The previous piece ended on a claim that workers are rerouting.1 AI has broken the entry ramp inside firms and simultaneously equipped the same graduates to build outside them. Fragmentation is the labor market outcome, not just a competitive dynamic.
Which leaves a question the piece deliberately didn't answer. If the firm is no longer the only viable platform, what actually keeps a builder inside one?
Salary alone won't do it. Salary is a floor, not a reason. Career progression won't do it either, because the ladder that made progression valuable was the same ladder that just collapsed.
The firm has to offer something the open market cannot — and the open market now offers almost everything.
There is one thing the firm still has that a solo builder does not. Infrastructure. Legal, financial, distributional, operational. The scaffolding that turns a working product into a durable business. That is the trade the next employment contract has to make explicit.
What Bureaucracy Built
The current employment contract is an artifact. It was designed for a labor market where the firm was the only serious platform for productive work, and the terms reflected that asymmetry.
Every standard IP assignment agreement is written as a blacklist. Everything you build during your employment belongs to the company, except for what you explicitly carve out in writing before you start. The clause was defensible when the firm supplied the tools, the customers, the capital, and the time. It is indefensible now, when a graduate can spin up production infrastructure on a laptop over a weekend.
But the deeper artifact is not the contract. It is what the contract enabled.
Promotion inside a bureaucracy is not primarily a function of output. It is a function of sponsorship — whether someone with authority over your career decides to advance you. Which means the operative skill for career progression has been, for decades, the ability to be liked by the right people at the right time.
The current contract does not reward what you built. It rewards who noticed. That is the artifact that has to be dismantled.
Every firm that has tried to reform promotion mechanics has run into the same wall: the reformers themselves rose through sponsorship, and their instincts about "merit" are calibrated to a system that measured social capital as if it were competence. The reform never sticks because the people implementing it are the product of what they are being asked to replace.
Now enters a new player... AI.
AI does not care about sponsorship. Neither does a live customer paying $12 a month for something that works. The market has already introduced a promotion mechanism that bypasses the entire adjudication layer. The question is whether the firm can incorporate that mechanism before the builders leave to use it directly.
What 20% Time Actually Was
Every serious attempt at this has run through the same failure mode. The most famous was Google's 20% time.
The origin was 3M, not Google. In 1948, 3M introduced its 15% time policy, allowing employees to spend a portion of their paid work time pursuing projects of their choice. That policy produced Post-it Notes, Scotch Tape, and a meaningful share of 3M's product catalog over the following decades.2 Google borrowed the model, expanded it to 20%, and produced Gmail, AdSense, Google News, and the prototypes of Google Maps.3
Both programs are cited constantly as evidence that structured autonomy produces innovation. Both are also cautionary tales.
The failure mode is instructive. 3M and Google both offered time. They did not offer ownership. Every product that came out of 20% time became a Google product, and every product that came out of 15% time became a 3M product. The employee who invented Gmail was paid a salary. The employee who invented the Post-it Note was paid a salary.
The upside of the innovation flowed entirely to the firm.
That worked in an era when the alternative for the employee was worse — where launching a product independently required capital, distribution, and infrastructure that only a firm could provide.
That era is over.
A twenty-two-year-old can now launch on a laptop and reach paying customers within a week. Structured autonomy without ownership is no longer competitive with structured autonomy plus ownership, which is what building alone actually offers.
Every 20% time program was designed to prevent leakage in a world where the firm was the only viable platform. In a world where it isn't, they leak by design.
The next generation of programs cannot repeat the ownership asymmetry. They have to solve it.
What Germany Already Codified
There is one legal regime that has already run this experiment at national scale, and it is worth studying carefully.
Germany's Arbeitnehmererfindungsgesetz — the Employee Inventions Act, or ArbEG — establishes a principle that runs directly against the standard US employment contract. Under German law, the employee is the initial owner of any invention they create, as a matter of inventor's natural right. The employer can claim certain inventions, but has to compensate the employee beyond salary in exchange for the assignment.5
The law splits inventions into two categories.
- Service inventions (Diensterfindungen) arise from the employee's role or from significant use of company resources. The employer can claim these, but owes the employee separate remuneration based on the invention's economic value.
- Free inventions (freie Erfindungen) fall outside the employee's role. These belong to the employee outright. If a free invention falls within the employer's business scope, the employee must offer the employer a non-exclusive license on reasonable terms — but ownership stays with the inventor.5
The empirical result is striking. Over 90% of German patent applications come from employees operating under this framework.6 The system that prioritizes inventor rights produces more inventions, not fewer.
The ArbEG does most of what the next American employment contract will need to do. It carves ownership by scope rather than by employment status. It requires the firm to pay separately for what it claims. It preserves the employee's underlying rights even when the firm has the primary commercial position.
It is not the model to copy directly. It is patent-heavy and was designed for an era of industrial R&D, but it establishes the philosophical precedent that matters most: the firm does not automatically own what the employee builds, and the burden of justifying a claim sits with the firm, not the worker.
That inversion is the missing piece.
The Whitelist Inversion
The mechanics of the next contract flow from one philosophical change.
Every standard IP assignment agreement today is structured as a blacklist. The company owns everything you build, with narrow exceptions carved out in advance. The employee bears the burden of proof for anything they want to keep.
The next contract has to be a whitelist. The company defines, in a document updated annually, exactly which product surfaces and market segments constitute its protected territory. Everything outside that whitelist is presumed employee-owned by default. The burden of proof reverses: the company has to establish that a build falls within the whitelist, not the employee has to prove that it doesn't.
The blacklist protects everything by default. The whitelist protects what actually matters. One is a legacy of a labor market that no longer exists. The other is designed for one that already does.
The whitelist has to be specific enough to be enforceable and narrow enough to be honest. A software company cannot whitelist "software." It has to name the products, the markets, the customer segments, the technical domains. If a build is genuinely outside the company's competitive surface, the company has no legitimate claim on it.
This does not eliminate the company's protections. It replaces vague overreach with defined territory. The company keeps what it actually competes on. It stops claiming everything else by default.
The whitelist has to be updated annually and negotiated with employee representation, not imposed unilaterally. Otherwise, the mechanism collapses back into a blacklist over time, as legal will progressively expand the definition until it covers everything.
Adjacent Build Rights: The Mechanics
The whitelist is the foundation. The build rights sit on top of it.
One register note before the mechanics, because it matters for how they are read. The inversion — employee ownership by default, the burden of proof on the firm, a protected floor, a defined exit — is the claim of this piece. The numbers attached below are proposals: opening positions for a negotiation that has to happen against real firms and real builds. Argue with the numbers. The inversion is not the negotiable part.
Scope. Any project an employee builds outside the whitelist, on their own time and equipment, without using company-confidential information and without selling to the company's clients, is presumed to be the employee's own. This is the default state, not a special permission.
The trigger. A build enters the shared regime when it hits sustained revenue: $4,000 monthly recurring revenue held for six consecutive months, or $48,000 in trailing twelve-month revenue, whichever comes first. Below the trigger, the company has no interest and no claim. Above it, both sides have a decision to make. The sustained threshold prevents a single lucky month from dragging a hobby project into a corporate negotiation before the employee has any leverage.
Right of first refusal. At the trigger, the company has 60 days to decide whether it wants to bring the product under its infrastructure. If it declines, the employee retains full ownership and walks away with a one-time modest right-of-first-refusal fee — cash, capped, not perpetual equity. No permanent stake in a business the company chose not to support.
If the company absorbs the product. The revenue share progresses on a defined schedule as the product scales, moving from a base split toward a maximum of 70% company / 30% employee — with the employee's 30% treated as a non-diluted floor of net economics from the product line. Not phantom equity. Not a bonus pool. A protected minimum that survives future rounds, restructurings, or reorganizations.
The floor is load-bearing — its protection, not its percentage. Thirty is the opening proposal; the non-dilutability is the mechanism. It is what makes the employee a principal rather than a beneficiary. Every existing intrapreneurship program has failed because the employee's stake could be diluted, restructured, or clawed back at management's discretion. A protected floor makes that impossible by contract. If the company cannot commit to some floor that survives every future round and reorganization, the whole system collapses back into a bonus program with better marketing.
A protected floor is not a concession the firm makes. It is the price of participating in the next labor market. Firms that refuse it will hire only from the cohort that had no other options.
The graduation clause. When a product's revenue produces sustained employee income of $750,000 per year or more, the employee has the right to spin the product out entirely, with the company retaining a defined minority stake and continued revenue share on the terms in place at graduation. This is the moment the firm has to prove it actually served the builder rather than captured them. If the graduation terms are visibly generous, the firm becomes a lifetime platform. If they are punitive, word spreads within a hiring cycle, and the firm loses access to the next cohort entirely.
The graduation terms are the single most important clause in the contract. Every other mechanism can be negotiated. The graduation moment is what tells every prospective employee whether the firm is a partner or a trap.
The Political Reality
This will be fought. It is worth naming who fights it and why.
Boards will resist because it makes future valuations messier. A firm with dozens of protected employee stakes across dozens of product lines is harder to value, harder to acquire, and harder to restructure than a firm with a clean cap table. That messiness is not a bug. It is the price of retaining people who could otherwise leave, and it will show up as a valuation premium over time as talent pipelines diverge between firms that offer this and firms that don't.
HR departments will fight harder, and less honestly. The current HR function exists in part to adjudicate promotion — to decide who is ready, who is a fit, who deserves the next step. Adjacent Build Rights replaces that adjudication with revenue verification. If a build hits sustained MRR, the employee has advanced by definition. There is no promotion committee to convene, no calibration cycle to run, no sponsorship to secure. The layer whose value proposition was managing that process becomes redundant for anyone operating under the new contract.
Legal will drag on the whitelist definition, because narrowing the company's default claim looks like a loss to a lawyer whose job is to maximize the company's claim. This is where the philosophical shift has to be defended at the top. If the CEO does not understand why the whitelist matters, general counsel will re-inflate it every year until it swallows everything again and the mechanism dies quietly.
The firms that adopt this will not be the largest. They will be the ones that have already noticed they cannot hire the top of the next graduate class under the old terms.
The Best Case Against This
The objections above are political — they come from the layers the contract displaces. These four are structural. They are the strongest ones I know, and they deserve answers at full strength rather than characterization.
"A firm carrying dozens of these is unacquirable." Mostly true — for the current buyer. A firm with thirty non-dilutable revenue-share obligations cannot be bought the way a clean cap table is bought, and no amount of premium talk changes that. What it can be bought as is what it actually is: a catalog. Publishing houses, music catalogs, and royalty aggregators trade constantly, and their buyers price exactly this shape — durable cash flows with contractual participants attached. What the clean cap table sells, in a world where builders have options, is partly fiction: 100% ownership of product lines whose builders would rebuild them elsewhere within a year of any acquisition that mistreats them. The ABR firm sells fewer points of ownership and a much smaller key-person discount. That is a different buyer, not a missing one.
"You will attract exactly the people planning to leave." Yes. That is the design, not the defect. The adverse-selection objection assumes the firm's alternative was retaining these people on the old terms — it was not; the old terms are why they never applied. Adjacent Build Rights rents the steepest years of an ascent the firm could never have owned, takes 70% of the product line built during those years, and converts the departure itself into recruiting evidence through the graduation clause. The counterfactual is not the same employee, captured. It is the empty chair.
"Annual negotiated whitelists are expensive at scale." They are. So is the current arrangement — its costs are just booked elsewhere: IP disputes on departure, the enforcement theater of clauses courts increasingly decline to honor, and the silent attrition of every candidate who reads an assignment agreement and self-selects out. The whitelist moves an unpriced cost onto the books, once a year, in daylight. For firms too small to carry even that, the minimum viable version below is the answer.
"If inventor-rights regimes work, why hasn't ArbEG spread in seventy years?" Because nothing forced it. Regimes like ArbEG are adopted where labor has structural power, not where firms have good taste — Germany codified it into a labor market with works councils and sectoral bargaining. The American firm spent those seventy years in a market where it was the only platform, which meant it never had to pay for what it claimed. The entire argument of this series is that the platform monopoly just ended. Non-adoption to date is evidence about the old market, not about the mechanism. The question is not why ArbEG failed to spread under conditions that protected the firm — it is what firms do now that those conditions are gone.
The Minimum Viable Version
The framework above assumes counsel and cap-table sophistication, and the firms most likely to move first — small, cash-tight, allergic to process — have neither. The deployable subset for a sub-twenty-person firm is four clauses:
- A named whitelist. One page. The products and customer segments the firm actually competes on, listed. No annual apparatus at first — just the list, and the default that everything off it belongs to the builder.
- A floor on the stake. Whatever split is agreed, it cannot be diluted or revoked. The number can be modest; the protection cannot.
- A defined spin-out path. One paragraph stating the conditions under which the builder can take the product and leave, agreed before there is anything to fight over.
- License, not assignment, for pre-existing work. Anything the builder brings in stays theirs; the firm takes a license to what it needs.
That is an afternoon with a template, not a legal department. It is missing the trigger machinery, the refusal clock, and the revenue schedule — all of which matter at scale, and none of which are what makes the contract credible. What makes it credible is the inversion, and the inversion fits on two pages.
The Firms That Survive
The previous piece argued that governance is planning for an economy that is fading. This is what governance for the emerging economy has to look like.
The firms that survive the next decade will be the ones that stop treating employment as an all-or-nothing capture and start treating it as a platform relationship. The employee retains their upside. The firm provides the infrastructure. Both sides win when the build succeeds. Neither side is coerced into staying past the point where the arrangement makes sense.
This is not radical. It is the pattern every other durable institution — universities, professional guilds, publishing houses — has used for centuries. The firm is late to it because the firm was the last institution to enjoy a labor market where it could avoid the pattern entirely.
One more honesty before the close. This contract is written for the builders a firm has to bid for, and that class is thin — most employees will never trip the trigger, and nothing here improves the median worker's position under compression. The next piece sizes that split properly.7 What the thin class changes is the terms: once the best of a cohort can walk in with a build instead of a résumé, the firm that captures everything by default stops being able to hire the top of any cohort at all. The contract is written for the few. The pressure it creates is felt by every firm.
Adjacent Build Rights is a specific implementation. The specific numbers can be argued. The 30% floor, the sustained MRR trigger, the graduation threshold — all of those are calibration points that need pressure-testing against real firms and real builds. What is not negotiable is the underlying inversion: the firm no longer owns everything by default, promotion no longer runs through sponsorship, and the employment contract stops pretending the labor market of 1980 is still the one being negotiated.
The firms that write this contract first will attract the graduates who currently see no reason to walk through their door. The firms that refuse will discover, one hiring cycle at a time, that the door was the point.