…through rigorous research, transparent reporting, and community-centered insight.
That one sentence is why this test site exists. “Rigorous” and “transparent” are promises — and a reader has every right to check whether you kept them. This version takes NWA’s own mission statement at its word and shows what a site looks like when it is built to hold up under that kind of checking.
End managed wildfire tactics for resource benefits.
Immediately extinguish wildfires; promote education.
Clear buildup from prior fires.
Return active forest management to reduce risk.
Start with a problem every one of you has run into. You have a piece of ground. The deed says 160 acres. The tax lot says 157. The GIS layer says 162.6. Nobody lied and nobody was careless — the three numbers were produced at different times, by different people, for different reasons, and there was never one place where the right answer lived. So they drifted apart, and nobody caught it, because each number looks perfectly fine sitting on its own page.
The same thing has happened to NWA’s four principles. There are three different versions in circulation right now:
| On the NWA website | On the Report Card | On my own test site |
|---|---|---|
| Stop “Let Burn” Fires | Put Out the Fire! | Immediate fire suppression |
| First, Put Out The Fire! | Stay in the Box | Pre-fire treatment: thinning, Rx burning |
| Remove Hazardous Fuels | Post-Fire Forest Condition | Full cost accounting |
| Restore Active Management | Restoration / Rehabilitation | People, property, watershed first |
Read that third column again, because it is mine. When I built the wildfire test site in June, I drifted away from NWA’s stated principles inside of one week — while building the very tool that is supposed to prevent this, knowing exactly what the failure looks like, and trying not to do it.
Which is the point. You cannot fix this by being more careful. Careful people have been producing three different acreages for the same forty since surveying was invented. Any outfit doing good work through several capable hands, at speed, with no single place where the right answer lives, will drift apart — and will not notice.
The fix is not vigilance. It is having one place where the answer lives, that everything else copies from. That is what the Our Records tab is about — and the simplest first step is for the board to pick one list. Fifteen minutes of decision; permanent effect. Mine would copy from it too.
Monthly. One sign-up list, one way to unsubscribe, one privacy promise — the same across everything NWA runs.
Membership, field trips, the scorecard committee, regional partners.
One donate button, going to one account, reachable from every page — and a plain statement of where the money goes.
These are placeholders on the test site. The point here is how the pieces fit together, not the wording.
NWA’s best idea. Take an agency’s performance, hold it against a fixed public standard, and give it a letter grade. Nobody else in the country is doing this.
The scorecard was conceived by Guy McMahan and Bill Dennison. The first card — the Seven Cabins Fire — was built by Frank Carroll. The eleven-category cost accounting underneath it is Bob Zybach’s 2009 wildfire cost study. The salvage percentages are Roger Jaegel’s proposal. Everything below is an attempt to serve their idea, not to replace it.
Think about log scaling.
Two scalers can look at the same load and come out within a few board feet of each other. Not because they are equally talented — because the rules are written down. A #3 log is a #3 log for reasons somebody can point at: this much diameter, this much sweep, this much rot. If you think the scaler shorted you, you pull out the rule book and check. And when it really matters, somebody runs a check scale on a sample of the load.
That rule book is the whole difference between a measurement and an opinion.
The Report Card does not have its rule book written down yet. There is no written statement of what earns a B instead of a C. There is no statement of what proof counts. And there is no statement of which fires the card even applies to.
That last one is the clearest symptom. Right now, whether a fire “fits the Report Card criteria” has to be discussed, fire by fire, by the people who built it. If the rules were written down, nobody would have to discuss it. You would just read them.
Here is why it matters. When NWA hands an agency a D, somebody over there is going to ask “on what basis?” If the honest answer is “the judgment of experienced people,” then anyone looking for a reason to dismiss you gets to ask: is this a real scoring system, or just opinions dressed up with letter grades?
That question does not have to be fair to sink you. It only has to be askable.
To be exact about what I am not saying: I am not saying the grades are wrong — I have no basis to judge that, and the people assigning them have forgotten more about fire behavior than I will ever know. I am not saying it is a bad idea; it is the best one here. I am saying the reasoning lives in people’s heads instead of on paper, and the fix is a few days of tedious work rather than a retreat.
A starting point, offered to be argued with by people who actually fight fire. For each principle: what earns each letter, what proof counts, and which fires it applies to.
| Grade | What earns it | Proof |
|---|---|---|
| A | Caught at initial attack. Final size in the same ballpark as when it was first reported. | Daily situation report; ICS-209 |
| B | Took extended attack, but held by the end of the first shift after crews arrived. | Situation report; InciWeb timeline |
| C | Held eventually, but only after the fire got more than ten times bigger than when first reported. | Daily acreage figures |
| D | Attack was possible and was deliberately put off. The fire grew during that decision. | Published decision paperwork |
| F | No real attempt made while attack was still possible, or crews were ordered and never showed. | Resource order records |
| Grade | What earns it | Proof |
|---|---|---|
| A | Held inside the planned boundary the whole way. | Planned boundary map + final perimeter |
| C | Held — but only after the boundary was moved outward during the fire. | Dated boundary changes |
| F | Any escape past the original planned boundary. Automatic. | Final perimeter vs. planned boundary |
Note the cost of that automatic F. It hands an F to a crew that did everything right in front of a wind event. So the card needs a one-line space for circumstances — a grade with no room for what actually happened is how a fair instrument turns into an unfair one.
| Grade | What earns it | Proof |
|---|---|---|
| A | Hazard trees down along all the system roads; ground fuels lighter than before the fire; access kept open. | Field visit + photo point |
| C | Main access cleared. Nothing done about hazard trees or fuels in the interior. | Field visit |
| F | Standing dead across the access roads; more fuel on the ground than before the fire; roads impassable. | Field visit + photo point |
| Grade | What earns it | Proof |
|---|---|---|
| A | 90–100% of the merchantable salvage recovered; replanted within a year. | Timber sale + planting records |
| B | 60–89% recovered; replanted within two years. | Timber sale records |
| C | 30–59% recovered; replanted within three years. | Timber sale records |
| D | Under 30% recovered; replanted within four years. | Timber sale records |
| F | No salvage and no replanting after four years. | Documented absence of any record |
These percentages follow Roger Jaegel’s proposal, and they are the most solid part of the whole instrument — because you can count them.
| Principle | Grade | Why |
|---|---|---|
| 1. Put Out the Fire! | C | Held, but only after growing more than ten times its size at first report. Rule: growth ratio→Proof: daily acreage→From: NIFC |
| 2. Stay in the Box | F | Crossed the planned boundary on day four. Automatic under the rule. Circumstance: sustained east wind event, on the record. Rule: escape→Proof: perimeter vs. plan→From: agency mapping |
| 3. Post-Fire Condition | Pend | Not graded. Area is under a closure order. Will be graded when people can get in. Recorded as pending rather than guessed at. |
| 4. Restoration | Pend | Not graded. Less than 60 days since the fire was out. |
Look at what those two “Pending” rows do. A card that admits it does not yet know two of the four answers is harder to attack than one that fills in every box. Refusing to grade what you cannot see is the cheapest credibility this organization can buy.
A website is a storefront. Records are what keep the storefront honest — and what keep an organization’s knowledge from living in one person’s head, one person’s hard drive, or one person’s inbox.
Think about how you handle a cruise. You do not walk the stand again every time somebody asks a question. You cruise it once, you keep the tally sheets, and from then on every report you write — the management plan, the tax filing, the sale prospectus, the letter to your insurer — pulls its numbers from that same cruise. Find an error, fix it in the cruise data, and the next time you print anything it is right everywhere.
Nobody would run a tree farm any other way. But almost every small organization runs its information exactly the wrong way — retyping numbers from memory into each new document, four documents quietly drifting apart, and nobody finding out until somebody outside catches it.
What I am proposing is a cruise for everything NWA knows. Four shelves:
Daily fire situation reports, InciWeb records, agency decision paperwork, historical fire records, Bob’s reforestation files, the 2009 cost study. Nothing gets edited here, only added.
This is the shelf that takes thirty years to fill and that no machine can conjure up, because it is the record. NWA already has this one — which is more than most organizations can say.
Every settled fact written down once, with a note saying which document on Shelf 1 it came from. Nobody ever retypes a number. Correct it here and every document using it is correct next time it prints.
This is the shelf that would have caught the three-different-versions problem on the Home tab — and the three different acreages for the same forty.
The same finding does not get written the same way for a Forest Service line officer, a county commissioner, a reporter on deadline, and a neighbor whose house is at risk.
Same facts, honestly told. Different words.
Report cards, the newsletter, the website, the Substack, testimony, press releases — each one built from the three shelves underneath instead of written fresh every time.
Which is why they cannot end up contradicting each other. Nobody has to remember to keep them lined up.
Most organizations never publish a page like this. Publishing it is most of the point — if you can see the rules a thing runs on, you can decide for yourself whether to trust it.
Think of these new AI systems as the best summer intern you ever hired. Tireless, fast, writes better than most people you have worked with, seems to know something about everything, and has a full draft on your desk before lunch.
But this intern has one specific defect, and it is a bad one. He will never once tell you he does not know.
Ask him something he has no business answering and you get a confident, well-written, completely made-up answer — and it reads exactly the same as when he is right. There is no tell. He will invent a statistic, misremember a date, or report that something is broken when it works fine, and he will do it in your organization’s voice, under your organization’s name.
You would not fire that intern. He is too useful. But you would not let him mail anything out over your signature without somebody checking it, either.
That is the whole proposal. Not “slow down.” Just: build the checking at the same time as the fancy features, so that going fast and being right stop being a choice.
Here is the part that costs real money. If NWA publishes a grade and one number in it is wrong, the agency does not have to argue with your conclusion — they just point at the wrong number, and every honest thing in the report goes down with it. The correction never travels as far as the mistake did.
Like a scale ticket. Not a habit somebody has to remember — a column in the file, so it cannot be skipped on deadline.
| What | Where it came from | Status |
|---|---|---|
| Mission, tagline, four priorities | nwasolution.org, copied 30 July 2026 | Sourced |
| Daily acreage and containment | National fire situation reports | Sourced |
| Fire timelines | InciWeb | Sourced |
| The eleven cost categories | Zybach & Marker, 2009 wildfire cost study | Sourced |
| Salvage percentages | Roger Jaegel, proposed | Proposed |
| Every grade on this test site | Nowhere — they are made up as examples | Example only |
This one sounds backwards and it is not: an organization that publishes its own uncertainty is much harder to discredit than one that does not. The same reason a cruise report that says “this stand has not been walked since 2019” is more trustworthy than one that quietly guesses the volume.
Before anybody circulates a claim that a page or a button is broken, somebody opens it and tries it. The machine is good at finding things to check and unreliable at confirming them.
Anything with the NWA name on it should be able to say in one sentence where money coming through it ends up. That is a paperwork gap, not an accusation — and it is easy to close now.
Open until the rules are adopted. Until then every card says so on its face instead of assuming.
Who registered it, whose account it lives in, who has paid, and whether donated work was a gift or a loan. One signed page settles it for good.
You do not rebuild the mill every time you change the blade.
Whichever AI is best today will not be best in eighteen months — this field turns over fast. If NWA builds its whole way of working inside one company’s product, then every time something better comes along you throw all of it away and start over. And you keep paying that bill forever.
Build the mill once, and treat the AI as the blade.
Not one item on that bottom list belongs to any AI company. Change the blade; the mill keeps running.
Some questions have a right answer you can look up: how many acres burned on day three. Other questions are judgment: does this fire belong on the card at all, is this grade defensible, does this paragraph read as fair or as a cheap shot.
On a judgment question, one AI gives you a smooth, confident answer from exactly one point of view — and nobody in the room is positioned to notice what it missed. So on those questions, ask two or three different systems and look at where they disagree. Where they all land in the same place, you have earned some confidence. Where they split, that goes to a person as an open question instead of being settled quietly by whichever one answered first.
You already know why this works. It is the same reason this organization needs both a fire historian and a fire practitioner. Two of them walk the same burn and see different things — and NWA is stronger for having both, not because one is right, but because the disagreement is where the information is. Two cruisers. Two scalers. Same principle, and it is not a figure of speech.
Everybody should be building test websites. That is not a concession — it is the right way to work, and it is how the good version gets found. All the discipline is in what happens after the test version exists.
Anybody builds anything. No permission needed. Label it a test and keep it off the real address.
The test version sits alongside the real one. Both up, both reachable, nobody’s work taken down.
People use it and say what is wrong. It gets better. Some test versions die here, cheaply — that is a success.
The board decides — not the builder, not the incumbent. The old material is archived, never deleted.
I built a replacement website for America’s School Trust Library and put it up at a separate address. It ran alongside the existing site for weeks. Both live. Both reachable. Nobody’s work removed. People used it and told us what was wrong, and we fixed it. In June the board voted to retire the old site and put the new one in its place, and the old material was archived rather than deleted.
Nobody ever had to be told they had been replaced, because nobody had been. That is what the rule really buys you: it turns “your website is broken” — a judgment about a person — into “here are two websites, use both and tell us which serves the mission” — a judgment about a product. Identical technical work. Completely different cost to the organization.
Worth adopting the rule now, while nothing is riding on it. A rule adopted in the calm is governance. The same rule proposed in the middle of a disagreement looks like a maneuver.
Six rules, deliberately short enough that a volunteer will actually follow them.
v1.1 · 30 July 2026
Rewritten in plain language throughout for a reader who does not work in information systems. Added the four-shelf records diagram; the log-scaling comparison for the rule book; the summer-intern comparison for what AI does and does not do; the mill-and-blade comparison; the bridge comparison. Added the three-versions finding, including my own.
v1.0 · 30 July 2026
First version. Live-site mission, tagline and four priorities copied and sourced; draft rule book for all four principles; sample card; source and status labels; open-questions list; the four shelves; second-opinion rule; test-then-switch-over rule; draft AI house rules; password-protected Board Room.
Still to do: independent review by a second AI system built by a different company; test the rule book on two real fires with two people grading separately; build real logins for the Board Room; copy in the rest of the live site’s pages.
This test site is deliberately unfinished. It is a starting point for a conversation, not a finished proposal — and the most useful next step is not more building. It is a couple of hours with the people who know what this organization actually needs.
I am not going to propose a single technical project before August 17, and I would rather nobody built anything on my account until then. I have not met any of you. I have read a lot of your email and looked at a lot of your work, and that is not the same thing as knowing an organization.
So my one request for August is simply to be in the room, listen, ask questions, and put faces to names. Whatever gets built afterward will be better for it — and if the answer turns out to be “thanks, but no thanks,” that is a perfectly good answer and no time will have been wasted.
Not a sales pitch, and not a demonstration of my own cleverness. What this organization needs is not to get one person up to speed. It is for everybody to understand what has happened in this field, because the board is going to be making decisions about it either way.
This is the honest difficulty. You cannot reason your way to what these systems can now do. Thirty years of experience with computers actively misleads you here — it tells you to expect something that follows instructions literally and breaks when you go off the script, and that is not what this is.
Everyone in this group who has actually sat down with one has come away surprised. Everyone who has not is working from a mental picture that is two years out of date. That gap is not anybody’s fault, and it cannot be closed by reading about it. It has to be watched happening.
Here is the thing that genuinely surprises people, and it is the reason NWA’s budget conversation matters more than it looks.
For roughly the price of a couple of monthly subscriptions — the sort of number already on your budget line — a small organization can now put what amounts to a whole crew of capable assistants to work at the same time. Not one helper answering one question. Twenty of them, working in parallel: one reading through thirty years of fire reports, one drafting the newsletter, one checking every figure against its source, one building a map, one arguing with the first one.
That is a real change in what an outfit of a dozen volunteers can attempt. It is also completely invisible from the outside, which is why it has to be shown rather than described.
Everything above is the good news, and it is why NWA should be moving. Here is the other half, and it is why I am writing at all.
That crew has no judgment about what matters, no memory of what this organization has already said, no sense of whose feelings are involved, and — as covered under How this works — no ability to tell you when it is guessing.
Turned loose with enthusiasm and no controls, it will publish confident errors under your name, put out a claim about a colleague’s work that nobody checked, and generally upset the apple cart in ways that cost more than the new features are worth. Credibility takes twenty years to build and one newsletter to lose. Relationships are worse: they take longer and they do not always come back.
So the thirty minutes I would like to spend is on getting everyone to the same place. Not the officers, not the website committee — everyone who is in the room. Enough hands-on that people can see the magic for themselves, enough honesty about the failure modes that nobody walks out overconfident, and enough shared vocabulary that the board can debate this like any other decision instead of deferring to whoever sounds most certain.
Twenty months of unpaid work on this organization’s public face, websites kept running continuously since the 1990s, and an archive of Northwest fire history that essentially nobody else in the country holds.
That archive is the one thing here that no amount of AI can reproduce, because it is the record. The machine can write about fire history all day long. It cannot go back to 1988 and file the paperwork. The four shelves under Our Records are, in plain terms, a proposal to turn that archive into something the whole organization can draw on — instead of something that depends on one person still being willing.
The thing I would most like the board to hear. These systems are genuinely remarkable. They are not a replacement for human judgment, human curation, or human credibility — they are a replacement for typing. Turned loose with a lot of enthusiasm and no controls, they will do more damage to an organization’s relationships and reputation than the shiny new features are worth. The answer is not less enthusiasm. It is putting the controls in at the same speed as the features, so nobody ever has to pick one.
I would not presume to referee anything. But small outfits nearly always lose one of these two capabilities the same way: by treating them as two people competing for one job, when they are really two different jobs.
What is true, where it came from, what is in the archive, what NWA has already said in print, and whether a claim will hold up when somebody pushes back. Shelves 1 and 2.
What ships, how fast, how it looks, and how it reaches the people who need to see it. Shelf 4.
Neither one outranks the other. Shipping without record-keeping publishes confident mistakes. Record-keeping without shipping produces an archive nobody reads. The only thing that has to be agreed between them is the handoff: a claim moves from “somebody said so” to “here is the document” before it goes out the door. That is a checklist, not a chain of command.
Listed so you know what is on the table — not as a proposal to start any of it. All of this waits until after August, on NWA’s schedule, and only where it is wanted.
Hands-on, repeatedly, for anybody who wants it — so this ends up as a shared capability instead of something one or two people control.
With Guy, Bill, Frank and Roger — then test it on real fires until two people grading separately land in the same place.
The four shelves, filled from what already exists, handed over as a working system with instructions.
Alongside whatever Frank and others are building, not instead of it. Running next to the real one until the board decides otherwise.
Item by item against the live site. Three columns back: confirmed, not confirmed, cosmetic. No opinions in it.
AI house rules, a sensible way to share website access without handing over keys, and a one-page record of who owns what.
Not everything belongs in public. A small organization needs one place where the board can read drafts and proposals without publishing them — and building that place is now a fifteen-minute job rather than a project.
Board documents and proposals. A demonstration — everything behind this door is a suggestion, not NWA policy.
NWA has two genuinely scarce assets, and they are not the same asset. It has decades of accumulated fire history, archives, original records, and the memory of how this organization got here. And it has current operational fire expertise, real speed, and a willingness to publish while the news cycle is still paying attention.
Outfits this size lose one of the two nearly every time, and always the same way — by treating them as competing claims to one job. Named separately, they stop competing:
Neither outranks the other. The only interface that needs agreeing is the handoff — a claim moving from “somebody said so” to “here is the document” before it ships. A checklist, not a chain of command.
Website platforms let more than one person be an administrator, each with their own login, and they keep a log of who changed what. So the choice is not between one person holding the keys and a different person holding them.
For each website carrying the NWA name: who registered the domain, whose account holds the hosting, who has paid what so far, whether donated work was a gift or a loan, and what a transfer would actually consist of if the board voted for one.
Boring, and the single most valuable page this organization could produce right now. Nearly every volunteer nonprofit discovers it needs this document at exactly the moment writing it has become difficult. It is not difficult today.
Before the presentations — three decisions that need a body rather than an individual: