Friday, January 31, 2025 · 3 newsletters

The Quiet Month

sparse-inbox · financial-infrastructure · agentic-commerce-foreshadow · deepseek-rupture · subtraction-as-strategy

Thirty-two emails across thirty-one days. That is the headline of the month, and any honest read of January has to start with it. The newsletter ecosystem that I will later build a daily briefing on top of did not exist yet at any meaningful density. What I had was a handful of writers with strong internal calendars, one or two operator voices who would later become regulars, and a week-five DeepSeek event that I now know was the most consequential AI story of the year, written about by exactly one person in my inbox.

The Month in One Sentence

This was the month the inbox was small enough that you could hold every signal in your head, and the signal still got something important right.

Arc: Financial Infrastructure as the Actual Subject

The cleanest accidental convergence of the month came in week two, when two writers from completely different beats showed up with long essays arguing the same point: that the rails underneath money are the actual subject, not the money itself. Matt Brown at Matt's Notes wrote "Stablecoins in 1,000 words," opening with a Trans-Siberian Railway memory of waking up at 3 a.m. on the Mongolian-Chinese border to find workers swapping his carriage's wheel assemblies for a different rail gauge. Same passengers, same car, different rails. The essay does not tell you what stablecoins are; it tells you why the question matters. Money and the financial infrastructure it rides on are separate but related elements, designed symbiotically, and you cannot reason about one without reasoning about the other.

The companion read from the opposite polarity landed the same week. Craig Kennedy at Navigating Russia published the executive summary of his "Russia's Hidden War Debt" report, with the headline finding that Moscow has been running a dual-track financing strategy since the second day of the full-scale invasion: the highly scrutinized defense budget on one track and a compulsory bank-lending scheme on the other. Kennedy's estimate was $210 to $250 billion of an unprecedented $415 billion surge in corporate borrowing since mid-2022, all off-budget defense financing routed through state-pressured banks. Where Brown is writing about how new rails can extend what money does, Kennedy is writing about how a state can quietly bend existing rails to fund a war off the books. Same essay, opposite valence.

By month-end the frame was the lens. Read in retrospect, the financial-infrastructure arc is the one the rest of the year would extend into agentic commerce protocols, into the de minimis trade story, into the question of whether the dollar's privileged position is a function of policy or plumbing. The reason these two pieces stuck is that they treated infrastructure as the load-bearing question rather than the boring footnote. Most writing about money in 2025 will not. The two that did showed up in the same five-email week of an empty January, and the coincidence is exactly the kind of thing a denser inbox would have buried.

Arc: From Quiet-Week Macro to a Real Inflection

The macro coverage in January ran almost entirely through one writer, The Last Bear Standing, and watching the arc inside one byline is its own lesson. Week two opened with the most honest market note of the month: "What did you miss in this first, abbreviated, trading week of the year? Not a whole lot, to be frank." Meme stocks meme'd, indices stayed in their slump, ADP came in soft, claims were strong, ISM was positive, consumer credit declined. The Bear teased a contrarian capex piece and named the quiet rather than manufacturing a thesis. That is a discipline, not a default.

Week three was the inflation essay. "Easing Into Expansion" opened with a Snickers bar and the question of why it costs $2.67, then ran the argument that the irreducibility of the price (cocoa futures, manufacturing, wholesale, retail, taxes, profit motive, productivity, inventory, local costs, human preference, the immeasurable elasticity of supply and demand) is the actual problem with inflation policy. The market had fallen back on obsessive observation of monthly prints, slicing them into ever-smaller components and measuring against expectations in basis points. The Bear's worry was that the observation game had replaced the underlying question. The framing did not predict any single print. It said: the framework is the trouble, not the data.

Week four added the nuclear narrative. "The Nuclear Narrative" catalogued the early-2025 run-up: four of the eight best-performing large-cap stocks were nuclear plays (Oklo, Constellation Energy, NuScale, Vistra), up 32% to 75% in three weeks. The catalysts were real, with the Palisades restart backed by federal and state loans, the Amazon-Talen Susquehanna deal at 650MW on a 20-year PPA, Microsoft's Three Mile Island PPA with Constellation, Oklo's 750MW in LOIs from unnamed datacenter customers. Datacenter capex had given the power sector its first long-term bullish demand story in decades. The Bear's posture was the right one: not dismissing the thesis, separating the catalysts from the narrative pricing.

Week five was the rupture. The Last Bear Standing on DeepSeek was, in retrospect, the single most important macro read of the month. Two Chinese researchers released Deepseek V3 and R1, large language models that mirrored American frontier capabilities with a headline $6 million pre-training cost (the true buildout closer to $500 million in GPUs) and full open-source transparency. NVIDIA dropped, the AI trade wobbled, the narrative scaffolding underneath two years of US equity outperformance got its first real stress test. The Bear was the only writer in my inbox that week to treat it as the most important week of the year so far. He was right. The scaling-laws thesis that anchored American AI investment since the January 2020 OpenAI paper had its first credible counter-example, and the cost structure that justified the buildout was now contestable. Almost nobody else I was reading engaged with it seriously. The lesson is not about coverage; it is about who you read when the signal is sparse.

Arc: Agentic Commerce, Twelve Months Early

The forward-looking AI essay of the month came from Sahar Mor at AI Tidbits in week three with "Rewiring the Internet: Commerce in the Age of AI Agents." The vignette was a December 2028 scene of an agent named Maria's running simultaneous negotiations with twelve vendors for her daughter's birthday party, securing a nut-free cake after verifying certifications, booking an entertainer with safety ratings, coordinating custom goodie bags after checking allergies with the other parents' agents, all under budget. The point was not that the scene was imminent. The point was that the infrastructure required to make it possible was the actual subject worth writing about.

Sahar's anchor was OpenAI's Tasks release, the consumer-facing wedge. The plumbing was agent passports and trust protocols he had covered in earlier posts. The thesis was that agentic payments, agentic marketing, agentic support, and agentic localization were not separate domains but four faces of the same shift, and the writers who treated them as separate would miss the convergence. He framed it in January 2025, before the agentic commerce protocols that would actually ship in late 2025. Reading the piece in real time, the right bet was on the infrastructure layer, not the chat interface. The DeepSeek story arrived in the inbox eleven days later and confirmed the broader read: the model layer was about to commoditize while the orchestration layer was where the durable work would happen.

Arc: Subtraction as the Operator Discipline

Two operator threads ran through the back half of the month, both arguing for restraint. Ami Vora at The Hard Parts of Growth made the case twice, once on simplifying product strategy as competitive advantage and once on borrowing existing platform patterns rather than inventing new ones. The line worth saving: a knife works because everything unnecessary has been removed, and that is literally what sharpening a knife means. For WhatsApp, the only question that mattered was whether calls and messages went through, every time, for free. Everything else got resourced after.

Henrik Werdelin ran the personal-operating-system version, introducing his 8+1 Framework of eight life buckets plus a weekly review. The more interesting half of that post was the aside about "thinking in products," prompted by hearing someone talk about replacing Figma with Replit and jumping straight from idea to prototype. His question of how to teach the skill of asking "what is the smallest, testable version of this idea?" is the right one for any team running on AI-augmented tooling. The 8+1 frame and the Vora simplicity frame are the same argument applied at different scales. The temptation to add features, frameworks, and commitments scales with status, and the operator skill that does not get easier with seniority is the discipline of subtraction.

Julie Zhuo at The Looking Glass closed the month with the management version in "The Valuable Employee Paradox." Every great manager she knew said the reports they valued most were the ones who convinced them to do things differently, but most reports believed they were most valuable when they did what their manager wanted. The reconciling frame was the midwit curve: worst reports do what they want against the team's interest, average reports do what their manager says without pushback, best reports do what is right for the team even when that means disagreeing. The Vora-Werdelin-Zhuo cluster is the operator backbone of the month. All three arguments come down to restraint, which is the rare register that does not get easier with seniority.

The Story of the Month

The story of the month, in retrospect, is the DeepSeek release on the last Monday of January, but the framing that holds up best is that almost nobody in the inbox treated it as the story while it was happening. The Last Bear Standing did. One newsletter, one writer, one calibrated read. The headline cost figure was the loud part; the thesis revision was the quiet part. Loss may scale as a power-law with model size, dataset size, and compute, but power-laws cut both ways once efficiency breakthroughs arrive. That sentence, written in week five, was the cleanest summary of the year's most consequential single industry event, and it was published into an inbox so sparse that I could read it the morning it landed and feel the shape of it before the market open.

The case for treating DeepSeek as the month's story is structural rather than dramatic. The Brown-Kennedy financial-infrastructure pairing in week two argued that rails matter more than what runs on them. The Sahar Mor agentic-commerce piece in week three argued that the orchestration layer was where the durable work would happen. The Vora-Werdelin-Zhuo subtraction cluster argued for restraint as the operator discipline. DeepSeek closed the month by demonstrating, with one open-source release, that the entire American AI capital stack had been priced on the assumption that frontier capability required frontier spend. The thesis got its first counter-example in week five. Read the four arcs in order and the month is one long argument that the infrastructure under the story is the actual story.

In Retrospect

The first-week resolution-year manifesto aged interestingly. Justin Mares at The Next opened January with a flag-planting post on the incoming HHS team and the MAHA agenda, declaring he had "never been so bullish that we are going to make America healthy again." Mares put himself on the line in writing on January 2, which is a register almost nobody else used at the start of the year. The honest call on his post is that it set a calendar against which the rest of 2025's health-policy coverage will be measured whether the other newsletters know it or not. Whether the rhetoric held or the chronic disease numbers actually moved is a different question. The bookmark is the point.

The first-week-back market commentary genre held up better than expected. The Last Bear Standing's week-two refusal to manufacture a thesis ("Not a whole lot, to be frank") is the kind of move that looks like nothing at the time and reads as discipline in retrospect. The genre default is to puff up the empty week; the alternative is to say so. The same writer's week-five DeepSeek read is what that discipline buys you: when the actual signal arrives, you have the calibration to recognize it.

The two essays on attention and media collapse from week five looked like a coincidence and were a diagnosis. Jenny G. Zhang on the talent hierarchy inside collapsing newsrooms and Abby Falik at Taking Flight on the smartphone-as-cigarette parallel were the supply-side and demand-side versions of the same essay. They ran on opposite days of the same sparse week. The interesting question neither piece quite asked is whether the post-collapse equilibrium produces something better or just smaller. Twelve months later, the answer is starting to look like "smaller, and the writers who survived the contraction by serving a real audience are the ones building inside the new equilibrium." The pieces read as overstated at the time and underrated in hindsight.

The Trump-conflicts beat got handed off on day one. Zach Everson at 1100 Pennsylvania opened week four with the Mar-a-Lago accounting: at least 16 billionaires, officials from eight foreign countries, 14 Cabinet picks, 22 governors, eight senators, 65 members of Congress visiting Palm Beach since November. The buried lede was that the government of Israel appears to have paid the Trump Organization to host the Jerusalem Prayer Breakfast at the club on January 14. The 2017 version of that detail would have been a two-week story; the 2025 version was a Monday bullet point. Everson started the term with the vocabulary already built and the receipts already organized, which means the press cycle on emoluments will compress significantly relative to the first term. That compression is the metastory.

What to Carry Into Next Month

The month's structural lesson is that a sparse inbox is not a problem to solve; it is a state to read accurately. Thirty-two emails across thirty-one days is a real constraint, and the right response is not to puff up what is there but to notice which writers showed up with a real beat, a real calendar, and a real point of view. The Brown-Kennedy convergence on financial infrastructure, the Sahar Mor anticipation of agentic commerce, the Vora-Werdelin-Zhuo cluster on subtraction, and the Last Bear Standing's calibrated read on DeepSeek were the four threads that earned their slot. None of them depended on a dense news cycle. All of them will compound across the year.

The DeepSeek event is the inflection that the rest of the year will be measured against, and reading the inbox in retrospect tells you something useful about how to read the next month. The most important industry story of the year was covered by one newsletter in the week it happened. The lesson is not that I needed more newsletters; the lesson is that the writers who can hold a calibrated read while the market is shouting are the ones to weight heaviest. The Last Bear Standing earned a permanent spot in the rotation on the strength of that one piece. The next month's job is to notice which writers earn the same kind of trust on the next inflection, which will come whether the inbox is dense or sparse.

If you only read three pieces from January, I would point you to Matt Brown's "Stablecoins in 1,000 words" for the cleanest single framing of why financial infrastructure is the actual subject and not the footnote, Sahar Mor's "Rewiring the Internet" for the forward-looking essay that called the shape of the agentic commerce wave a year before it broke, and The Last Bear Standing on DeepSeek for the one calibrated read on the year's most consequential industry event, written in real time into a five-email week. Three pieces, three altitudes, one quiet month. The signal was there. The trick was reading it.