Skip to main content

Writing an ALIGN.md

This guide covers practical advice for authoring an effective ALIGN.md. The spec defines the format. This page helps you write one that's honest, specific, and useful.

Start from the Dealbreakers​

Write your Dealbreakers section first. What would kill a partnership six months in? What has killed partnerships in the past? What do you know about yourself that you usually don't say out loud until trust is built?

Work backwards from there. Every other section should support or contextualize the Dealbreakers. If your Values section doesn't explain why something is a dealbreaker, one of them is wrong.

This is counterintuitive. Most people want to start with Identity or Mission because those feel positive and energizing. But the Dealbreakers section is where the real value of ALIGN.md lives. It's the section that saves everyone time.

Section-by-Section Guide​

Identity​

One paragraph. Facts, not framing. State what you do, how long you've been doing it, and what stage you're at.

Bad:

We are a revolutionary AI education platform transforming the future of learning.

Good:

Applied AI Society is a community and educational organization making the world
applied AI literate. Founded in 2025. Pre-revenue, community-funded. Based in
Austin, TX.

The bad version is marketing. The good version is information. An agent evaluating alignment needs information.

Mission​

State the goal in terms that could be measured. If your mission statement could belong to any organization in your space, it's too generic.

Bad:

Our mission is to empower individuals through AI education.

Good:

Make the world applied AI literate. North Star: number of people certified who
can demonstrate applied AI competence in their domain.

The North Star metric is optional but valuable. It tells a potential partner what you're actually optimizing for.

Values​

Each value should have a cost. If living this value never requires you to say no to something attractive, it's not a real value. It's a talking point.

Bad:

- We value innovation
- We believe in collaboration
- We prioritize excellence

Good:

- Sovereignty over convenience. We will not build systems that create
  dependency on any single vendor, including us.
- Signal over noise. We'd rather publish one useful thing than ten
  impressive-sounding things. This means we say no to most content
  opportunities.

Test each value: "Can I name a specific time this cost us something?" If yes, it's real. If no, rewrite it.

Capabilities​

List what you can deliver today. Separate current capabilities from things in development. An agent evaluating "can this partner deliver what we need?" needs to know the difference.

Bad:

- World-class AI education
- Cutting-edge research
- Global community

Good:

- Applied AI curriculum covering agents, context engineering, and harness design
- 500+ active community members across 8 university chapters
- Event infrastructure (Applied AI Live series, hackathons)
- In development: certification program, corporate training pilot

The bad version could describe any AI education company. The good version describes exactly one organization and exactly what it can do right now.

Looking For​

Describe partner archetypes, not generic relationship types. "Strategic partners" is meaningless. "University departments that want to co-develop applied AI curriculum" is an archetype an agent can match against.

Bad:

- Strategic partners
- Investors
- Thought leaders

Good:

- University partners who want to bring applied AI literacy to their students
- Community partners who run meetups, hackathons, or study groups
- Sovereignty-aligned sponsors (no hyperscaler dependency, open source friendly)

Dealbreakers​

This is the hardest section to write and the most important. Most people instinctively soften their dealbreakers because stating them feels confrontational. Resist that instinct.

The 6 Months Test: Think about every partnership, collaboration, or relationship that went wrong. What went wrong? When did you first notice? Could you have known earlier if you'd asked directly?

Those answers are your dealbreakers.

Bad:

- We prefer partners who share our values
- Misalignment on goals

Good:

- Partnerships that require exclusive vendor lock-in
- Organizations that gate foundational knowledge behind enterprise pricing
- Partners who want brand control or co-branding approval rights over our content
- Relationships where we'd be a marketing channel rather than a genuine collaborator

Tips for writing dealbreakers:

  • Be specific about the behavior, not the character. "Dishonest people" is a character judgment. "Organizations that misrepresent their AI capabilities to clients" is a behavior you can observe.
  • Include structural dealbreakers. Some misalignment isn't about values. It's about structure. "Partners who require 90-day payment terms" is structural. "Organizations without a technical co-founder" is structural. These matter.
  • Include the ones that feel awkward to state. If you have spiritual convictions that affect partnership decisions, say so. If you won't work with certain industries, say so. If you have strong opinions about specific technologies or business models, say so. The point of this section is to be upfront about the things that usually stay unsaid until they cause problems.
  • Update this section when partnerships fail. Every failed partnership teaches you a new dealbreaker. Add it.

Current Priorities​

This is the only section that should change frequently. Update it quarterly. Include what you're focused on, what you're not focused on, and what would make a new partnership especially timely.

Bad:

- Growing the organization
- Expanding our reach

Good:

# Current Priorities (Q2 2026)

- Landing first university keystone partnership
- Launching corporate upskilling pilot
- Austin as Applied AI capital initiative
- NOT currently focused on: international expansion, venture funding

The date in the header matters. An agent reading this in Q4 should know it's stale. The "NOT currently focused on" items prevent wasted outreach.

How to Engage​

Make the first step obvious. Don't send people to a generic contact page. Give them the specific email, form, or channel that reaches the right person.

Bad:

Visit our website for more information.

Good:

Email gary@appliedaisociety.org or DM @AppliedAISoc on X.
If you publish an ALIGN.md, include the link. We'll run bilateral evaluation.

The last line is powerful. It tells potential partners that publishing their own ALIGN.md is the best way to get a serious conversation.

The Union Principle: Personal ALIGN.md Files​

If you run multiple organizations or wear multiple hats, your personal ALIGN.md synthesizes all of them. This is harder than writing an org-level ALIGN.md because you need to show how the pieces fit together.

Guidelines:

  • Lead with the synthesis, not the list. Your Identity section should describe how your roles connect, not list them separately.
  • Surface tensions honestly. If your ministry work and your business have different priorities, that's useful information. Don't hide it.
  • Show the hierarchy. Which commitments take priority? If your faith comes first and your business comes second, say so. Partners need to know what they're working with.
  • Reference the org-level files. Link to each organization's ALIGN.md if they exist. Your personal file is the synthesis. The org files are the details.

Bad:

# Identity

I am the co-steward of AAS. I also co-founded Imagos Labs. I also have a
faith blog.

Good:

# Identity

Gary Sheng. Co-steward of Applied AI Society (making the world applied AI
literate), co-founder of Imagos Labs (cultural infrastructure at the
intersection of AI, culture, and sovereignty), and author of FaithWalk OS
(a documented Christian worldview that shapes every decision). These aren't
side projects. They're the same mission from different angles: shepherding
people through the Great Transition with agency, dignity, and truth intact.

Common Mistakes​

Making It a Pitch Deck​

If your ALIGN.md reads like it was written to impress, start over. The audience is an agent evaluating fit, not an investor evaluating upside. Include the things that would make some partners say "this isn't for us." That's the point.

Omitting Dealbreakers​

The most common mistake. People leave out dealbreakers because stating them feels aggressive or limiting. But unstated dealbreakers don't disappear. They just surface later, after time and trust are invested. Every dealbreaker you omit is a future conversation you're postponing.

Being Too Generic​

If you replaced your organization's name with a competitor's and the ALIGN.md still made sense, it's too generic. Every section should contain information specific to you. Values like "integrity" and "excellence" tell an agent nothing because every organization claims them.

Not Updating Current Priorities​

An ALIGN.md with outdated Current Priorities is actively misleading. Partners will reach out about things you're no longer focused on. If you can't commit to quarterly updates, omit the section entirely rather than leaving stale information.

Writing for Humans Instead of Agents​

ALIGN.md is designed for agents to parse. This means: clear headers, bullet points, short sentences, concrete language. Don't write flowing prose that requires interpretation. Don't embed important information in the middle of paragraphs. Structure everything so an agent can extract and compare it section by section.

Keeping It Current​

Set a quarterly reminder to review your ALIGN.md. The review process:

  1. Current Priorities. Update the quarter, add new focuses, remove completed ones.
  2. Capabilities. Did you ship something new? Move items from "in development" to the main list.
  3. Dealbreakers. Did a partnership fail? Add the lesson.
  4. Looking For. Have your needs changed? Update the archetypes.
  5. Values. These should change rarely. If they're changing quarterly, they weren't values.

The identity, mission, and values sections should be stable. If you're rewriting them every quarter, the problem isn't ALIGN.md. The problem is that you haven't clarified those things for yourself yet.

Checklist Before Publishing​

  • Does the Dealbreakers section include at least 3 specific, testable items?
  • Does every value have a cost (something you'd say no to)?
  • Does Capabilities separate current from in-development?
  • Does Current Priorities include the quarter and year?
  • Is it free of marketing language, superlatives, and pitch-deck energy?
  • Would you be comfortable if your most skeptical potential partner read it?
  • Can an agent cross-reference your sections against another ALIGN.md?