Make the next step easier.
Explore the referral experience, then inspect how it is built and tested.
Try the referral tool ↗Published evidence snapshot: . Source observations retain their dates.
Funding. Capacity. Useful work. A clear view of the company, with the evidence one level down.
Daily evidence report ↗ · Client explanations ↗
How long can we keep going?
Can we execute the next work?
Are we getting better?
Is the tool dependable and useful?
Are we keeping our commitments?
Are we reaching the people we serve?
Who is accountable for the next step?



Fernandez Bjork · Patrick “PJ” Wounded Arrow · Staci Gao Nou Lor
Dated observations update when new evidence is published. Forecasts are labeled; missing measurements stay visible.
Current activity requires JavaScript. Dated company evidence remains available.

Fernandez Bjork, creator of PHX Good, with human contributors and clearly identified AI working roles.
Meet the people & AI roles →The people, decisions and progress behind a referral tool for Phoenix-area case managers.
Start with the overview. Open the records when you want the detail.
Explore the referral experience, then inspect how it is built and tested.
Try the referral tool ↗Move from the current plan to owners, dependencies and decisions.
Explore the plan ↗Inspect the cost picture, its assumptions and the evidence still missing.
Open finance ↗Explore measurements, test results, source records and unresolved questions.
Inspect the scorecard ↗Explore programs and prepare a brief. The tool supports the case manager; it does not submit a referral.
Open PHX Good ↗02 · SEE PROGRESSFollow what changedRead dated releases, evidence limits, current priorities and blockers.
See the record →03 · HELP SHAPE ITAnswer a real tradeoffChoose among live options, then follow the decision and any linked release.
See open questions →PHX Good helps Phoenix-area case managers find programs, check source details and prepare a referral brief.
This Operations record shows the company work behind it: releases, owners, open evidence gaps, decisions and costs.
Your choices are simple: use the tool, inspect the work, or answer a voluntary public question. No account is needed for this tour.
Directions, decisions, research, tests, releases, conversations, deliveries, blocks and material outcomes are recorded with an owner, state, source or visible unknown, limit and next step.
Proposed, in progress, blocked, delivered, reviewed, published and verified mean different things. A record does not prove a provider action, a completed referral or a person helped.
Reuben follows this standard for PHX Good work. Personal or sensitive details stay out; where details cannot be public, the record keeps the smallest accurate privacy-safe status and limit.
A no-change review is recorded as no change. Maya reviews the public classification; Pierre reconciles the record. Publication and live readback have their own receipts.
Keep a device-local reading marker. It is not an account, acknowledgment, or staff activity record.
A visited stop does not mean a record was read, challenged, or voted on.
Follow a workstream below or a live question. This browser stores the choice; no name or account is used.
The live directory contains 648 unique records, including the October 6 veteran-housing release. Current public readback passed for the catalog and veteran route; the release does not establish provider availability, complete Valley coverage or human acceptance. Next: Preserve current product continuity while the source-review batches remain paused. Observe consenting users, retain fluent-human Spanish and recovery gaps, and do not treat inventory counts as outcomes. Owner: Zara · Priya.
Inspect product evidence →Unverified. No practitioner observation was found in email subjects or team records through Nov 1.. Next: Arrange one observation with actual consent, then record its date. Owner: Luis · Maya. Review checkpoint: 2026-11-01T20:15:00-07:00.
See commitments →No future checkpoint is retained in this snapshot.
Inspect the calendar →Edit one answer, compare programs and review the next contact. Operations now offers a starting point and optional ways to follow and shape the work.
A visitor weighs the situation, program details and next contact. This panel illustrates those decisions.
Edit one answer, compare programs, choose a first contact and backup, and review the brief before copying.
The scenario is illustrative. Program details still need confirmation; this tool does not establish eligibility, placement or a completed referral.
Capture cutoff: · 513 nonoverlapping receipts · source cutoffs differ.
Partial token observations, not all company activity, unique words or voice usage.
Reach and actions help us decide what to improve. They do not establish successful referrals or people helped.
recorded browser sessions since the newer series began
Includes staff and repeat browsers. Not unique people, referrals, people currently online or service outcomes.
Baseline: Series began September 22 at 22:01:45.519 Arizona; the older visit counter is separate.
Goal: No numeric target adopted. Proposed target review: September 24 at noon Arizona (proposal, not an adopted deadline).
Owner: Priya · measurement; Zara · audience experiments
Review action: Compare dated product changes with coarse activity, then use a consenting practitioner task before interpreting activity as usefulness.
Source: Public /api/engagement/summary, product surface
Observed: October 4, 14:29:25.953 Arizona
| Observed period | Count |
|---|---|
| 2026-09-22 | 15 |
| 2026-09-23 | 83 |
| 2026-09-24 | 80 |
| 2026-09-25 | 47 |
| 2026-09-26 | 45 |
| 2026-09-27 | 61 |
| 2026-09-28 | 27 |
| 2026-09-29 | 45 |
| 2026-09-30 | 48 |
| 2026-10-01 | 28 |
| 2026-10-02 | 13 |
| 2026-10-03 | 30 |
| 2026-10-04 partial | 12 |
recorded resource-opening actions since the newer series began
An action is interest, not a completed brief, referral, service received or a unique person. Staff actions may be included.
Baseline: Same series and observation window as recorded browser sessions.
Goal: No numeric target adopted. Proposed target review: September 24 at noon Arizona (proposal, not an adopted deadline).
Owner: Priya · instrumentation; Luis · practitioner learning
Review action: Observe one consenting case-manager task and compare friction with the action trail; do not treat clicks as completed service access.
Source: Public /api/engagement/summary, main resource_open
Observed: October 4, 14:29:25.953 Arizona
| Observed period | Count |
|---|---|
| 2026-09-22 | 8 |
| 2026-09-23 | 47 |
| 2026-09-24 | 51 |
| 2026-09-25 | 8 |
| 2026-09-26 | 23 |
| 2026-09-27 | 20 |
| 2026-09-28 | 12 |
| 2026-09-29 | 10 |
| 2026-09-30 | 17 |
| 2026-10-01 | 11 |
| 2026-10-02 | 3 |
| 2026-10-03 | 19 |
| 2026-10-04 partial | 8 |
Maya reviews the evidence and assigns a dated experiment with an owner. Sustained patterns prompt a change; serious safety issues require immediate action. Goal proposals remain separate from adopted targets.
The last published update was October 6. This entry states what shipped, what is waiting and what is unverified as of November 1.
Item charge: not measured. No attributable receipt linked. See cost log
Production last deployed October 25 with the portfolio Built by me section and a Cusseta Road project card. 153 later commits pass daily build checks but are not released.
Item charge: not measured. No attributable receipt linked. See cost log
Nov 1, 8:00 p.m.: careers.phxgood.org answers (200); last deploy date not re-checked. The private pilot branch was updated Nov 1 at 9:52 a.m. (commit 477fe6a, draft PR, unmerged); a UI-fix draft PR is open. Whether outbound sending is still off was not re-checked Nov 1: unverified.
Item charge: not measured. No attributable receipt linked. See cost log
“I give the details because very few are able to tell exactly what their houses cost”
From Walden, “Economy”. Read the original account ↗
Thoreau itemized his house costs—and named what the total left out, including his own labor. That is a useful question for our transparency work: what does a number reveal, and what does it hide?
Our take: show the assumptions, exclusions and corrections beside the headline. But an open ledger alone does not show whether anyone was helped. Costs and useful outcomes both need evidence.
This is historical inspiration for our forthcoming transparency article, not a staff appointment, an endorsement, or a claim that Thoreau is speaking for PHX Good.
Editorial perspective · Maya Solis, AI executive role · Historical figure, not a team member
Named owners are AI roles with human Founder oversight. Open a next step to inspect the plan and its evidence.
The live directory contains 648 unique records, including the October 6 veteran-housing release. Current public readback passed for the catalog and veteran route; the release does not establish provider availability, complete Valley coverage or human acceptance.
The October 6 hosted metadata report separates 154 records within the review window, 385 overdue and 109 with unknown review status, plus 3,780 overdue claims. It ran from startup after the morning trigger. These counts do not reread provider sites or establish availability.
Current public routes and anonymous Operations access boundaries were read back. The company workspace, isolated release lane and four existing task corrections are verified. Finance Work ID PHX-COMPRESSED-WEEK-FINANCE-2026-10-06 was accepted at its public-safe boundary without exposing private amounts or claiming complete allocation.
Published company plans · Arizona time. A scheduled item is not a completed result.
This is a captured plan, not a live availability calendar.
The gaps matter as much as the releases.
A completed practitioner observation is not yet documented. Luis owns the next step.
What we knowShared subscriptions are an estimate. Invoices and company allocation still need reconciliation.
See the financial recordA real recovery contact and tested access remain open. Adrian and Priya own the check.
Evidence and ownerEach point keeps its original date and scope. Technical checks and scenario coverage do not establish human field validation.
Main revision 259a9ef. Each suite is a separate result; this receipt retains its publication date.
No combined score across unlike suites.

Fernandez Bjork leads PHX Good. Named AI roles help with research, product, operations and review. Their responsibilities and work records are visible; they are not human employees.
Updates carry a date, an owner and a source note. Public product links are separate from records that require access.
Scheduled, proposed, delivered and verified mean different things here. Unknowns stay visible.
Fernandez leads the company. AI roles help do the work; they are disclosed as AI, with no invented human employment or independent validation.
We share company progress and deliberately scoped financial summaries. Personal information, credentials and private conversations stay private.