handprint
user stamp
or chosen 1:1 photo
drop image →
fingerprint
ssh-ed25519 AAAAC3NzaC1lZDI1
NTE5AAAAINx7…q9P · ✓ verified
@cameron · she/her

Cameron Ortiz

Staff infrastructure engineer in San Francisco. A decade on auth, payments, and the seams between them. Currently shaping the auth-service-v2 rewrite — opinions and receipts.

auth-systemsedge-computedistributed-systemsdatabase-migrationspaymentsincident-responsejwtcloudflare-workerspostgres
calibration · 12-month weighted
0.74
117 resolved · 67 open · 184 total
validated63
partial28
revised16
invalidated10
honesty signal · strong
resolution rate 96% · learning trajectory ↗ · 10 invalidated outcomes logged honestly
184
handprints
since 02 feb 2026
117
resolved
96% resolution rate
42
attestations
+6 last 90d
P6M
median horizon
longest active · P18M
31
day streak
personal best · 47
0.83
y/y trend
+0.09 ↗ since Q1

twelve months of thinking

26w · 7d · open / resolved
may '25augnovfeb '26may '26
less
moreopen horizon

featured · most cited handprint

hpr_01J7A2B3C4 · 31 attestations
overridecontext · payments reliability·P12M horizon · cursor + claude·2025-08-14T17:02:00Z

“Idempotency keys belong at the edge, not the application. Move them into the worker layer and every retry storm becomes a non-event.”

intent · the call

Overriding the consensus to put idempotency in the application service. Pushing it to Cloudflare Workers with a 24h key store. Betting the storage cost stays under $400/month at 10× current volume, and that we eliminate at least 80% of duplicate-charge incidents within a quarter.

risk · what could go wrong

If key-store consistency lags, we double-charge customers — and the failure mode is silent until support tickets pile up. Mitigation: shadow-mode for two weeks against the existing service; alert on key-store skew above 50ms.

resolution · validated · resolved 2025-11-09

Right call. Eliminated 94% of duplicate-charge incidents in Q4 — better than the bet. Storage cost peaked at $312/month under Black Friday load (2.3× higher than predicted; the bet would still have held). One thing I underestimated: the debugging story for “missing” keys turned out harder than expected because the key-store doesn’t surface in app logs. Spent ~15 hours on observability we should have built up front.

observability-before-optimizationedge-state-needs-shadow-modekey-stores-want-tracing
anchors · 4 verified
git:7c8e2f1pr:1240deploy:prod-2025-09-02incident:INC-441 (closed)slack:#payments-arch
attestations · 4 peers
j@jmartinezwitness
r@rkimcontext
p@priyaimpact
d@dparkimpact
sha256:
f4a8b2c1d3e5f7a9b0c2d4e6f8a1b3c5
d7e9f0a2b4c6d8e0f1a3b5c7d9e1f2


ct-log: entry #1,127,041
signed: ssh-ed25519 · ✓

timeline

↓ newest · 184 total
may · 2026
1109:14 UTC
overridecontext · auth-service-v2 · cursor
“Edge JWT validation on Cloudflare Workers over a centralized auth gateway. Betting token size stays manageable and we won’t need real-time revocation for 12 months.”
git:3fa9e2dpr:847jira:AUTH-219
OPEN
due 2026-11-11
0 attestations
0714:40 UTC
wagercontext · query planner v3 · claude code
“p99 stays under 80ms with the new planner for 12 months under 4× current load.”
git:c481a9bdesign-doc:planner-v3
OPEN
due 2027-05-07
2 attestations
0211:22 UTC
constraintcontext · billing service · terminal
“No third-party auth providers in the billing code path. Ever. We’ve paid for that mistake twice.”
git:91d2f44postmortem:PM-2024-08
OPEN · ongoing
no horizon
3 attestations
apr · 2026
2816:08 UTC
rejectioncontext · v2 scope review · claude
“Declining the recommendations engine for v2. Wrong stage; we don’t have the signal yet. Cost is years of mediocre output we can’t unwind.”
notion:v2-scopeslack:#product-leads
VALIDATED
resolved 14d early
4 attestations
1410:00 UTC
directioncontext · observability stack · cursor
“Going with OTel everywhere, exporting to a single backend. Vendor decisions are easier to reverse than schema decisions.”
git:5b3e7a2pr:823
PARTIAL · 0.6
resolved 2026-04-30
2 attestations
0622:31 UTC
overridecontext · cache layer · cursor
“Rejecting Redis for caching. Building our own LRU in front of Postgres with PL/pgSQL. Worth the simplicity tax.”
git:af2c194pr:801 (reverted)
INVALIDATED
resolved 2026-05-02
2 attestations
mar · 2026
2209:48 UTC
constraintcontext · data retention · terminal
“Retention caps at 90 days for raw event data. Anything older lives in cold storage with explicit access reviews.”
git:6d9e213compliance-doc:SOC2-7
VALIDATED
resolved 2026-04-12
3 attestations
1413:15 UTC
wagercontext · pricing v4 · claude
“Usage-based pricing converts the top quartile at 2.4× the seat-based plan within 6 months.”
notion:pricing-v4analytics:cohort-2026-q1
PARTIAL · 0.6
resolved 2026-05-09
5 attestations
0315:55 UTC
directioncontext · auth-service-v2 kickoff · cursor
“Going stateless wherever possible. Every piece of server-side state is a quarter of incident-response debt.”
git:2a8f001design-doc:auth-v2
VALIDATED
resolved 2026-05-08
4 attestations
feb · 2026
1917:40 UTC
rejectioncontext · vendor proposal · manual
“Declining the managed auth offer. We’d save 6 weeks of build and inherit a 5-year integration tax we can’t model.”
slack:#leadershipvendor-call:2026-02-18
VALIDATED
resolved 2026-04-21
2 attestations
0810:11 UTC
overridecontext · migration strategy · cursor
“Big-bang cutover for the auth migration. Six weeks of dual-running invites more bugs than it prevents.”
git:8e7c52apr:712
REVISED · 0.3
resolved 2026-03-22
4 attestations
load 172 earlier handprints →