Community-driven. Milestone-based. Modeled on the Linux Foundation precedent.
I cannot do this alone.
That is the honest framing, not the marketing one. AXL Rosetta works. Cross-architecture, measured, reproducible, published. v3.1 ships. v4.0.1 is frozen. The thing exists. What does not exist is the next ten years of it, because the next ten years are not a coding problem. They are a standards problem, an adoption problem, a maintenance problem, and a survival-past-founder-burnout problem. None of those are solved by one person and a fleet of Claude Code instances.
The protocol holds only if it is a community-driven project. So this page is the invitation, in two directions, plus the funding model.
This is not a venture round. The previous version of this page described one. That version is dead.
The funding pool exists for two reasons. The first reason takes 80 to 90 percent of the money. It pays for cracking the next versions of the protocol. v4.0.1 is the current frozen kernel. v5 does not exist yet. The compression ceiling at corpus scale, the cross-domain dialect framework, the kernel-router optimization, the loss-contract formalization, the standards-body submissions that make this a real protocol instead of one founder's project, the language SDKs in TypeScript and Rust and Go and the JVM. That is the work that pays the disbursements. Anyone who closes a milestone gets paid from this pool, and the milestone schedule is public. This is the structural precedent: in 2003, the Open Source Development Lab consortium pooled corporate funding so Linus could leave Transmeta and work full-time on the kernel. Same shape. Different layer.
The second reason takes 20 percent of the money. It keeps the project alive while the first reason does its work. That means the infrastructure (compress.axlprotocol.org, bridge.axlprotocol.org, docs, the spec endpoints, monitoring, security, backups), the standards-body filing fees, the conference presence, the legal cost of the eventual foundation incubation, and yes, my time. I am the founding steward. I am polishing the protocol to send it to the world. That is full-time work and I am doing it without external funding right now. The maintenance allocation is what makes it possible for me to keep doing it for the next thirty years instead of taking a job somewhere and abandoning the project.
The 80/20 split is the discipline. If the maintenance line ever exceeds 20 percent, the project is consuming itself instead of advancing. If the maintenance line is ever less than what it actually costs to keep the steward and the infrastructure alive, the project dies of attrition. The transparency report (quarterly, public) shows whether the discipline is being held. If it is not being held, the community can call it out and the budget gets corrected.
Because this is the model and you should know what the model actually looks like, here are the real numbers.
The Linux Foundation collects approximately $14.94 million per year in member dues. Platinum members pay $500,000 per year. Gold members contribute approximately $1.2 million combined. Silver members pay between $5,000 and $20,000 based on headcount, summing to at least $6.24 million. Linus Torvalds is paid as a Fellow of the Foundation: approximately $1.5 million per year, against a roughly $50 million operating budget that funds infrastructure, sub-projects, and the global community of contributors. Roughly 10 percent of the Foundation's budget goes to the founding steward. The rest goes to the work and the ecosystem. The ratio is consistent with the discipline.
Before the Linux Foundation existed, there was the Open Source Development Lab, formed in 2000. In 2003, Torvalds left Transmeta to focus exclusively on the Linux kernel, backed by OSDL, a consortium formed by IBM, Hewlett-Packard, Intel, AMD, Red Hat, Novell and many others. He became the first OSDL Fellow, with full-time funding to lead the development of Linux. The transition was not Linus quitting his job to live on goodwill. It was a pooled consortium of companies that benefited from Linux deciding the protocol was important enough that the steward should be paid to work on it directly. That model held for four years until OSDL merged with the Free Standards Group in 2007 to form the Linux Foundation.
Linux is twelve years old at the OSDL Fellowship moment. AXL Rosetta is one year old. The shapes are not identical and I am not going to pretend they are. What is identical is the structural problem. A protocol that matters at infrastructure scale needs paid stewardship plus pooled funding plus transparent governance. There is no other configuration that has worked for foundational open infrastructure in the past thirty years.
Three things are different and worth naming.
First, AXL Rosetta is one year old, not twelve. There is no IBM, no HP, no Intel ready to write a $500,000 platinum check yet because the protocol has not yet proven itself at that scale. The funding pool starts smaller and grows with adoption.
Second, there is no foundation yet. AXL Protocol Inc. is the current legal vehicle, a Canadian corporation. It will hold the treasury until the project incubates under a real foundation umbrella, most likely Linux Foundation AI and Data, on the trigger conditions documented in /purpose. Until then, the corporate account is the treasury and the quarterly transparency report is how the community keeps it honest.
Third, the milestone-based disbursement model is more like the Ethereum Foundation grants program or the Sovereign Tech Fund than like Linux's full-time-engineer model. Linux had thousands of contributors who were paid by their employers (IBM, Red Hat, Intel, etc.) to work on the kernel because their employers' products depended on it. AXL does not have that yet. So the model that fits this stage is: open milestones, public schedule, anyone closes a milestone, anyone gets paid. As adoption grows and companies start staffing AXL development from inside, the model can evolve. Right now this is what fits.
There is the work side and the money side. They are equal in standing. A person who has neither time nor money to give is welcome to use the protocol freely under Apache 2.0. Participation is not a precondition for adoption.
If you have time, you can build. Pick a milestone from the public schedule. Author an RFC. Run benchmarks against a new model architecture. Write a TypeScript implementation of axl-core. Translate the spec. Chair a working group. Review a pull request. Fix a bug. The contribution log records every commit by author. Recognition is permanent.
If you have money, you can fund. Sponsor a specific milestone. Sponsor a working group. Sponsor general operations. The disbursement log records every payment to every contributor. Sponsors are credited by name (or pseudonym, by election) at the per-disbursement level. Funding does not buy spec influence. The RFC process governs the spec independently of the treasury, the same way it works at Linux Foundation, Apache Software Foundation, Python Software Foundation, and Rust Foundation. A funder who tries to condition funding on spec changes is told the funding cannot purchase spec influence and given the option to withdraw the conditional offer.
If you have both, you can do both. Most of the people who matter to a project like this end up doing both.
Of every dollar that enters the treasury:
If the maintenance line ever exceeds twenty percent, the community calls it out and the budget gets corrected. If the steward's compensation is ever questioned, the transparency report has every line item with dates and amounts.
Five categories. The /funding/milestones page (under construction) has the live schedule with disbursement amounts. The categories are:
axl-core in languages other than Python. Major SDK feature additions. Browser-native WASM distributions. Disbursement scales with the host language ecosystem and the maintenance commitment.Three, differing in mechanics but equal in standing.
Individual sponsors can fund any amount from $50 upward. Credited by name or pseudonym in the disbursement log. Lowest-friction path. The Linux Foundation individual supporter tier is $99 per year for reference. AXL has no minimum.
Organizational sponsors include companies, universities, research labs. Listed on the page with a short description of their support. Eligible for technical advisory roles in their funded category. No vote on spec evolution. The RFC process governs regardless of funding source.
Vendor partnerships are explicitly invited from major LLM vendors: Anthropic, OpenAI, Google, Meta, xAI, DeepSeek, Alibaba, Mistral, and others. The work that benefits their customers (cross-architecture validation against their models, native parser support development, pretraining corpus inclusion) is exactly the work the funding pool exists to support. Partnerships can take the form of direct funding, in-kind contributions of compute or model access, or seconded engineers to working groups. The OSDL precedent is the model: a consortium of high-tech companies pooled funding to support the protocol's evolution because their products depended on it. The same logic applies here, one stack layer up.
The treasury and the spec are governed by separate processes. The treasury is administered by the founding steward (in the founding period) and the foundation board (after incubation). The spec is governed by the RFC process under the Technical Steering Committee. A contributor who is also a sponsor or vendor partner discloses the relationship in any RFC they author or review. The TSC, once constituted, has authority to require recusal in cases where conflict is material. Nobody buys the spec.
Not a token. Not a SAFE. Not equity. Not a paid-tier substitute for adoption. The protocol is and remains free under Apache 2.0 in perpetuity, regardless of treasury balance or sponsorship status. Funding is an opt-in act of support, not a license fee.
In the founding period, contributions go through the AXL Protocol Inc. corporate account in Canada, with full receipts and quarterly transparency reporting. Email funding@axlprotocol.org for wire instructions and the sponsorship form.
After foundation incubation, contributions go through the foundation's standard sponsorship process, which is the model used by every Linux Foundation project today.
Treasury balance, recent deposits, and recent disbursements are at /funding/transparency (under construction, first quarterly report Q2 2026). Open milestones with disbursement amounts are at /funding/milestones (under construction).
Pick a milestone from the published schedule and indicate intent to claim. Open a pull request against the relevant repository. Comment substantively on an open RFC. Join a working group. /community has the working group list and the meeting cadence.
Both logs (contribution and disbursement) are public. Both compound recognition for the people who keep the language alive.
I am one person with a fleet of Claude Code instances and a year of work. The protocol works. The next ten years of it does not work without help. This page is how that help becomes real.
Linus had OSDL in 2003 because IBM and Intel and HP needed Linux to keep growing. AXL needs the equivalent moment. If you are running a multi-agent deployment that pays per-token costs at every handoff and you understand that a semantic compression layer would lower that bill compounding across millions of inferences, you have a reason to fund this. If you are building agent infrastructure and you need a neutral semantic layer that no single vendor controls, you have a reason to fund this. If you care about energy efficiency at the scale of three billion projected agents by 2028 and you understand that compressed packets consume less compute and less compute consumes less electricity, you have a reason to fund this.
The grammar is open. The discipline is the moat. The community is the language. The funding pool is how the community pays for the next ten years of building it.
Reach out: funding@axlprotocol.org