Fernandez Bjork

PHX Good: referrals for Valley case managers, built in the open. Human-led, AI-assisted, human board forming. Army NCO. Watch live:

[Fernandez Bjork @fdbjork]

Every building on record in Phoenix, 1880 to 2020. 1,117,055 lights.

That same January, 7,419 people were counted without a home. There were 4,006 shelter beds.

60 seconds. All public data.

[Fernandez Bjork @fdbjork]

Launch is tomorrow.

Shipped at 1:04 a.m.: a shelter listing that pointed to a centralized line current regional sources say no longer takes requests.

Today: update the Operations snapshot and close the remaining route-caution gap.

Still open: no observed case-manager session.

[Fernandez Bjork @fdbjork]

5:30 check: 10/393 program records have overdue review evidence. Dates don’t confirm service availability. Operations is dated Sep 27; poll response prepared, not public. Next: finish authenticated update and review overdue sources. https://t.co/WGgVIb9PxW

[Fernandez Bjork @fdbjork]

PHX Good | Tuesday, two days out.

Today’s targets: refresh the dated Operations snapshot, recover readable source, and run the product review.

Morning check: 2 of 30 conflict-flagged records lack an English or Spanish route caution. Defect found, not fix shipped.

[Fernandez Bjork @fdbjork]

PHX Good · Tuesday, September 29
Two days to our October 1 introduction target. Readiness remains under review.

For case managers, the next step needs to be clear about what is known, what conflicts and what still needs checking.

TODAY'S PRIORITIES
• Make warnings easier to see. This morning's engineering review reported two records missing bilingual route cautions. The repair candidate now passes its production build and 571 tests, with 11 optional tests skipped. It is not deployed; the underlying provider conflicts remain unresolved.
• Finish source work. Last night's review completed 11 of 131 selected claim-level reviews. The remaining 120 carry forward: 100 unfinished comparisons and 20 blocked source reads. No catalog changes were published from that review.
• Use the recovered current source to test repairs and prepare the Operations refresh. Operations still shows September 27. The overdue responses to the closed polls also remain open.

WHAT CHANGED SINCE LAST EVENING'S POST CUTOFF
• The partial catalog review was filed, with four review-pending candidates and one additional candidate held. Proposed does not mean published.
• Internal advisory and management records were filed; they do not establish a human board vote or launch approval.
• Reporting corrections now recognize yesterday's two published briefings, retain their late delivery, and distinguish partial compute evidence from total company cost.
• This morning's engineering check produced a reproducible finding. The technical owner recovered an archive matching the current repository source and prepared and tested the repair. Recovery is complete; deployment is still pending.

LIVE STATE RECHECKED AT 09:00 ARIZONA
The hosted 08:30 evidence-date check ran: 393 catalog records, 383 within their review windows, 10 overdue programs and 71 overdue claims. This checks recorded dates; it does not reread providers or confirm availability. No new product release was verified in this review.

Today's operational checkpoints, Arizona time: product review at 09:00; finance at noon; compute at 16:00; feedback at 16:30; Founder accountability report at 17:00; source review at 18:30. These are commitments, not completed results.

Actual company cost and verified people helped remain unmeasured. We will show useful work without turning activity into a claim about outcomes.

For case managers: https://t.co/FoQ0Fzs2gV
The dated public work record: https://t.co/zho07QR3GN

Reporting window: September 28, 17:46 through September 29, 09:00 Arizona. The hosted report ran at 08:30; the live state was rechecked at 09:00. Coverage is partial; earlier work retains its original date.

[Fernandez Bjork @fdbjork]

PHX Good · September 28 · Evening accountability
Three days to our October 1 introduction target. Readiness remains under review.

The purpose is practical: help Maricopa County case managers find a dependable next step for the people they support. Here is what moved today, what remains unfinished, and what we can measure.

WORK FOR THE PRODUCT
• Reviewed six official source URLs: five were fully readable, one partially readable. Prepared two seasonal reminders. No new service change was established by that review.
• Prepared a correction specification for a food-access referral instruction. Implementation and live verification remain open.
• Prepared seven responses to closed community polls. We missed the 10 a.m. publication deadline; the responses are still unpublished.
• Checked public page content. The hosted 5:30 p.m. evidence-date check also completed: 383 of 393 records are within review windows; 10 programs and 71 individual claims need review. These counts do not confirm service availability.

BACKGROUND WORK
• Published this morning’s briefing and infographic at 9:03.
• Retained the daily plan and original deadlines.
• Filed partial management answers and a later Chair briefing. The launch-recommendation packet is incomplete. This is AI-assisted internal review, not a human board meeting or approval.
• Completed a version-four architecture review covering evidence, recovery and publishing. Implementation is not verified.
• Updated our publishing procedure to keep one scorecard format, reviewed social coverage and prepared the next posts.

These are nine grouped workstreams, with completed, prepared and partial stages shown in the graphic. They are not nine fully closed commitments. Earlier website releases keep their original dates. We have not verified a new live product release today, and restoration of the displaced guides remains open.

THE SAME SIX-AREA SCORECARD
Pilot v2.1, displayed out of 10; evidence cutoff 5:46 p.m. Arizona:
Help — not measured; complete run denominator missing.
Trust — 0.0 mechanical calculation. The fixed penalty for 29 distinct conflict objects overwhelms 97.46% review-window coverage. This flags a scoring limitation; it is not a zero-quality verdict.
Execution — not measured; original due-work inventories and partial acceptance still need reconciliation.
Media — 5.0, one of two originals due by this cutoff published. The midday post was missed. This evening edition is not counted before publication.
Finance — not measured; receipt and bill coverage incomplete.
Compute — not measured as an area score; partial token evidence exists.

Coverage: two of six areas, 40% of the weighting. Overall company grade: not final.

COST AND COMPUTE
Four bounded receipts record 3,952,360 input-plus-output tokens. This is partial usage, not the full company total or a dollar charge.
$16.80/day is a shared-stack budget estimate. Actual daily company cost and the human-equivalent cost ratio remain unverified.

Useful preparation belongs in the record. People helped requires separate evidence; that outcome remains unmeasured.

NEXT
Verify and ship the correction and guide restoration; publish the overdue poll responses; reconcile costs and the full due-work inventory. Public Operations still carries yesterday’s snapshot, and its refresh remains open. Later evening work will retain its actual completion time.

The tool: https://t.co/L2wKyI7c8A
Operations: https://t.co/jjHvSR205s

[Fernandez Bjork @fdbjork]

PHX Good · Monday, September 28
Three days to our October 1 introduction target.

Today's purpose: make the next step more dependable for case managers and the people they support.

PRIORITIES
• Restore the 70 guides displaced by a later release. The repair is prepared; restoration is not yet live.
• Review overdue sources and finish the carried-over correction work. Our live catalog has 393 records. Today's hosted check flags 10 overdue program-review windows and 71 overdue claims. Those are aging evidence dates, not proof that services changed.
• Close the feedback loop: a management response to the closed polls is due at 10 a.m., with actual votes, reasons, owners and next steps.

WHAT HAPPENED SINCE THE EVENING RECORD
The 5 a.m. public check found no live change. The hosted evidence checker ran at 8:30:32 a.m. A morning planning record and seven reviewed poll responses were prepared. Publishing those responses is blocked on administrator access; none is live yet. These are internal outputs; no new product deployment is verified in this morning review.

YESTERDAY'S WORK, CORRECTED
Our corrected record accounts for ten release workstreams, including the website, Operations, portfolio, film, resources and guides. The separate 118-of-131 source-review batch published no changes; that was never the whole company's output. The 70-guide regression remains open.
https://t.co/tRUGtet4l0

TODAY'S CHECKPOINTS (ARIZONA)
10 a.m.: poll-management response.
Noon: management answers and finance evidence/unknowns.
4 p.m.: compute receipts or explicit coverage gaps.
5 p.m.: Founder accountability report and recovery register.
6 p.m.: tentative asynchronous AI advisory review. The Founder retains the launch decision.

STILL OPEN
Guide restoration, publication of reviewed corrections, current Operations records, practitioner evidence and reconciled company cost. The $16.80 daily figure is a shared-stack budget estimate, not actual measured compute. Plans and tests do not establish people helped.

Product: https://t.co/FoQ0Fzs2gV
Company record: https://t.co/zho07QR3GN

Evidence window: previous review cutoff Sep 27 at 10:30 p.m. through Sep 28 at approximately 9:02 a.m. Arizona. Coverage is partial across company tools; earlier work retains its original completion date.

[Fernandez Bjork @fdbjork]

Correction: the earlier recap missed today’s Codex releases. Here is the reconciled company record.

TEN RELEASE WORKSTREAMS
Operations front door; mobile program browsing; homepage clarity; the 18-second film; resource updates; portfolio, essay and film; Ops charts; record rules and caching; 35 general guides; 35 veterans guides.

Three service-content changes are live now: two listings added and one refreshed. The 70 guides were published and verified earlier, but a later release is serving the older pages. Restoration is prepared, not yet live.

COMPANY GRADE: NOT FINAL
Sep 27 is a pre-pilot dry run. Three of six areas are measurable, covering 60% of the weighting:
• Trust 0.0/10: the formula hits its floor on unresolved catalog conflicts. This is a control alarm, not a verdict that the product has no value.
• Execution 1.8/10: 2 of 11 frozen commitments accepted on time. Completed work is listed separately from deadline performance.
• Media 6.7/10: 2 of 3 planned original posts.
• Help, Finance and Compute: not yet measurable from complete company-wide records.

The mechanical result is 1.717/10 across measured areas; it is not an official company-quality grade.

COST AND EVIDENCE
$16.80/day is the shared-stack budget estimate ($504.12 ÷ 30), not measured compute or reconciled company spending. Five partial usage receipts captured 2,405,465 tokens; full-day coverage and dollar cost remain unknown. Cost per useful change and people helped are not measured.

Tonight’s review completed 118 of 131 planned source comparisons; its six change candidates remain unpublished. A separate code-quality candidate passed 382 checks and remains undeployed.

Next: restore the guides, resolve catalog conflicts, and bring all company runs and costs into the same evidence record. Same six areas and layout each day; dated corrections when evidence changes.

[Fernandez Bjork @fdbjork]

Correction: the earlier recap missed today’s Codex releases. Here is the reconciled company record.

TEN RELEASE WORKSTREAMS
Operations front door; mobile program browsing; homepage clarity; the 18-second film; resource updates; portfolio, essay and film; Ops charts; record rules and caching; 35 general guides; 35 veterans guides.

Three service-content changes are live now: two listings added and one refreshed. The 70 guides were published and verified earlier, but a later release is serving the older pages. Restoration is prepared, not yet live.

COMPANY GRADE: NOT FINAL
Sep 27 is a pre-pilot dry run. Three of six areas are measurable, covering 60% of the weighting:
• Trust 0.0/10: the formula hits its floor on unresolved catalog conflicts. This is a control alarm, not a verdict that the product has no value.
• Execution 1.8/10: 2 of 11 frozen commitments accepted on time. Completed work is listed separately from deadline performance.
• Media 6.7/10: 2 of 3 planned original posts.
• Help, Finance and Compute: not yet measurable from complete company-wide records.

The mechanical result is 1.717/10 across measured areas; it is not an official company-quality grade.

COST AND EVIDENCE
$16.80/day is the shared-stack budget estimate ($504.12 ÷ 30), not measured compute or reconciled company spending. Five partial usage receipts captured 2,405,465 tokens; full-day coverage and dollar cost remain unknown. Cost per useful change and people helped are not measured.

Tonight’s review completed 118 of 131 planned source comparisons; its six change candidates remain unpublished. A separate code-quality candidate passed 382 checks and remains undeployed.

Next: restore the guides, resolve catalog conflicts, and bring all company runs and costs into the same evidence record. Same six areas and layout each day; dated corrections when evidence changes.

[Fernandez Bjork @fdbjork]

PHX Good · Sep 27 · Scorecard baseline
118/131 listings checked; 0 changes published from tonight’s review.
8:52pm log: 9 runs, 5 drafts, 2 internal builds.
Cost: ~$17/day (budget estimate, not measured compute). People helped: not measured.
Next: resolve findings and publish.

[Alabama Football @AlabamaFTBL]

1-0 mindset! 👊

[Fernandez Bjork @fdbjork]

@Pexoai_offical @kimmonismus Thank you so much!

[Fernandez Bjork @fdbjork]

@CityofPhoenixAZ PHX Good refreshed its published-reference page for 20 W. Jackson and added a separate Cholla Library heat-relief listing. The City’s published 2026 heat-relief season ends Sept. 30; check current City guidance or 211 before travel. https://t.co/gVgFskMU0l

[Fernandez Bjork @fdbjork]

15-second explainer commercial, 16:9, warm off-white textured paper background, subtle grain and vintage halftone dots.

Cute anthropomorphic rose mascot: pink-red flower head with a simple smiling face, tiny hat, thin stem and two leaves, soft screen-print look.

Clean modern sans-serif type. Small section labels in the corners like “// 01 - hook”. Website https://t.co/LGnT0v9tGP in the top right.

0-3s: mascot centered in a dotted circle. Big text fades in: “Find help. Know what to do next.”

3-7s: mascot slides left. Three rounded cards appear: Shelter & housing (tiny house + cactus, 44 listings), Food & groceries (bag + apple, 41 listings), Rent & utilities (lightbulb + bill, 39 listings). Caption: “Maricopa County. Your city. A starting point. Catalog as of Sep 26, 2026.”

7-11s: phone mockup of a “Prepare a brief” form. Text beside it: “Prepare a brief. You send it. We don’t.” Small note: “Example brief · not a live referral.”

11-14s: two source cards with checkmarks and dates. Headline: “Clarity, without false promises.” Subtext: “A source check is never a promise of a bed, funding, or admission.”

14-15s: dark background, mascot + “PHX Good / Many cities. One community.” + button “Start a referral brief” + “https://t.co/LGnT0v9tGP · Free. No account needed.”

Calm, helpful, no voiceover, simple motion only, no camera shake.

[Fernandez Bjork @fdbjork]

I posted this PHX Good video and people asked how it was made.This is the prompt I used. I wrote it for the product, not for a generic “AI showreel.”PHX Good is a free referral guide for Maricopa County case managers: find a program, see whom to contact, know what to ask, and prepare a brief. It does not send the referral or promise a bed, funding, or admission.Prompt:Create a premium motion-design explainer for PHX Good (https://t.co/LGnT0v9tGP).Style: clean motion graphics, cream paper background, black ink, dusty-rose accent. Simple dotted-print desert rose mascot with Phoenix mountain contours. Modern sans-serif type. Tiny section labels in the top-left: // 01 – hook, // 02 – needs, // 03 – brief, // 04 – sources, // 05 – next. Smooth UI motion: cards, phone mockup, source lines. Calm music. No voiceover. Stay illustrated. No photoreal people.Sequence: // 01 – hook: “Find help. Know what to do next.” // 02 – needs: cards for Shelter & housing, Food & groceries, Rent & utilities. Line: “Maricopa County. Start close to home.” // 03 – brief: phone mockup of city → need → first contact + backup + what to ask. Line: “Prepare a brief. You send it. We don’t.” // 04 – sources: cards flip to source lines. Line: “Clarity, without false promises.” // 05 – next: PHX Good / “Many cities. One community.” / “Start a referral brief” / https://t.co/LGnT0v9tGP / Free. No account https://t.co/MWtJa2xzdP not promise beds, funding, or admission. Tone: useful, honest, local.If you work with Valley case managers, the tool is here: https://t.co/LGnT0v9tGP

[Fernandez Bjork @fdbjork]

@achxvi I used this prompt to make a commercial for my company that helps people in phoenix find resources for free!

[Fernandez Bjork @fdbjork]

AHCCCS, SNAP, housing waitlists—these can be hard to explain clearly.

PHX Good has 10 short explanations in English and Spanish, with copyable text, next steps and official sources.

For case managers and the people they support:
https://t.co/tzLMgABA7s

[Fernandez Bjork @fdbjork]

PHX Good now has 190 programs. Our focus: current official sources, clearer next steps, and referral briefs that stay useful. We’re putting daily source reviews and evidence checks on a schedule—and keeping unknowns visible. https://t.co/2ZnW4d3pYQ

[Fernandez Bjork @fdbjork]

Released: 100 source-reviewed additions, including 10 veteran/military programs. PHX Good now lists 190 programs, plus 10 copyable English/Spanish explanations.

Source review ≠ confirmed openings. 75 older records remain overdue for review.

https://t.co/tzLMgABA7s

[Fernandez Bjork @fdbjork]

PHX Good | September 26

Today’s work: verify 100 additional programs, including 10 veteran resources; write 10 plain-language explanations case managers can share; and repair unfinished site and scheduling work.

These are targets. Results will follow the evidence.

https://t.co/zho07QR3GN