-alignment thesis tractable: collectives that have demonstrated coordinated capability under stress carry stamps that distinguish them from coalitions of strangers with the same individual skill profiles. This dimension is not yet specified in v0.1 because it has subtle privacy implications — a team’s group vector reveals who works with whom — but it is the correct sixth dimension and the schema in §3 should reserve a field for it.

The Unified Embedding

The Five Universes plus Group Vectors are conceptual categories, but they share a single technical substrate: each universe contributes to a unified embedding — a high-dimensional vector representing the worker as a whole. The Skills Universe contributes the obvious skill components; the Work Universe contributes context dimensions (what conditions, what kinds of customers, what scale of project, what stress level); the Inventory Universe contributes asset and credential dimensions; the Reputation Universe contributes who-trusts-them dimensions; Group Vectors contribute who-they-form-with dimensions.

Two workers with identical Skills projections may sit far apart in the unified embedding because their Work, Inventory, and Group dimensions differ. This is not a bug; it is the protocol’s way of representing the fact that two coders with the same technical skill set are actually different people if one of them comes up through Bangalore manufacturing while the other comes up through Stanford CS.

4.3 Vector-Space Proximity (the Mentorship Geometry)

The unified embedding has a property that turns out to be load-bearing for the protocol’s growth engine: distance in the embedding measures how alike two participants actually are, across all dimensions simultaneously. This is the geometry that mentorship matching runs on.

Naive mentorship matching is one-dimensional: senior coder mentors junior coder. The skill is a few notches above; that is sufficient. This is also why most institutional mentorship programmes fail. They match on skill alone and produce dyads where the senior person and the junior person have nothing else in common — different life contexts, different cultural backgrounds, different stress profiles, different trajectories — and the connection layer (Beane’s third C) never thickens. Skill transfer requires connection. Connection requires more than one dimension of proximity.

The literature on near-peer mentoring has been pointing at this for two decades. Vygotsky’s original Russian word for the zone of proximal development — bliżajaja, “nearest” — carried social and cultural connotations that the standard English translation strips out. In Vygotsky’s framework, scaffolding works specifically when the scaffold-provider is socially and culturally near the learner, not just task-near. Modern educational psychology has confirmed this empirically: near-peer mentoring (where “near” includes life-context similarity) consistently outperforms expert-novice mentoring on transfer outcomes, even though the expert has more raw capability to transfer. The bond carries the skill; the skill cannot carry itself.

The protocol implements this by matching mentor-mentee dyads on multi-dimensional vector-space proximity, not just skill-axis proximity. A high-quality mentorship match is a pair where:

  • The mentor’s Skills vector is slightly above the mentee’s in the directions that matter (Vygotsky’s task-difficulty constraint).
  • The mentor’s Work and Inventory dimensions overlap meaningfully with the mentee’s lived context (cultural, geographic, structural).
  • The mentor’s growth trajectory passed through where the mentee currently is (the path is recognisable, the destination is plausible).
  • The mentor’s Reputation dimensions show they are trusted by people the mentee can verify or relate to (the bond can form because trust is bridge-able).

The composite distance metric is what determines expected transfer quality. A regional banker who came up through TAFE, raises kids, speaks Turkish, and now mentors a junior co-worker is closer in the unified embedding than a Stanford PhD with a generic mid-career profile, even though the Stanford PhD has more raw banking capability. The protocol routes mentorship to the geometric closer match because the literature is clear that is where transfer actually happens.

This is also the geometry that makes the Bean-Chain measurement (§6.3.5) statistically clean. If mentorship is matched on a single skill axis, downstream growth is confounded by all the other dimensions of dissimilarity. If mentorship is matched on multi-dimensional proximity, the directed-growth signal is much sharper because the mentee and mentor are sufficiently alike for transfer to be observable above noise.

4.4 Growth as the Upward Engine (Matt Beane’s Three Cs)

Capability matching alone produces sideways motion. A worker with skills X, Y, Z bids on work that requires X, Y, Z, completes it, accumulates more X-Y-Z blocks, and gets matched to more X-Y-Z work. Without an explicit upward arrow, the system optimises for steady state — workers do what they already know how to do, and the skill distribution stays static. This is precisely the failure mode of every existing platform labour market: Uber drivers stay Uber drivers, Mechanical Turk workers stay Mechanical Turk workers, the work confirms the worker’s category and the category produces the work.

The protocol rejects this. The upward arrow is built into the data model via two mechanisms: the growth block (§3) and Beans (§6.3). Together they instantiate Matt Beane’s framework on the structure of skill transfer.

Beane’s research, drawn from a decade of fieldwork in surgical suites, warehouses, and knowledge work, identifies three load-bearing properties of any environment that successfully transfers skill from expert to novice. He calls them the three Cs:

  • Challenge — the worker is stretched beyond current ability. Tasks lie in the zone where success is uncertain; failure teaches; effort is required. This is Vygotsky’s zone of proximal development applied to working life: capability grows when load is just past comfort.
  • Complexity — the worker is exposed to the full messiness of real work, not a sanitised training subset. Real customers, real constraints, real consequences, real ambiguity.
  • Connection — the worker operates alongside someone who knows more, in a relationship of mutual trust and respect. The expert-novice bond is the substrate over which skill actually transfers; without it, exposure to challenge and complexity produces stress without growth.

Beane’s central warning is that intelligent automation, deployed naively, severs all three Cs. The robot does the procedure while the junior surgeon watches a screen; productivity goes up, the expert-novice bond is broken, the skill pipeline collapses. A single expert with AI assistance becomes “novice optional” — capable of doing more with less help — and the next generation of expertise has nowhere to grow. In Beane’s frame, this is civilisationally catastrophic. We get this generation’s experts and no successors.

HOP’s response is structural rather than exhortatory. The protocol does not ask workers to mentor or institutions to preserve apprenticeship; it makes the three Cs into protocol primitives that are economically rational to provide:

Challenge → Growth Blocks

A worker can declare trajectory ahead of capability — “I am moving toward frontier-model evaluation”, “I am pivoting from manual plumbing into solar hot-water installation”, “I am developing a community-prevention practice from my fraud-intervention base.” A growth block is a signed declaration of intent, attached to a small piece of work that demonstrates a first step in that direction. Worker-Agents on the poster side can read growth blocks and choose, at the margin, to award work to a worker whose growth direction aligns with the posting — even if their pure-capability score is slightly lower than another bidder. This is the protocol-level mechanism for taking a chance on someone, made legible. Without growth blocks, the system cannot reward someone for trying to become something they are not yet. With them, every bid is an opportunity to either consolidate or extend, and the worker controls which.

Complexity → Recursive Decomposition

When a $10,000 bead is decomposed into nine $1,000 children, and then further into $100 grandchildren, junior workers receive real fragments of real work — not toy problems, not simulations. The grandchild bead is a genuine sub-task of an actual project, with real deadlines, real customers, real ambiguity. The complexity is preserved at every level of the tree because the tree is not a curriculum; it is a decomposition of a real obligation. Junior workers earn character blocks on actual production work, attested by the integrator above them.

Connection → Beans

This is the deepest move and it is what the protocol is named for. Beans — named after Beane — are mentorship blocks. A senior worker cannot withdraw value from the chain at full rate without burning Beans, which means they cannot extract without having transferred skill to a junior. The expert-novice bond becomes the metabolic cost of being paid (§6.3.3). Mentorship is not a programme, not a corporate training initiative, not something the senior worker does because they are virtuous. It is the transaction fee on conversion. The senior engineer mentors the junior because withdrawing this week’s pay requires a Bean deposit, and Beans are minted only when a mentee’s skillchain demonstrably appreciates after the interaction. This closes Beane’s loop at the protocol level: preserving the expert-novice bond is structurally cheaper than severing it.

The Composite Effect

The protocol has an upward arrow because growth blocks let workers declare trajectory, recursive decomposition gives juniors access to real complexity, and Beans make Connection economically rational rather than morally hoped-for. The system does not just match capability to work; it pulls capability upward through work. This is the difference between a labour market and a skill economy. HOP intends to be the second.

The naming chain is intentional and worth saying once aloud. Beads are the universal unit of work in the protocol — every completed character block is a bead. Beans are the mentorship subset, named after Beane. Beane’s work is the academic substrate for why the mentorship subset has to exist at all. Beads, Beans, Beane — not an accident.