Presented live or sent ahead and read cold.

Nine decks from the Fun files: a council briefing, a pilot pitch, a market read, a sales deck, a class, a case study and the thinking behind them.

Four follow El Paso’s incentive deals and the Meta campus, from who builds the data center to why a public-records request cost $1,500. One reads Meta’s new AI agent the week it went big. One sells a first step in B2B analysis, one runs an afternoon class that ends with a film, and two set out the ideas underneath.

Every slide is on this page.

Press the F to see a deck.

Council briefing

Who builds the data center?

Aug 2026 · 12 slides

See every slide ↓

A readiness audit built for one council member. The Meta campus needs 53–71% of El Paso’s entire electrical-contracting workforce at once, and the county trains 32 electricians a year.

Built to survive forwarding: every figure sourced, no one cast as a villain, and the two questions the record couldn’t answer reported as unanswered.
Market read

Is Muse fun?

Sep 2026 · 12 slides

See every slide ↓

Meta’s new AI agent, read the week it went big: what’s fun, what isn’t, what 2.5 million downloads mean for AI adoption, and whether it’ll be out for the holidays.

Every claim carries its source on the slide. The not-fun list comes from the reviewers, not from us.
Sales deck

Some vs. none

Sep 2026 · 10 slides

See every slide ↓

Where AI in B2B actually starts: not good vs. great analysis, but some vs. none. The competitor is inertia, and the offer is one question, one dataset, one answer, one week.

Written to be left behind after the meeting. It says what we won’t sell as plainly as what we will.
Class deck

Make a film with AI

Sep 2026 · 16 slides

See every slide ↓

The run of show for a hands-on AI video class: story, character, a shot plan with Claude, permission, and a lineup where everyone’s character meets. Everyone leaves with a film.

Every example is from a real film on the reel. The frames on the slides are the frames the method made.
Case study

Broken by Design

Jun 2026 · 11 slides

See every slide ↓

One public-records request, followed through five failures: a $1,440 fee, a waiver never ruled on, a request closed by silence, and two justifications for the same 80 hours.

Each failure is tied to the statute it tests, and it ends in seven reforms the city can adopt without waiting on Austin.
Pilot pitch

380 Governance Audit Pilot

Jul 2026 · 7 slides

See every slide ↓

A 30-minute briefing for city management. The city’s own auditors found five control gaps, and the reform stops at the data-center door. The ask is a ten-agreement pilot.

The first line of the pitch is that no one broke a law. What’s proven is stated flatly, and the dollar figure is the smaller, defensible one.
Thesis carousel

Governance as Antagonist

Jul 2026 · 9 slides

See every slide ↓

Governance isn’t the problem; the absence of it is. A nine-frame argument for the person willing to make the system visible.

Built as a LinkedIn carousel: one idea per frame, one question at the foot of each.
House thesis

Compounding Gains

Jul 2026 · 6 slides

See every slide ↓

Why work should be built as assets that compound instead of expenses that evaporate, and why only shipped assets count.

Six slides, one engine, five fronts: the operating thesis behind the rest of this page.
Policy timeline

Economic Governance Evolution

Jul 2026 · 10 slides

See every slide ↓

El Paso’s path from the 2024 audit findings to the 2026 data-center framework: what changed, what didn’t, and why the Meta agreement stayed.

Reads as a timeline an aide can brief from: one year per slide, then the standards, the mechanisms and the outlook.

1 · Council briefing

Who builds the data center?

Aug 2026 · 12 slides

El Paso · August 2026

Who builds the
data center?

A readiness audit of El Paso’s skilled-trades workforce and contracting capacity, measured against the announced Meta / BlackRock campus.

Prepared by Luis Ruiz · Fun

Findings only. No engagement is being proposed and no fee is attached to any of it.

The announcement

What has been committed publicly

1 GW

final capacity

operational 2028

$10B

Meta investment

$12B bond, BlackRock-led

4,000+

construction workers

at peak

300

permanent jobs

once operational

The load ramp is on the regulatory record

2026 · approved

220 MW

2027 · projected

440 MW

2029 · projected

1,000 MW

Bridge generation

McCloud Generation — 366 MW across 813 modular gas generators, adjacent to the site, exclusive to the campus for an initial five-year bridge period from 2027 and separate from El Paso Electric’s transmission system.

Sources: Meta newsroom announcement, 21 July 2026 · El Paso Electric filing with the Public Utility Commission of Texas.

Fun · Who builds the data center?02

The question

This audit does not ask whether the campus is good for El Paso.
It asks a narrower question that has an answer:

Can El Paso staff and contract
what has already been announced?

Two things follow from the announcement itself: the peak workforce is a published number, and the dates are filed. That makes readiness measurable rather than a matter of opinion.

Fun · Who builds the data center?03

Method

How this was measured

Four design choices, made so the findings can be checked rather than taken on trust.

Rules fixed before the data was pulled

Every threshold was written down and dated before collection ran. The rules file is checksummed into each report, so a reader can confirm the test was not fitted to the result.

A control case, not a benchmark

Findings are measured against Abilene, where a comparable buildout already happened and the outcome is documented — not against an abstract standard.

“Insufficient” is a real answer

Where the public record cannot answer a question, the audit says so. Missing evidence is never counted toward the conclusion and never treated as zero.

Every figure traceable

All findings come from federal sources — the Bureau of Labor Statistics and the federal education data system — each carrying its source and retrieval date.

Fun · Who builds the data center?04

Finding one · The arithmetic

The project needs most of the sector at once

Electricians are roughly 30–40% of data centre construction craft. Against a published peak of 4,000 workers, that is 1,200–1,600 electricians at the same time.

2,263

people employed in electrical and wiring contracting in El Paso County

173

establishments — an average of 13 employees each

An average establishment of 13 employees is a residential and light commercial firm. Data centre electrical scope is self-performed by contractors an order of magnitude larger — which is a separate question from whether the workers exist.

Share of the county’s entire electrical contracting workforce required by one project

53%

at 30% trade share

71%

at 40% trade share

The denominator is the whole sector, including everyone already committed to residential and commercial work. It is a ceiling on available local labour, not a pool of idle workers.

Source: BLS Quarterly Census of Employment and Wages, El Paso County, NAICS 23821, 2025 Q4.

Fun · Who builds the data center?05

Finding two · The pipeline

El Paso trains 32 electricians a year.

32completers a year across the county

1,200–1,600electricians required at peak

Spread generously over the three years from announcement to operation, the annual requirement is 400–533. Local programmes cover 6–8% of it.

This is a floor, not a ceiling: it counts degree-granting institutions only and excludes union and merit-shop apprenticeships.

Source: IPEDS completions, El Paso County institutions, CIP 46.0302, 2023–24 award year.

Fun · Who builds the data center?06

Finding three · The control case

Abilene is not a comparison. It is a preview.

Abilene · before the Stargate ramp

1.36%

of all private employment was in electrical contracting

What followed:

local electrician pool depleted
contractors imported out-of-state labour
Texas opened licence reciprocity with three states
site wages ran to roughly double local subcontractor scale
housing costs spiked

El Paso · today, before its ramp

0.83%

of all private employment is in electrical contracting

0.61×

El Paso enters the same curve with roughly six-tenths the trade density of the metro where the labour market was documented as breaking.

Source: BLS QCEW, Taylor County (2023 Q4) and El Paso County (2025 Q4), NAICS 23821, as a share of total private employment.

Fun · Who builds the data center?07

The limits of the public record

Two questions the audit could not answer

Both concern whether any of this is enforceable. Neither could be answered, because the records are not public.

Insufficient evidence

Can local firms capture the work?

Prime and subcontract awards for the campus are not published, so local participation cannot be measured. The firm-size distribution needed to judge whether local contractors could hold the scope is also unavailable.

Insufficient evidence

Is there any binding local-hire commitment?

The operative incentive agreement text is not in the public record reviewed, so it cannot be tested for local-hire terms, apprenticeship requirements, wage floors, reporting duties, or clawbacks.

The audit does not treat a gap in the record as evidence for its own conclusion. Both questions are reported as unanswered, and both are now the subject of records requests.

Fun · Who builds the data center?08

Why the gap matters

The three questions that could be answered from public data all point the same way.

The two that could not are precisely the two about whether anything is enforceable.

Answered	Is the demand real, dated and sized?	Yes — filed with the PUC
Answered	Is the local trades workforce large enough?	No — 53–71% of the sector
Answered	Is the training pipeline scaled to it?	No — 6–8% of annual need
Unanswered	Can local firms capture the work?	Awards not published
Unanswered	Is there binding local-hire language?	Agreement text not public

Fun · Who builds the data center?09

Filed

Two records requests, one clock

Both were filed with the City through the Public Information Center on 13 August 2026 and acknowledged the same afternoon.

W211728-081326

Contracting and award records

The incentive agreement and amendments; contractor and subcontractor lists; local participation requirements and reporting; permits with contractor of record; economic impact assumptions.

W211732-081326

Community benefits and workforce provisions

The executed agreement and exhibits; any local-hire, apprenticeship, or wage terms; compliance reports filed to date; clawback provisions; the department responsible for monitoring.

Statutory response

August 2026

The tenth business day after filing, under Chapter 552 of the Texas Government Code.

A third request went to Workforce Solutions Borderplex for apprenticeship registrations — the data most likely to revise the training-pipeline finding upward.

Fun · Who builds the data center?10

What would close the gap

Six elements that make a commitment enforceable

The Community Benefits Agreement is where workforce terms either get written or don’t. These are the elements the audit tests for — offered as drafting considerations, not as proposed final text.

1 Local hire percentage

A stated share, with a defined geographic hiring area.

2 Apprenticeship utilisation

A required share of hours worked by registered apprentices.

3 Wage floor

Prevailing wage or a stated floor, named explicitly.

4 Reporting cadence

How often, in what form, and by whom — named in the agreement.

5 Independent verification

A verifier independent of both the City and the recipient.

6 Clawback on non-performance

A defined consequence — recapture, reduction, or termination.

A commitment with no reporting duty and no consequence attached counts as absent for this test, however it is worded.

Fun · Who builds the data center?11

Verification

Everything here is reproducible.

Every figure carries its source, its identifier, and the date it was retrieved. The thresholds were fixed and dated before collection. Anyone can re-run the analysis and obtain the same result, or apply different thresholds and see what changes.

1 Workforce and firm counts

BLS Quarterly Census of Employment and Wages, El Paso and Taylor Counties, NAICS 23821, 2019–2025

2 Training completions

IPEDS completions, El Paso County institutions, CIP 46.0302, 2019–20 to 2023–24

3 Demand figures

Meta newsroom announcement, 21 July 2026; El Paso Electric filing with the Public Utility Commission of Texas

4 Abilene outcomes

Contemporaneous reporting on the Stargate campus, Taylor County, 2024–26

Luis Ruiz · Fun · luis@funistheproduct.com | Findings only. No engagement is being proposed and no fee is attached to any of it. This is governance-audit analysis, not legal advice.

Fun · Who builds the data center?12

1 / 12

2 · Market read

Is Muse fun?

Sep 2026 · 12 slides

Fun · Release notes · September 2026

Is Muse fun?

What’s fun, what isn’t, and what Meta’s new AI agent means for AI adoption.

Luis Ruiz · Fun · luis@funistheproduct.com

The release

Meta’s new agent runs your errands

Sept 8 · Launch

Muse arrives

A personal AI agent you text or talk to. It works in the background across your email, calendar, browser and shopping, then reports back.

iPhone, Android, web, Mac and WhatsApp. U.S. only, 18 and up. Free to start, with paid plans for heavier use.

Sept 23 · Meta Connect

Muse gets a voice and a body

Voice mode that keeps working while you talk
Realtime Avatar, a face for that voice
New connectors: Walmart, Best Buy, Sephora, Instacart, PayPal, Notion, GitHub and more
Muse Charm, a pocket device made for talking to Muse

Sources: Meta Newsroom, Sept 2026 · Meta, Everything We Announced at Connect 2026

Fun · Is Muse fun?02

Downloads in about two weeks

2.5 million

902,000 in the first six days. The No. 1 free app on the U.S. App Store, ahead of ChatGPT and Claude.

Sources: Sensor Tower via Bloomberg, Sept 21 · CNBC, Sept 21, 2026

Fun · Is Muse fun?03

What’s fun

Less chatbot.
More capable friend.

The fun is in how it feels. You hand off an errand the way you’d text a friend who’s good at this stuff.

Fun · Is Muse fun?04

What’s fun

Four things Muse gets right

It talks while it works

Voice mode holds a long conversation while Muse does the task in the background.

It runs the errand

Say what needs doing. It works across your email, calendar, browser and stores, then reports back.

It lives where you text

The Muse app, WhatsApp, the web or a Mac. No new habit to learn first.

You pick its voice

Describe the voice you want: faster, slower, even a specific accent.

Sources: Meta Newsroom · Meta, Everything We Announced at Connect 2026

Fun · Is Muse fun?05

What’s not fun

A friend who keeps
asking for more keys.

The catch: Muse works better the more of your life it can reach, and reviewers say it keeps asking for more.

Fun · Is Muse fun?06

What’s not fun

Four places the fun stops

It keeps asking

Wired’s reviewer was pushed, again and again, to connect an email inbox and banking.

Training is opt-out

Your chats can help train Meta’s AI unless you switch that setting off.

It reads past the ask

An Inc. reviewer found Muse reading private messages nobody asked it to read.

The errand comes second

Wired’s verdict: it put learning about the user ahead of finishing the task.

Sources: Wired (Reece Rogers) · Inc. (Jason Aten) · Meta Newsroom

Fun · Is Muse fun?07

Meta’s answer

Meta built real locks

Secure VM

Muse runs in its own isolated machine, with your data stored apart.

Sentinel

A second agent has to approve anything Muse sends to the internet.

A vault

Your logins sit where Muse can use them but never read them.

Audit trail

A running record of everything it did and plans to do.

Coming later

Confidential VM, encrypted with keys only you hold.

The open question

This is serious engineering. The question left is trust: you’re still giving Meta the run of your email, calendar and browser.

Sources: Meta Newsroom · Meta Help Center · eesel AI, Muse review

Fun · Is Muse fun?08

The Fun take

Fun has rules

Play needs a boundary. Here are three that keep Muse fun.

Start with one errand

Give it one job this week, not your whole life. Add access when a task needs it.

Check the training switch

Decide whether your chats help train Meta’s AI before you connect anything.

Read the receipt

The audit trail shows what it did and what it plans to do. Read it like a bank statement.

Fun · Is Muse fun?09

What it means for AI adoption

Muse sells a friend

How AI is usually sold	How Muse is sold
Do more with fewer people	A friend who handles things
Learn to write prompts	Text it like a person
Another app to master	Lives in WhatsApp
Productivity software	Something to talk to

That’s the Fun thesis out in the wild. AI adoption stalls when it’s sold as subtraction. It takes off when life gets more interesting.

Fun · Is Muse fun?10

What it means for AI adoption

Downloads are curiosity.
Adoption is January.

2.5 million people tried Muse. The number that matters is how many still let it into their inbox, calendar and wallet after the holidays. Every item on the not-fun list is a reason to disconnect.

Fun · Is Muse fun?11

The question everyone’s asking

Will it be out for the holidays?

The app, yes. The Charm, maybe. Muse on glasses, not yet.

Product	What Meta has said	Our read
Muse app	Out since Sept 8 on phone, web, Mac and WhatsApp. U.S. only, 18 and up.	Yes. It’s already here.
Muse Charm	Zuckerberg wants it ready to ship for the holidays. No price yet. Not in mass production; only a few units built.	Maybe. A target, not a date.
Muse on Meta glasses	Coming in the next few months, with no date given.	Not yet. Don’t gift glasses expecting Muse.

Sources: Meta Connect 2026 recap · iThinkDiff · CNBC · checked Sept 24, 2026

Fun · Is Muse fun?12

1 / 12

3 · Sales deck

Some vs. none

Sep 2026 · 10 slides

Fun · AI for B2B

Some
beats none.

Where AI in B2B actually starts: not good vs. great analysis, but some vs. none.

Luis Ruiz · Fun · luis@funistheproduct.com · September 2026

Most decisions get no analysis at all

Gut feel

The call gets made by whoever is most confident in the meeting.

Stale dashboards

Built once for a question nobody asks anymore. Still on the wall.

Nobody looked

The data exists. The question was never asked of it.

Fun · Some vs. none02

The comparison is wrong

What vendors pitch

Good vs. great

“Our model is more accurate than theirs.” Assumes you already have analysis to improve.

What buyers actually face

Some vs. none

Any honest look at the data beats the decision made without one.

Fun · Some vs. none03

You don’t have to beat an analyst.
You have to beat nobody looking.

Directionally right in ten minutes is a win over never done.

Fun · Some vs. none04

The competitor is inertia

Why buyers stall	What we do about it
Starting feels expensive	Make the first step small and fixed-fee
Nobody knows where to begin	We pick the one question worth asking first
“Our data isn’t ready”	Start with the dataset you already have
AI projects never seem to end	One answer, one week, then decide

Fun · Some vs. none05

The first step: one of each

Question

The decision you’re about to make without data.

Dataset

Whatever you already have. No migration, no cleanup project.

Answer

A written finding you can act on or argue with.

Week

Fixed scope, fixed fee. Credited if you continue.

Fun · Some vs. none06

Value is what you would have missed

Not “your analysts got 30% faster.” Instead: here’s what you couldn’t see before.

The churn signal

Accounts drifting before they cancel.

The pricing gap

Where you’re leaving money, or customers, on the table.

The wasted spend

Budget going to things nobody checked.

Fun · Some vs. none07

No baseline means we show the work

With nothing to compare against, you can’t tell if the answer is right. So every finding comes with its receipts.

Sources

Exactly which data it came from.

Assumptions

What we took as given, written down.

Confidence

How sure we are, and what would change our mind.

Fun · Some vs. none08

What we won’t sell

No guaranteed lift.
No magic dashboard.

We sell the finding and the method behind it. What you do with it is yours. That’s not a hedge. It’s why you can trust the answer.

Fun · Some vs. none09

Next step

Bring one question.

We’ll bring back one answer, with the work shown, in a week.

Luis Ruiz · Fun
luis@funistheproduct.com

1 / 10

4 · Class deck

Make a film with AI

Sep 2026 · 16 slides

Anise Baca, a woman in her sixties, running as a white medical exoskeleton assembles around her. A frame from the film Legs.

Fun · AI video class · El Paso

Make a film with AI.

Plan it before you render it.

A hands-on class in five parts. Everyone leaves with a finished short film.

Luis Ruiz · Fun · luis@funistheproduct.com

Why this class

Sound familiar?

The first image was great. In the second shot, it was a different person.

You spent your credits finding out what the tool can’t do.

It looks like AI because nothing in it was decided.

This class fixes all three by deciding before anyone presses render.

Fun · AI video class02

The class