Every Substack topic we have, in one place

The model congress

544 AI agents, one per sitting member of the House and Senate, each grounded in that member's real voting record. Connor and Rebecca put their plan in front of them and made them negotiate it, cut deals, count votes, and say what could pass. Built as discovery, not advocacy: the finding is what the simulation did, not what anyone wanted it to say.

Then on August 6, while the essay was still being written, three members of Congress introduced a bill the model says cannot pass. It has been checked line by line against the simulation, including the part where the model appears to have named two of the sponsors months in advance. That check is sitting inside this essay right now and is the strongest candidate to be cut out and run as its own post.

Receipts

Full simulation results, the 13 provisions, every vote count, and the HR 10044 run from Aug 19.

Came from

Chosen as essay 2 on the June 8 call. Outline sent to Connor Aug 19.

Source doc

AI Congress essay outline

Connor asked for this one on Aug 25.

Why no one knows what programs are working

Nobody can tell you which workforce programs actually work. Connor's receipt is Merit's own: it took eight years to get government data speaking to the efficacy of their programs. Eight years, for the organization that wanted the answer most and had every reason to chase it.

The argument runs from there to the money. A country spending $28.2 billion a year on workforce development cannot see what it is buying, which is the same hole underneath funding on outcomes instead of enrollment, and underneath every corporate reskilling pledge on this list.

Receipts

Merit's own eight-year wait for government efficacy data. The $28.2 billion federal workforce spend and the 12 percent average wage gain, both already checked.

Came from

Connor in Slack, Aug 25: one of two pieces we would like to write asap.

Connor asked for this one on Aug 25.

The philanthropy problem

A hundred billion dollars a year flows through philanthropy and produces hundred-thousand-dollar nonprofits. Philanthropists went paralyzed at the exact moment AI demanded they move. Named as problem one in the launch essay and then deferred to a future piece, which is the whole reason to write it.

Connor's framing, Aug 25: how philanthropy actually flows in this country, and how it is a drop in the bucket compared to what we should actually be deploying.

Receipts

Rebecca's funnel math from the Sept 29 2025 call. The higher-ed pledge movement.

Came from

Rebecca's funnel math, Sept 29 2025 call, then teased in the published essay. Connor named it Aug 25 as one of two he wants to write asap.

The emperor has no skills gap

The gap is real, it is why Merit exists, and it is also the alibi. A roofing company that cannot find skilled workers is offering $15 an hour against Amazon's $18. A glass factory that could not find workers needed air conditioning. Operators saying this at their own expense is the whole angle.

Receipts

The roofing wage ladder, the glass factory, Anthony's coinage of the phrase, Cappelli 2012, Summers 2015.

Came from

Connor's pitch 7, plus Rebecca's argument that the skills gap is a cover story. Was going to be essay 2 before the congress piece took the slot.

Labor versus capital

The country taxes labor and protects capital, and the AI transition runs straight down that groove. Not a new observation. Rebecca's register on it is the reason to publish it.

Receipts

The tax argument, condensed out of an earlier draft of the launch essay. The Elon line needs verifying before it runs.

Came from

Cut from the launch essay for length, and flagged at the time as deserving its own piece.

Source doc

Can America (Please) Make A Plan?

Unfucking the American Dream

Connor's umbrella frame, and the one he imagines as a book cover with a flag in place of the U. The dream was not passively broken, it was actively undone, so the job is to un-undo it. Every other idea on this list hangs off it.

Receipts

Connor verbatim, April 8 2025.

Came from

Connor. Recurring across calls since 2025.

Everyone's LinkedIn posts read the same

A piece making fun of how identical LinkedIn has become, written in the exact style it is making fun of, pushed hard. The only comedy piece on the list.

Receipts

None needed.

Came from

Connor, in Slack, June 16 at 10:54pm. I said yes in the thread.

Reskilling is not enough

The most credible contrarian claim either of them can make: they run the largest reskilling nonprofit in the country and they are the ones saying reskilling alone will not save anyone. Ran as the heart of the launch essay. There is more of it left than got used.

Receipts

The roofing ladder, the reskilling supercut, Brookings retraining data, Summers on evasion.

Came from

The spine of the whole publication.

Doers, not pontificators

The identity claim. We are people who are operators, we are creators, we are not pontificators. The anti-op-ed channel, 80 percent emotional to 20 percent wonky.

Receipts

Connor verbatim, June 8 call.

Came from

Connor, June 8. Also the line the whole positioning was built around.

The third lane, between doom and hype

Neither camp. We don't know what's coming, we just refuse to not-plan. Pack the go-bag while everyone else argues about the asteroid.

Receipts

The asteroid frame from the launch essay.

Came from

Pitched as an angle for the launch essay, June 8. Defines what the Substack is.

Where is the plan

No plan from Biden, none from Trump, none from the think tanks, none in the 100-plus federal AI bills or the 1,000-plus state ones. Ideas without a wheel to attach them to. This became the launch essay and still has a back half that was cut.

Receipts

The OpenAI industrial policy whitepaper, Bernie's sovereign wealth fund op-ed, the bill counts.

Came from

Became the published launch essay, June 24.

Source doc

Can America (Please) Make A Plan?

The fights-its and the fuck-its

A taxonomy of who packs a go-bag and who drinks on the roof. Most American policymakers have a serious case of the fuck-its. Cut from the launch essay for length, survives whole as a standalone.

Receipts

A friend of Connor and Rebecca's coined it.

Came from

Pitched as an angle for the launch essay, then cut for length. Available intact.

Representation bias

Connor's coined term. The people who shape the AI and jobs conversation are college-educated and report on their own circles first, so 53 million workers without degrees are structurally absent from it. Three exposures non-degree workers carry that graduates do not: their entry-level roles are equally automatable, they have no credential to prove competence when they pivot, and they have no network to catch them.

Receipts

The Sam Altman argument. Merit's national poll, 51 percent call the economy poor. The Ezra Klein and Citrini framings.

Came from

Connor. SKDK built their whole launch-post concept on this.

Source doc

SKDK memo, May 7 2026

Women and the invisible disruption

AI could do to high-school-educated women in admin work what deindustrialization did to high-school-educated men in manufacturing. Roughly 12 million jobs. Molly Kinder's thesis at Brookings, and Merit's angle is that those 12 million jobs are literally their recruiting pool.

Receipts

Kinder's full automation-risk table by occupation. Diane the employee of the year.

Came from

Meghan and Kristen via the SKDK thread, June 10. Named as the Pivotal and Lean In play.

Source doc

SKDK memo, May 7 2026

Entry-level jobs are already declining

Stanford's Canaries in the Coal Mine on ADP payroll records: since ChatGPT, employment for 22 to 25 year olds in the most AI-exposed jobs fell 13 percent relative to older workers in the same jobs. The bottom rung, the job where you are allowed to be new, is the cheapest thing for AI to take.

Receipts

Brynjolfsson, Chandar and Chen, millions of payroll records.

Came from

Found while checking the numbers for the launch essay.

AI is already in the layoff data

Challenger, Gray and Christmas logged 87,714 US job cuts attributed to AI in the first five months of 2026, about 22 percent of all layoffs, past the 54,836 for all of 2025. The angle is the fight, not the stat. Rebecca wants to cut it because most people she talks to think AI is an excuse companies hide behind. That disagreement is the piece. When because of AI becomes the default corporate explanation, the framing has already won and the workers are out either way.

Receipts

Challenger data since 1993. Rebecca's margin comment on the draft.

Came from

I flagged it in the draft thread: we should write about this. The most recent one added before Connor's two.

Corporate pledge theater

AT&T pledged $1 billion for reskilling. Amazon $700 million. PwC $3 billion. Microsoft $4 billion. The pledges exist. The scale does not. Announcements built to manage political risk.

Receipts

All four pledge figures documented and sourced.

Came from

From the launch essay research, kept as a standalone.

Who actually needs reskilling

The real candidate is employed, working nights and weekends, and has kids. Programs that require you to not have a life are not programs, they are designed for someone who does not exist.

Receipts

Rebecca's verbatim walk-through of a first-generation CS grad, June 8 call.

Came from

Rebecca, plus Connor's pitch 9: stop telling unemployed people to become better students.

Good people, bad outcomes

Mark Nunnelly, former Dunkin' Brands CEO, asked whether he would consider mass AI layoffs, smiled and said: well I am a capitalist, aren't I? Not a bad person. A person inside a system.

Receipts

The Nunnelly exchange. The historical record of companies fighting new rules and then living with them.

Came from

Connor and Rebecca, from a meeting.

CEO incentive transparency

Don't hate the player, hate the game. 99 percent of executives expect AI-driven layoffs within two years. A CEO at dinner: I know it's terrible for society, but to be clear, my incentive is to automate labor with AI as quickly as I can. Rebecca trusts that guy more than the ones who won't say it.

Receipts

The dinner quote. The Mercer 99 percent stat.

Came from

Already ran as a Rebecca LinkedIn post. The only one on this list that has run anywhere.

Fund results, not promises

$28.2 billion a year of federal workforce spending returns 12 percent wage gains. Not enrolled. Not credentialed. Employed.

Receipts

The $28.2 billion federal spend, and the gap between people who finish a credential and people who land a job.

Came from

A provision in Connor and Rebecca's plan. Long-running Merit position.

Source doc

Can America (Please) Make A Plan?

Why nonprofits don't scale

Merit is one of the most successful reskilling nonprofits in the country and its co-CEOs will say they would not build it this way again. Philanthropic capital moves too slowly and too small to fund a transition this size. Only public money scales.

Receipts

Merit's own per-learner economics.

Came from

Connor and Rebecca, repeatedly. Pairs with the philanthropy problem.

UBI versus UDI

A floor that stops people rioting but gives them no dignity, purpose or mobility is not stability. Universal dignified income as a platform for participation rather than a floor under collapse.

Receipts

Connor and Rebecca's plan.

Came from

A position they have come back to across several calls.

Source doc

Can America (Please) Make A Plan?

Care work as the future of labor

What is left is caring for a human or caring for a robot. Both categories grow: healthcare, childcare, elder care, and the maintenance work that keeps AI systems running. Connects directly to the women piece.

Receipts

Connor and Rebecca's plan.

Came from

From the plan notes.

Source doc

Can America (Please) Make A Plan?

The skills gap myth has a history

Cappelli in 2012, Summers in 2015, Brookings throughout. Researchers have said this for over a decade and the field nodded and moved on. Bloomberg Opinion ran the same thesis in June 2026. The distinctive thing is not the claim, it is who is making it and what it costs them.

Receipts

Cappelli, Summers, Brookings, the Bloomberg piece.

Came from

Built while checking the numbers for the launch essay.

Mission Employment Organizations

The provision that funds 501c3s to place Americans in jobs. Passed the simulation with 308 of 544. Also the provision that most directly funds Merit's own model, which has to be said out loud in the piece rather than left for a reader to notice.

Receipts

P3.2 and its vote count.

Came from

From the simulation results.

Source doc

Can America (Please) Make A Plan?

The keystone provision

P2.6, strategic-sector and advanced manufacturing, read as beat China on the right and good union jobs on the left. 474 of 544. Injected after the original ballot, so it was discovered rather than designed.

Receipts

The provision text and its rating.

Came from

From the simulation results.

Source doc

AI Congress essay outline

What died in the coalition

The most honest finding of the simulation is what did not survive. Every funding mechanism went net-negative: sovereign wealth fund, automation tax, robot tax, excess-profits tax, foundation payout requirements, carried interest. SNAP and Medicaid expansion drew 235 hard no's and got cut before the final vote.

Receipts

Full net-support numbers per mechanism.

Came from

From the simulation results.

Source doc

AI Congress essay outline

Democrats are getting higher ed wrong

Rebecca, Dem-aligned, on the record. When the coverage framed federal accountability as bad because Yale's music program could lose grants, she wrote: Democrats, we are getting this very wrong on higher education. High risk, flagged as such, hers alone to write.

Receipts

Rebecca's iMessage, verbatim.

Came from

Rebecca. Filed as high risk, and it is.

Robin Hood AI

Merit as the first organization to hand real AI tools to the workers who need them most, so they walk into interviews with portfolios instead of resumes. Most people are replacing workers with AI, here we would be leveling them up.

Receipts

Merit's own AI course work.

Came from

My pitch to Brittany on the June 3 call. Not yet something Connor or Rebecca have said they believe, and it is flagged that way.

The cite-to-recruit loop

Not a topic so much as a mechanism worth writing up: publish a piece citing a researcher prominently, DM them the published piece, then book them. Built for Molly Kinder after two direct asks got ghosted.

Receipts

The Kinder sequence, written out and never run.

Came from

Mine, from the June 10 SKDK thread.

Human accountability in AI's competitive edge

SKDK's second big idea, built as an alternating monthly LinkedIn series rather than a Substack piece. AI delivers information, humans deliver accountability. Anchored on the FaceTime story: a learner working overnight janitorial shifts refused to walk into the hospital's cyber department, so his coach FaceTimed him and stayed on the line through the elevator ride. Six-month shadowing, then a full-time cyber job.

Receipts

The FaceTime story. Research on near-peer mentorship.

Came from

SKDK, May 7 ideas doc. Never actioned.

Source doc

SKDK memo, May 7 2026

America needs a new kind of company

The IRA carbon-credit model applied to job creation. Rachel at Rockefeller already built the good jobs criteria.

Receipts

Rockefeller's criteria exist.

Came from

Pitch 1 in the March 2025 pitch doc.

The sliding scale of work

Replace the employed and unemployed binary with a continuum. Unemployment costs roughly $100k a person.

Receipts

The per-person cost figure needs verifying.

Came from

Pitch 2 in the same doc.

Expand the military, not AmeriCorps

The largest job trainer in America already exists and is already bipartisan.

Receipts

None yet.

Came from

Pitch 3 in the same doc.

The rare earth reckoning

The mining infrastructure gap as a national security crisis. Sits close to the keystone provision thematically.

Receipts

None yet.

Came from

Pitch 4 in the same doc.

Pre-legislate automatic triggers

Set-it-and-forget-it legislation that fires on economic indicators. Both trigger variants died in the simulation, so the piece has to answer that rather than ignore it.

Receipts

The simulation result cuts against it, which is usable.

Came from

Pitch 5 in the same doc. Whoever writes it has to answer the simulation result, which went the other way.

Source doc

Can America (Please) Make A Plan?

Invest upstream, not in the wreckage

Workforce money funds people after the layoff the way healthcare funds end of life.

Receipts

None yet.

Came from

Pitch 8 in the same doc.

The 90-day window we waste

The WARN Act gives 60 to 90 days of notice that we treat as dead time instead of training time.

Receipts

The WARN Act itself.

Came from

Pitch 10 in the same doc.

The let-them-eat-cake economy

The haves genuinely do not know what the have-nots are living.

Receipts

None yet.

Came from

A musing in the same March 2025 doc.

Cobalt collar

Reframing blue-collar work: cobalt, sapphire, titanium.

Receipts

None yet.

Came from

A musing in the same March 2025 doc.

I packed boxes at Amazon

A learner op-ed in a learner's voice, not Connor's or Rebecca's. A different register for the publication entirely.

Receipts

Merit's learner story bank, barely touched.

Came from

A musing in the same March 2025 doc.

Why higher ed stayed silent

The restaurant food-poisoning analogy. Undeveloped.

Receipts

None yet.

Came from

A musing in the same March 2025 doc.

I didn't drop out, I made an educated decision

The college dropout reframe, in a learner's voice.

Receipts

The learner story bank.

Came from

From the April 8 2025 call.

Paying for teachers in love

Why we expect caregivers to work for feelings. Pairs with care work as the future of labor.

Receipts

None yet.

Came from

From the June 8 2026 call.

Scale and impact run inverse

Received nonprofit wisdom. Have we been lucky, or did we find something real.

Receipts

Merit's own numbers.

Came from

From the April 8 2025 call.

The co-CEO roadshow where everyone said no

The meta relationship story. Folds into doers not pontificators if it does not stand alone.

Receipts

Connor and Rebecca's own history.

Came from

From the Sept 29 2025 call.

What companies owe their employees

Employer obligations, which Merit is already modeling internally.

Receipts

Merit's internal work.

Came from

Added March 2026.

What Congress actually agrees on

Simulation color that got cut, along with which Republicans said yes and why, and what America wants from AI jobs policy. Folds back into the model congress piece unless promoted.

Receipts

The full run data.

Came from

From the simulation cut list.

Source doc

AI Congress essay outline

The union self-perpetuation trap

An organization's first goal becomes sustaining itself. Rebecca's read is that it is unpublishable as written, everyone would freak out. Listed because it exists, not because it should run.

Receipts

The Delaware teachers-union quote.

Came from

Rebecca ruled it unpublishable herself.

The Nan response

I hope you're right, but in our experience. Explicitly held back from the launch essay as too gentle for a first piece.

Receipts

Nan's own words, written down.

Came from

Held back from the debut on purpose, as too gentle for a first piece.

The reskilling supercut and the duck-and-shark bit

Two formats rather than topics. The supercut cuts together clip after clip of people saying reskilling, then hard-cuts to not the way you're thinking. The duck-and-shark bit is the list of what Congress does instead of the plan, cut from the launch essay, and works as a Note.

Receipts

Both written. The supercut needs an edit.

Came from

Connor's idea. I loved it.