The Importance of Finding Receipts, An Ode to Why Details Decide Policy, Debt, and Democracy
Challenging Summary. Taxation. K-Scoring. Rising Fastballs. Thunderbird. Competing Break-up Albums.
I’ve been belated in my Newsletter duties. This is due to a writing project of size and scope, that darn day job I keep, and a ballooning coding project. Also, this article has a political bent around process and debt; people have opinions here. I imagine it’s tough writing politics consistently. Anytime I try to dip my toe in even mundane topics (the stuff I like to wonk on about), my output slows. So my apologies, next time I’m tackling Ninjas and Werther’s Originals Butter Candies. Until then, go forth. And feel free to skip to the end if you want to hit the other sections. Thanks for reading.
Hard to say how projects begin and where they ultimately end, some come through a certain happenstance and burn out quickly. Others never finish, constantly iterating and adding a feature here and there. Scope creep is real in experimentation.
This time, my friends, I read a bill!
Not really. Large Language Models (LLMs) have wrought vast summarization upon us; ChatGPT (or name your tool of choice, Claude is cool too) can summarize the news, create vacation packing lists, and draw up nightly dinner recipes. But what do we lose in the borderline fairy-tale land of summary? A missed quote in an article, or worse, a fake fact? What if we forget that extra sweatshirt for the late-night beach walk? And what about saffron, would the extra teaspoon have perfected the paella?
Do the details matter?
Or are we only left to our own stories and echo chambers? Take any Congressperson, Democrat or Republican, few appear to dig deep enough while many simply rehash the same monotonous talking points. That’s why politicians look alike. Like old media, these stilted suits repeat the same stories. Love him or hate him, Donald Trump is one of a kind.
Here, where 65 percent of Americans can name all three branches of government while, sadly, approximately 15-17 can’t name any, I’m going to be my own cheerleader. I’ve perused more than a few Congressional bills that have found the cutting room floor. Some managed to be passed into law. I know, this is an odd hobby, but the law becomes personal when non-compete clauses come calling, local easement regulations encroach on your local park, or, well, er, other stuff. Believe me, “Demons do hide in that six-point font.”
Unless you’re in a lawyering profession, how much do we pay attention to legalese? Have you ever read the user agreement on an iPhone? Heck, Tim Cook might have parental rights to our children under paragraph 39, clause 22. Those tiny letters probably aren’t important until a John Wayne look-alike, wearing a cowboy hat, rings the doorbell and tosses a subpoena on your welcome mat.
That’s not saying we don’t argue about what’s inside these ominous bills. Think about the odd Truths, Tweets, and articles that are consumed daily.
Our politicians shout on Fox News and inside CNN green rooms about odd policies. Take the rare bipartisan consensus with the so-called millionaire freeloaders amendment, targeting folks who pulled down a million annually yet drew unemployment insurance.
The outrage!
Turns out, the numbers are minuscule; the monsters drawn up on political whiteboards aren’t necessarily real.
These are stories, narratives to claw back funding or expand it. But Presidential legacy isn’t made through intent or solid policy. It’s sometimes created through odd unintended consequences of policies. There’s a libertarian argument; you’re better off not having any laws, avoiding the unintended consequences altogether.
Rewind the clock 25 years, I’d argue few defining moments in Congressional history exist; it’s mostly noise. I often write that the process matters, the how we do things, instead of the outcome.
Tax Cuts, Wars, and Borrowing, Tweaks that Opened Debt’s Doors
I know, Congress passes comprehensive tax policy each term. Loopholes are opened. Others are closed. In the early 2000s, Republicans argued that these new cuts would be paid off through certain budgetary measures, an increase in Gross Domestic Product (GDP) with a promise of more taxable revenue. Note, it’s not a poor argument. This has happened in our nation’s history, notably moving our effective tax rate from the 70s to the 30s during the Reagan administration. It’s more complicated, I get it. Don’t flood me with hate mail.
However, here’s the part we struggle saying in precise language: in the modern income-tax era, the United States fought long, sustained wars (Afghanistan and Iraq) while cutting federal taxes and financing the conflicts largely through borrowing. Prior, the country tended to weigh war’s cost with higher taxes, bonds, or both (World Wars, Korea, and late-Vietnam). And we moved into an all-volunteer military force in 1973.
How does a country weigh the cost of war? What’s the measure of overthrowing a vile dictator? What about Afghanistan? Would the answer change if an explicit tax showed on your next return? Maybe? Maybe not?
Without the check, no matter your answer, we began the march of the debt brigade while creating a strong competitor in Asia. Note, it could also be argued the US took it’s eye off China by allowing permanent membership in the WTO. Worse, enforcement of Intellectual Property provisions lagged and was uneven, giving Chinese firms room to learn and copy in key sectors as US consumers saw lower prices. See the current state of the automotive industry if you want a master’s level course about the trade-offs of short-term gains versus long-haul consequences in markets.
Anyway, tax cuts aren’t tax cuts unless they are paid for. And since Congress never cut spending, we’ve shifted the burden to future generations.
But another magic trick has been revealed through the media’s relentless pursuit of Trump’s taxes. It’s not unlike US Intelligence searching across the desert for Saddam Hussein’s Weapons of Mass Destruction. No nuclear stockpiles, only a boastful dictator telling stories, but the effort discovered A.Q. Khan’s network, which proliferated secrets to Libya, Iran, and North Korea, leading to their programs.
Sure, Trump has been accused of paying next to nothing according to most sources. But ProPublica did find its own version of the US Intelligence’s Pakistani scientist. Turns out, ultra-high-net-worth households minimize realized income using “buy, borrow, die” living on loans against appreciated assets.
Gingrich and Pelosi Power Shift, Attributed to Elon
My second unintended consequence falls outside my initial 25-year window. Newt Gingrich swept into power with the famed Contract with America. This led to a balanced-budget push, a political win for the Grand Old Party. However, Newt also cut the everyday Congressperson’s budget, leading to more power rolling into the speakership. Lean offices, smaller staff salaries. Prior to these changes, more work on legislation and oversight occurred in committee, less happens there because the speaker decides what comes to the floor. Nancy Pelosi consolidated this power in her tenure.
There was some hope that parts of this would roll back with Kevin McCarthy. He promised to pass a budget, bring smaller bills to the floor. More votes; true committee work; fewer social media tirades. But since he’s gone, the newly elected speaker has leaned on omnibus bills again. No idea why Kevin was forced out, nothing to see here.
I suppose these details take time to digest.
Here is the Senator from the great state of Tennessee, Bill Hagerty: “Thank God Elon Musk bought Twitter, because that’s the only way we would even know what’s in this bill ... Elon Musk’s transparency helped make that happen.”
Well, thank you. I like to think you’re elected to read these things. Give him credit for honesty.
In fairness to the Senator from the Great State of Tennessee, the challenge in our current legislative process is that these bills rarely bubble up in committee. Newt. Nancy. Paul. They blew this ages ago, twist people’s arms to find something they like about it, and maybe make a few concessions. What’s the problem with adding a few billion for Disaster Relief? Fair question, is it still debt? Or is money free? Do Alaskans notice federal taxes anymore given the immense amount of flow back? I know CRs keep the government open for business, creates compromise. Yet, the exception, often at the cost of transparency, is now the rule.
So here we are, the Big Beautiful Bill passed, and few understand what’s in it. My favorite quote comes from Nancy: she’s often cited as saying, “You’ll know what’s in it when it passes.”
Changing Processes, Better Outcomes?
So these two unintended tweaks have created a storm where immense debt is welcomed and few read the fine print. The consequences are gone. Soon, net interest is projected to rival or exceed single mega-programs like Medicare or Social Security, without adding any new services. It’s just math. Yet, I laugh when the news asks this gotcha question to Congress, “Have you read the bill?”
Why?
It’s impossible. These monsters clock in at over 1,000 pages. I sifted through the Voting Rights Act, took notes. How long did it take? I won’t lie; this is boring stuff, no page-turners here. I labored six months with a highlighter in hand. And trust me, if you have trouble sleeping, doctors should prescribe these for a little late-night reading.
AI Summaries Overwhelmed
So how can our friendly LLM of choice help solve this problem?
The challenge with any document summarization is the immense noise with anything political. Chatbots are also subjected to doomsday talk on climate change, stories about Hillary Clinton eating children’s faces, Jewish space lasers, immigrants eating cats... I could keep going, but you get the idea. The machine isn’t an all-knowing being. These models have limited context windows; even the largest, with huge token limits, truncate information for speed, performance, and guessing what the response should be to maintain engagement.
Summarization is challenging.
ChatGPT is the fastest-growing application in history. The scale, from zero to hundreds of millions, is hard to imagine. Think of all the hacks that went into making Google Search what it is today; now condense that time span into eighteen months.
But in doing so, the company has to shortcut what users want to know and that means articles, internet searches, opinions become truncated. That’s not necessarily bad; frankly, it’s a practical reality.
Yet, the One Big Beautiful Bill Act is an omnibus, stretching into thousands of words. So what do you do?
I downloaded the text from Congress’s website, segmented the 150,000 words into a database leveraging Python embeddings, and built a Retrieval Augmented Generation (RAG) system.
Showing receipts, I chunked the bill at ~1,000–1,500 tokens, embedded said chunk, set the top-k scoring and relevant passages per question, and constrained the model to answer from retrieved spans only. The last point is critical to prevent drift—logging the prompts and section hits to linked quotes and section numbers.
Long-form summarization does have trade-offs. Sometimes, sending the data in its entirety through AI models shows higher accuracy at higher cost and performance. Here, the embeddings guide the AI to statistically significant portions of the bill. And by focusing the model on the document, I filter out, yes, the silliness. No talk on Sydney Sweeney’s jeans. Or the imminent demise of hospitals across the United States, a trend unrelated. It’s just the bill. And what it does.
Change-Worthy Notations, Winners and Losers, and Gaming the Market
I asked three questions. The first, What are the most impactful provisions from the provided legislative text, focusing on those likely to drive real-world change? Surprisingly, the answers ranged from closing certain business write-offs and ending trade carve-outs taken advantage of by shipping companies. I once wrote about the perils of fast fashion; for me, this is a solid change. And there were changes to student loan evaluations, benefiting farmers and other businesses. Here was a portion of the noise-free response:
The most consequential provisions include the major changes to customs and remittance taxation, permanent tax reforms affecting high-income individuals and corporations, and sweeping changes to federal student loan programs. These measures will have broad impacts across industries (e-commerce, higher education, finance, clean energy), affect millions of individuals, and signal significant shifts in US policy priorities.
My next question, Who are the winners and losers?
This legislation marks a significant shift away from federal support for clean energy, climate action, and environmental justice, while enhancing support for traditional energy, large-scale agriculture, and business investment. The winners are primarily established industries and higher-income individuals/entities, while clean energy, environmental, and community-focused groups lose ground. Several groups face mixed outcomes, depending on future implementation and market responses.
And Investment winners and losers? Specifically, where would you put your money to make a buck?
I won’t post the full details, including the specific stock pickers, here because giving investment advice in newsletters has certain legal implications. If you ask in the standard tools, they’ll highlight high R&D companies, specifically tech. Here the answers were more focused on REITs, certain healthcare segments that would benefit through child tax credits, any banks or investing conglomerates focused on rural economies. The shift is noted here.
What do the Receipts Teach Me?
Well, we’re fallible. Stories matter. And, the details are important, even the ones we lose on the cutting room floor, because they strip out the emotion.
Yes, summaries are fairy tales. But the receipts are the true reality even if they’re in that damn, tiny font. Thanks for reading. Onward.
Notes and the Left-overs:
Do we really need an all-powerful speaker leading the most powerful branch?
I plan to open-source the application I used to analyze the One Big Beautiful Bill, launching to subscribers later this year. There is a certain amount of work so it can be used/packaged for ease that takes a little time.
Originally, I drafted this comparing The One Big Beautiful Bill to the Inflation Reduction Act; both being a wishlist of stuff without a unifying theme. Technically, Republicans or Democrats could vote for either blindfolded, they are a canary in the coal mine because everyone will find something you like or hate depending on party affiliation.
Note, I left off a few key pieces of legislations including The Dodd Frank Act, Patient Protection and Affordable Care Act, and the Cares Act. They were either more about outcomes versus process. Or too new to analyze long term impact.
Headlining picture comes from my own hand, taken at the Tennessee Capitol. Yes, I ran it through a half-dozen photoshop filters.
I’m using Nancy’s quote to prove a point. Note, it’s been argued that this is presented out of context.
I’ve do run all articles through a plagiarism checker, two different tools. Grammarly also checks for AI. Interesting that the citations above weren’t detected, probably the RAG model changes the writing tone.
Be Cool, Pass The JPLA On …
Baseball Analytics (What I’m Reading):
The Athletic has a detailed piece on the analytics behind the mythical rising fastball. I loved this article, a worthy read. Spin rate. Speed. Trajectory.
For a St. Louis Cardinals updates, my initial model had them a game under .500 and they are well behind the pace. I’m not sure this team is worth a more detailed post mortem review, but I’ll probably build a factor into next year’s approach on being a buyer/seller at the deadline. Personnel change does impact results, and I didn’t have that built into my thinking. In fairness, they’ve been consistent the past few years, often improving in the second half.
Email Management (What I’m Tinkering With):
I don’t know why, but I change email applications often, bouncing between Apple Mail and Outlook. There are a host of other technologies out there too. But, I’ve started to notice more want to run your email through their own services for AI enhancements. That’s all well and good, but it does open up privacy issues. Do I really want Microsoft having access to my Gmail? And vice-versa?
This led me to Thunderbird. What I love about it is that my email stays local, leveraging IMAP. Also, I can customize the heck out of it from colors to fonts. Now, this is challenging to setup, but Claude or ChatGPT streamlines this by building an .xpi file based on your preferences and fonts.
Free your mailbox.
Competing Break-up Albums (What I’m Listening to on the Morning Run):
Jason Isbell released Foxes in the Snow in March, his first solo effort since the acclaimed Something More Than Free (considered by many to be one of the top albums ever produced). The new release has multiple winks and nods to his ex-wife, Amanda Shires.
Well, she released her response, which might be the break-up album. I know, that’s a tall order because of the competition. Rumours. Blood on the Tracks. Jagged Little Pill. Still, it’s beautiful, haunting, and her lyrics, the rhyme and melody, prove she’ll never be erased.
When Words Cut:
“He erases the details in our history. No matter how clear I keep the memories. He rewrites them so he can sleep.”
