How the count is kept honest

survey.csv gives the day an answer arrived, never the time, and nothing the person said about themselves (age band, sex, coverage, region, state, stage, who answered, the ages at onset and diagnosis). One row carrying all of that could point at a person, so those answers are published only as counts, at /api/survey, and any answer fewer than 11 people gave is not shown there.

gap.csv gives the day, never the time.

This is the whole method, the same code we run:

# Check our count without trusting us. Python 3, no packages.
import csv, json, hashlib, re, urllib.request

url = "https://waypointledger.org/api/export/corrections.csv"
req = urllib.request.Request(url, headers={"User-Agent": "Mozilla/5.0"})  # the site's edge refuses Python's default agent
lines = urllib.request.urlopen(req).read().decode().splitlines()
rows = list(csv.DictReader(l for l in lines if not l.startswith("#")))  # line 1 is a comment: the table version

def cell(s):    # a cell a spreadsheet would execute is stored with a leading
    if s[:1] == "'" and re.match(r"[=+\-@\t\r]", s[1:2]):   # apostrophe: strip it
        return s[1:]
    return s

def num(s):
    if s == "": return None
    return int(s) if "." not in s else float(s)

prev = "0" * 64
for r in rows:
    fields = {
        "received_at":   cell(r["received_at"]),
        "price_id":      cell(r["price_id"]),
        "verdict":       cell(r["verdict"]),
        "believed_usd":  num(r["believed_usd"]),
        "table_version": cell(r["table_version"]) or None,
    }
    canonical = json.dumps(fields, sort_keys=True, separators=(",", ":"))
    prev = hashlib.sha256((prev + canonical).encode()).hexdigest()
    assert prev == r["row_hash"], f'row {r["received_at"]} does not match'

print("head:", prev)  # compare with /api/integrity

One detail the code above handles: a cell that would otherwise begin with =, +, -, @, a tab or a carriage return is stored with a leading apostrophe, so a spreadsheet reads it as text instead of running it. Strip that one apostrophe before you hash. It is the only thing we do to a published value, and it is reversible.

The first prev_hash is 64 zeros, every later one is the row_hash of the row above, and the last row_hash and the number of rows match the head published above. That shows no row was dropped, added or moved since the head you compare with.

It cannot show that a published answer matches what was hashed; for that we walk every chain over the stored rows ourselves at /api/integrity?verify=1 and print the result, whether or not it matches.

# survey.csv and gap.csv: check the links row by row. Python 3, no packages.
import csv, urllib.request

for name, table in (("survey", "survey_responses"), ("gap", "gap_reports")):
    url = f"https://waypointledger.org/api/export/{name}.csv"
    req = urllib.request.Request(url, headers={"User-Agent": "Mozilla/5.0"})  # the edge refuses Python's default agent
    lines = urllib.request.urlopen(req).read().decode().splitlines()
    rows = list(csv.DictReader(l for l in lines if not l.startswith("#")))
    prev = "0" * 64
    for n, r in enumerate(rows, 1):
        assert r["prev_hash"] == prev, f"{name} row {n} does not link to the row above"
        prev = r["row_hash"]
    print(name, len(rows), "rows, head:", prev, "- compare with", table, "at /api/integrity")