The Perimeter Is Payroll

Run a hypothetical, labeled as such: two resignations land on the same Friday, same building, same system in both heads. A senior lawyer on the firm's payroll and a senior engineer on the MSO's payroll have each spent two years inside the same orchestration · the decomposition logic, the prompt chain, the eval suite. The lawyer's departure is a dead end this series has already walked to its wall. The engineer's departure walks into Waymo v. Uber. When Anthony Levandowski left Waymo, he left a forensic trail of roughly 14,000 downloaded files; a federal court granted partial provisional relief · it ordered Uber, his new employer, to wall him off from all LiDAR work while the case ran, though the broader injunction Waymo sought was denied · and the files themselves never came back: Levandowski took the Fifth rather than return them, and compliance with the court's order cost him his job, with Uber firing him in May 2017 for failing to comply. The case settled mid-trial for about $245 million in Uber equity, roughly a 0.34 percent stake, and Levandowski was later prosecuted on the download trail itself. Hold the two departures side by side and the asymmetry is exact. A court can order the engineer's new employer to wall him off from a field of work, and the engineer can be made to answer for his logs. No instrument can take the lawyer out of practice; Rule 5.6 voids any agreement that would try.

The asset did not change between the two heads. Only the employer's name on the W-2 did. The perimeter is payroll.

1 · The unbindable half, in two sentences

"The UI Is the License Agreement" owns the lawyer side of this asymmetry in full: Rule 5.6 voids the covenant, the trade secret cannot practically be recovered from a memory, and the moat must never be shown to a head that is not bindable · "bindable head" is that memo's coinage, used here as its. A reserved companion piece will take up the departing lawyer's economics, where equity aligns what no covenant can hold; this memo does not walk that ground. It is about everyone else in the building: the bindable population.

2 · Org design is IP strategy

Start with what the staff side actually signs, because all of it is live. Non-competes where state law allows them. Employee and client non-solicits. Invention assignment. Garden leave. Access architecture beneath the paper, deciding who can see what at all. With the FTC's 2024 non-compete rule dead · vacated nationwide in Ryan v. FTC (N.D. Tex., August 2024), the agency's appeals dismissed in September 2025, the rule formally removed from the CFR by a Federal Register final rule of February 12, 2026 · the entire stack reverted to state law, with the FTC retaining only case-by-case Section 5 enforcement against particular non-competes (its September 2025 enforcement statement, followed by actions in 2026), which means it is now a placement question: the same role under the same instruments is bindable in one state and not in another, and bindable on one payroll and not on the other.

The advisory literature has circled every one of these pieces and never made the move. Sidley's second installment on private equity investment in U.S. law firms notes the limits on enforceable non-competes for lawyers among the structure's challenges and stops there; Bloomberg Law's coverage runs the same instruments as diligence flags to be priced into a closing; the bar-journal register, Washington Bar News among it, catalogs the lawyer-side restraints and expressly declines the staff side, referring it out to the state's non-compete statute. The closer the literature gets to the staff covenant stack, the more deliberately it stops short · which is the white space this memo occupies. All of it treats the org chart as given and asks whether the paper on top of it is signed. None of it asks the prior question: which entity employs which role, holding which fragment of the system, in which state. That question is the design, and the covenants are reinforcement bolted to it. Decide it role by role · who may hold the decomposition logic, the prompt chain, the eval suite · and the answer is always the same: only heads on the MSO side of the employment line, because only those heads can be bound at all. The covenant perimeter is drawn before employment counsel opens the form file. Here is the instrument it produces.

Exhibit · The Perimeter Placement Table (generic roles; a design instruction, not advice)

The design must now survive the state where the engineers actually sit.

3 · The California stress test

California is this memo's pre-draft gate, and it is passed by narrowing, not by argument. Business and Professions Code 16600 voids the headline instrument outright, and the 2023 amendments closed the workarounds: AB 1076 amended Section 16600 itself to codify Edwards v. Arthur Andersen's rejection of narrow tailoring · the section is to be read broadly, voiding the restraint no matter how narrowly tailored · and added Section 16600.1, which makes it unlawful even to include a non-compete in an employment contract, required employers to send individualized void-notices to current and recent former employees by February 14, 2024, and makes violations actionable as unfair competition under Section 17200. Section 16600.5 (SB 699) makes the void covenant unenforceable regardless of where or when it was signed and hands the employee a private right of action. Nor do the non-solicits ride to the rescue: Edwards killed the customer non-solicit, Dowell v. Biosense Webster refused the trade-secret exception for one drafted too broadly, and AMN Healthcare v. Aya voided an employee non-solicit while declining to follow Loral v. Moyes, the older authority that allowed them · their statewide viability is at best uncertain, and this memo concedes it rather than lawyering around it.

So what holds, for the engineer row, in the engineer's home jurisdiction?

First, invention assignment. Labor Code 2870 protects only what an employee builds entirely on her own time, without employer equipment or trade secrets, and that relates neither to the employer's business nor to its actual or demonstrably anticipated research or development. A compliant assignment still captures anything made with the MSO's resources, anything relating to its business even off-hours, anything resulting from work performed for it. The prompt chain and the eval suite are squarely inside that capture.

Second, trade-secret conduct claims under CUTSA · and here the concession comes first, because it is this series' own published canon. What lives only in an engineer's memory is nearly as hard to recover as what lives in a lawyer's; head-memory futility does not check W-2s. The engineer-side strength of trade-secret law is not that it reaches heads. It is that engineering work leaves what lawyering does not: download logs, repository access records, commit histories · conduct and forensics · and that courts will layer remedies on that trail which Rule 5.6 makes unthinkable against a lawyer, up to and including ordering a departing engineer's new employer to wall him off from a field of work, as the Waymo court ordered Uber to do. The perimeter holds against engineers not because the law is stronger there but because the remedy is permitted there.

Third, confidentiality duties, which 16600 does not touch and which survive departure in full. Fourth, access architecture · the same discipline "The UI Is the License Agreement" pointed outward at lawyers, turned inward at the factory floor: tier the build so no single bindable head holds the decomposition logic, the prompt chain, and the eval suite at once, and the memorized fraction is a fragment.

The honest line: in California the perimeter is covenant-light · assignment, conduct claims, confidentiality, tiers · and that is a perimeter, not a wish. The thesis is not Delaware-only. It is narrower in California and stated as such.

4 · The legislature moves the perimeter

Illinois HB 5487, passed May 31, 2026 and awaiting the governor's signature as of early June 2026, bans post-termination non-competes inside the MSO arrangements it regulates, and "The MSA Is the Cap Table: How to Read a Legal MSO Like a Bond Indenture" has already priced that tail · the enacted-statute notch · so this memo cites the price and does not re-run the register. The point here is structural: a legislature drafting a statute against the staff covenant stack is evidence the stack is load-bearing; nobody legislates against paper that does nothing. And the consequence is one row of the exhibit: if the ban is read to reach the MSO's own staff covenants, an Illinois engineer row loses its non-compete and drops to the California subset · assignment, conduct claims, confidentiality, tiers; if it is read more narrowly, the row keeps the full stack for now. Washington has just confirmed the pattern at full statutory volume: ESHB 1155, signed March 23, 2026, bans the non-compete outright effective June 30, 2027 while expressly preserving non-solicits, confidentiality, and invention-assignment covenants · the latest evidence that legislatures are killing the headline covenant and leaving the covenant-light stack alone. The design answer is the same either way: build every row so it still holds on the covenant-light stack, because the covenant-light stack is the one no legislature has touched.

5 · The ethics hook has nothing to grab

One constraint remains that sounds fatal and is not, and the precision matters because none of the authorities says what they are casually cited for. Texas Ethics Opinion 699 addresses departing lawyers; its relevant line is dictum, observing that lawyers leave firms free of "standard contractual restrictions, such as non-compete restrictions, that often constrain the subsequent employment of non-lawyer employees." Philadelphia Bar Opinion 96-5 voided an employee non-solicit "to the extent that it applies to employees of the Firm who are attorneys" · the limitation doing the work by negative implication. NYSBA Opinion 1151 addresses only the lawyer's own right to practice. None of the three affirmatively blesses a staff covenant; together they establish only that Rule 5.6's prohibition runs to lawyers. The practitioner compilations that collect these opinions · Epstein Becker Green's survey of ethical issues for attorneys in restrictive covenants is the thorough one · read them as constraints on the lawyer-employer and stop there.

Stopping there misses the structural point nobody has printed. Whatever residual ethics argument survives against staff covenants runs through a lawyer's participation in imposing them · the rules govern lawyers, and they reach nonlawyers only through a lawyer's conduct. In the MSO structure, the entity covenanting the engineer is not a law firm and has no lawyer as party to the covenant. The MSO signs; the firm does not. The hook that was always weak by negative implication now has nothing to grab. That is not a loophole to be celebrated; it is one more consequence of the same boundary, and it points the same direction as everything above: the instruments live on one side of the payroll line.

6 · The practice pushes back

The strongest objection comes from the veteran COO, and it deserves printing at full strength:

"You have drawn an elegant fence around the wrong field. The people who actually internalize this system are the lawyers who use it daily · your own memo says competent use is indistinguishable from memorization. Binding the engineers protects the blueprints while the product walks out the unbindable door every evening. And to chase covenants, you would split my workforce across two employers, doubled HR and fractured culture included, for instruments California voids and Illinois just legislated against. The perimeter you can defend does not contain the value, and the value I must contain has no perimeter."

Concede the first half entirely. It is this series' published thesis, and it is the reason the two memos are companions rather than rivals. The interface decides what may enter an unbindable head; this memo decides which heads may hold the rest. Neither works alone: covenants without surface discipline guard the blueprints while the finished goods leak through the lawyers' daily use, and surface discipline without placement leaves the factory itself sitting in heads no instrument reaches. The fence around the engineers is not a substitute for the keyhole in front of the lawyers; it is the other half of the same wall.

Concede the cost too, because it is real and the objection prices it fairly. Two employers means two HR functions, two handbooks, two cultures to keep from curdling into us-and-them. Nobody should build an MSO for the covenants. The structure exists for reasons settled elsewhere in this series · the economics, the topology, the capital · and the covenant perimeter is the option that structure creates for free. The instruction in this memo is not to build the fence; it is to stop leaving a fence you have already paid for unexercised, with the factory's keys distributed as though both payrolls were one.

And say the narrowing before the objector does: in California the exercised option is covenant-light. Assignment, conduct, confidentiality, tiers. Narrower, and still standing.

7 · The perimeter map

The decision rule fits in a breath: the factory goes on the MSO payroll; the firm payroll sees surface only; no hybrids. The exhibit in Section 2 is that rule spelled out role by role, and it is built to be argued with at a whiteboard, one row at a time.

The rule's hardest case is the one every COO will raise at that whiteboard, so answer it here. The dual-hat hire · the brilliant associate who wants to build the prompt chain, the lawyer-engineer who would be the best orchestration designer in the building · is forbidden. Not discouraged; forbidden. The moment the chain enters a head with a license to practice, it has entered a head Rule 5.6 guarantees can leave with it, and no MSO job title, no secondment, no carefully drafted dual employment agreement changes which rule governs that head. The most talented candidate for the factory may be the one person who can never be allowed inside it. That is the price of the structure, stated without flinching, and it is the whole memo in a single staffing decision: the same system, defensible or indefensible depending on nothing but who signs the offer letter. The perimeter is payroll.

Jopese operates a legal MSO, and this memo therefore describes a design problem it owns.

This memo is published by Jopese, a legal management services organization operated by HIRO PARTNERS LLC, a Texas limited liability company. It is offered for educational and analytical purposes only. It is not legal, tax, or investment advice, and it is not an offer to sell or a solicitation of an offer to buy any security or service. Jopese is not a law firm and does not provide legal advice or legal services; legal services are delivered by an independent law firm under a separate engagement in which Jopese does not participate. The two-resignations hypothetical and the Perimeter Placement Table are hypothetical illustrations of generic roles and instruments; they do not describe any actual person, employment arrangement, agreement, or transaction, including any to which Jopese is a party, and they are not advice about the enforceability of any instrument in any jurisdiction. References to specific companies, publications, ethics opinions, court doctrines, statutes, and regulatory developments are drawn from public sources and are provided as market commentary, not as an endorsement, a recommendation, or a representation of any relationship.