Monday, March 31, 2025 · 1 newsletters

The Quiet Before The Vortex

sparse-inbox-coherent-signal · regime-change-in-equities · proof-of-work-as-human-need · ai-from-demo-to-org-chart · state-level-policy-over-federal · writing-as-the-act-of-thinking

Six weekly wraps, an inbox that never hit triple digits, and a month that started with a writer admitting his slow-down program was still being driven by the part of him that refused to stop, and ended with the Rose Garden tariff announcement landing as a wrecking ball. The month did not feel loud while it was happening. Looking back, it was the quiet before the regime change.

The Month in One Sentence

March 2025 was the month a thin inbox kept telling the same story from different rooms, that the structures we built to keep moving were the things in the way, and then the macro tape arrived in the last week to make the inside argument an outside one.

Arc: A Markets Regime, Called Three Weeks Early

The cleanest single arc of the month belonged to one writer. The Last Bear Standing rang the bell on February 28 with "Max Stupid," called the regime change on March 9 with a piece on the technical breakdown, sketched the bifurcation thesis on March 30 with "Haves vs. Have Nots," and was already collecting on the call by April 6 with "The Vortex." Reading the four pieces in order is the cleanest single-author market arc I tracked all month.

Week one was about a momentum break. The S&P had fallen 5% off its all-time highs. Palantir, Robinhood, Hims and Hers, and Applovin were yacking up their blow-off tops. NVIDIA beat earnings on Wednesday and got punished with an 8.5% plunge on Thursday, trading back to its DeepSeek lows. The line that landed: when good news is punished, expectations have run ahead of reality. Bitcoin was swimming with cement boots. The reflex rallies were still showing up, but they were getting weaker.

Week two was the technical confirmation. The S&P 500 closed on its 200-day moving average for the first time in eighteen months; the Nasdaq 100 broke its 200-day for the first time in two years. ADP reported the weakest payroll gains in seven months. The Challenger report logged the largest monthly job cuts since COVID. The Atlanta Fed's Q1 GDP nowcast sat at negative 2.4%. Tariff flip-flops were getting called "less 3D chess and more flailing ad-lib." The framing essay called it a regime change, not a dip, and the three-way confirmation (technical breakdown plus macro deterioration plus policy noise) was the kind of convergence that usually marks one.

Week five was the bifurcation read. Citing Citrini Research's Macro Memo, the argument moved on from "is there a recession" to "what kind." The top 10% of American households now account for roughly half of all consumer spending. The dominant risk is no longer a labor-market recession but a wealth-effect recession driven by asset prices falling and the top decile pulling back. That is a different recession than 2008 or 2020. It is one where the median household is structurally weaker but the tape only turns when the upper-quintile portfolios crack.

Week six was the timestamp. The Rose Garden tariff announcement landed on April 2, dropped inside week six of the month's reading. The framing essay was already written: this is not a negotiating posture, this is a wrecking ball, this is intended to dismantle the status quo. The median stock peaked in December. DOGE took office in January. Tariff-talk arrived in February. Soft data started screaming in March. By April, we had passed the point of no return. When the writer who has been calling the turn for months says we are past the point of no return, and the news catches up the same week, the piece becomes the timestamp on the regime, not one more market note.

Arc: Proof of Work as the Defining Human Question

The most cohesive non-market thread of the month was a slow-building cluster of writers, none of whom were talking to each other, all of whom were circling the question of what kind of effort still belongs to us when the tools get easier. The arc started inside, with a writer running a slowdown program admitting he was still moving too fast, and ended outside, with a series launch that named the question directly.

Week one was the inner confession. Steven Schlafman at Where the Road Bends wrote "Slow Enough to Matter," about running Downshift while still being driven by the Achiever who refused to stop. He called the wound "ever-becoming," and traced it to the fear that at 45, if he let go of momentum, he would fade into irrelevance. Jenny G. Zhang wrote a quieter version of the same instinct in "The truth about writer's block," looking back at three novellas written by age 13 and then six years of silence, having internalized the lesson that writer's block was a thing that happened to writers. The grace note was Tanya Windman's profile of Kettie Jean, with Kettie's line that her humanitarian work taught her the emotions and connections between people are far more powerful than the big issues. Three writers, one move: a story about how serious work happens is what is keeping us from doing it.

Week three was when the question got a name. Julie Zhuo at The Looking Glass opened a series on AI and how it might change us, and the first post, "Our Souls Need Proof of Work," was the one to read. Her frame: ignore the tired hard-work-versus-YOLO debate and notice instead that our happiness seems to require effort we can feel, work that is deeply personal because it is hard for us specifically. The implicit threat from AI is not that it takes our jobs, it is that it takes our friction. Henrik Werdelin ran the companion read almost by accident, in his "AI Perfection Paradox," landing in the same neighborhood: when AI makes flawlessness effortless, imperfection becomes the premium signal, like aristocrats in tattered coats. The Winning Therapy posts ran the Jensen Huang gardener story from the other direction. After 30 years tending one small basket of dead moss, the old gardener told Huang he had plenty of time. Huang called it the best career advice he ever got.

Week four was when writers started naming the activity, not the artifact. Jack Butcher returned to Substack after a three-year silence with "Write because you have nothing to say," and the claim was the inverse of the post he had abandoned the feed with: the act of forcing yourself to write is the only thing that gives you something to say. The Lingthusiasm episode on Sapir-Whorf was the companion read, landing on the point that the activity precedes the insight. By week five, Steven Schlafman closed the loop on his own arc with "The Gift of Not Being Needed," about waking up with COVID on the morning of the long-planned Downshift team off-site, watching the retreat he had built start to slip, and learning what it means to be sidelined from the thing you have made central. The arc that opened week one with a writer admitting he was still moving too fast ended week five with the same writer being forced to stop. The forced stop was the gift the program could not give him.

Arc: AI From Magical Demo to Org-Chart Surgery

The AI conversation moved a notch this month, and the move was small but durable. Early in the month the writing was about builder tooling and the awe-versus-bubble debate. By the last week of the month, three writers were independently circling the same harder question: not what the models can do, but what they do to the teams, stacks, and interfaces around them.

Week one was builder tooling. Sahar Mor at AI Tidbits ran a roundup that flagged Anthropic's foundational patterns for building effective agents, LlamaIndex's Agentic Document Workflows, DeepSeek's prompting findings for reasoning models, and PDF and web-automation tooling including francedot/acu and the OpenAI playground preset for structured table data. The single most useful operator tip was Anthropic's counterintuitive note that placing long documents (20K+ tokens) at the top of a prompt, before the query, lifts Claude's accuracy by roughly 30%. The piece was a builder's note. The frame was that the picks-and-shovels names were getting repriced while the builders were quietly compounding.

Week two was the measurement question. Mark Humphries at Generative History wrote the month's most useful single AI piece, on how to measure LLM performance in the humanities and social sciences, where there are no right answers and the existing benchmarks have run out of room. Math, science, and coding benchmarks were getting saturated. There was no equivalent measurement for qualitative analysis, argument, use of evidence, or prose style. Without that, the public AI discourse split into two unfalsifiable camps. The piece did not resolve the question, which was the point. The argument that the absence of good qualitative benchmarks was the story stayed useful all month.

Week three was the surface-versus-substance read. Shruti Gandhi at Array argued in "The Infrastructure Behind AI's App Layer" that most so-called app-layer investments are really infrastructure plays in disguise, because the real engineering work in any defensible AI product is in data pipelines, fine-tuning, and integration frameworks, not in the UI on top. The shiny chat interface is a costume. The differentiation is underneath. Read alongside Julie Zhuo's "Our Souls Need Proof of Work" the same week, the two pieces are arguing the same observation from opposite sides. Underneath the app layer, the real engineering is plumbing. Underneath the AI productivity pitch, the real question is what kind of effort still belongs to us. Both observations age well.

Week five was the org-chart move. Julie Zhuo co-wrote "The Death of Product Development as We Know It" with Henry Modisett at Perplexity, arguing the classic EPD stool and the mocks-then-PRD-then-build pipeline are both dying. When a single operator can prototype a full flow in an afternoon, the gating function on a product team stops being "can we ship it" and starts being "did we pick the right thing to ship." Matt Brown ran the structural companion in "Vertical AI: Beware What You Wrap," arguing the real question for any vertical AI startup is which layer of the stack you sit on top of. Wrap a system of record and you become a feature the incumbent ships next quarter. Build a genuinely new system of intelligence and you have a chance. Sahar Mor flagged voice as the next interface, with OpenAI's Advanced Voice Mode, Gemini Flash, and Sesame all crossing into real-time synthetic speech in the span of a few months. Three writers, one move. The conversation was no longer about awe. It was about operations.

Arc: State Capitols and Shipping Regulators Did the Real Policy Work

The week-three read that compounded the most through the rest of the month was the one almost nobody else was running. Justin Mares at The Next argued the MAHA story was at the state level, not the federal one. West Virginia led with a sweeping ban on artificial food dyes. Texas was voting on a similar bill the same week. Twenty-six states had legislation in flight. Americans were consuming about 15 million pounds of food dye a year, five times the 1955 number, with the average child getting roughly 45 milligrams a day. The Red 3 history was the detail to remember: the FDA banned it from cosmetics in 1990 after rodent cancer studies, then for 25 years took the position that the same dye was safe to eat. The state-level action was the story, not the press conferences.

Ben James ran the structural companion the same week with "How To Boil the Mediterranean Sea," the month's strongest climate piece. The framing: 253 million Hiroshima bombs of heat accumulated in the ocean between 2023 and 2024. The mechanism was mostly the 2020 shipping-fuel sulfur rule, which had removed an aerosol that had been masking warming. From May 1, the new Mediterranean Emissions Control Area would cut sulfur an additional 80%, which by his back-of-envelope had already added warming equivalent to all of France's historical CO2 emissions. Cleaner air, hotter oceans. The trade-off was honest. The two pieces shared a quiet structural argument: the policies that move the needle on health and climate are not the federal ones the press covers; they are state legislatures rewriting food chemistry and shipping regulators tweaking fuel rules.

The arc held up through the rest of the month and is the one to carry forward. Federal posture got cable news. State capitols and shipping regulators got the actual outcomes.

The Story of the Month

The story of the month is the regime change in U.S. equities, called three weeks before the rest of the press caught up. The Last Bear Standing's "Max Stupid," March 9 regime-change post, "Haves vs. Have Nots," and "The Vortex" form the single cleanest market arc I read in any inbox this year. The case is not that the writer was lucky. The case is that the technical, macro, and political signals were stacking week by week, and exactly one writer was willing to call them what they were before the news made the call obvious. By the last week of the month, when the Rose Garden announcement landed, the framing essay was already written.

The reason this is the story of the month, and not the proof-of-work cluster or the AI org-chart pieces, is that the regime change reframes everything else. The career-confession cluster that landed in week six (Alec McNayr admitting his upper-mid executive role inside Warner Bros Discovery was "literally and figuratively, clothing that didn't fit," Abby Falik writing "Trust the Bird, Not the Book") was not coincidence. Outside forces give people permission to ask inside questions. The reason the proof-of-work cluster compounded across the month was that the macro tape was quietly preparing the ground for it. The reason the AI conversation moved from awe to operations in the last week of the month was that the operators were starting to feel the regime change too. The market piece is the timestamp. The everything-else is the texture.

In Retrospect

The "sparse inbox is a quiet inbox" read aged poorly. Every weekly wrap this month opened with some version of "a thin week, eight to sixteen emails." The reflex was to treat the volume as the signal and conclude that not much was happening. The actual signal was that the thin inbox was unusually coherent. When the news flow is quiet, the writers who do publish are the writers with something they cannot not say. That is when the through-lines get easier to see, not harder. The right read going forward is to read a thin inbox harder, not faster.

The "bear case is still contrarian" read aged in nine days. The week-two post on the regime change in U.S. equities was framed cautiously: the bear case is no longer a contrarian position, but the convergence of signals suggests a regime change rather than a dip. By the last week of the month the bear case was not contrarian at all. By the first week of April it was the consensus, and the early callers were collecting. The right read in retrospect is that when three independent classes of signal (technical, macro, political) converge in the same week, the regime change is already happening, and the only question is the lag on the news catching up.

The SXSW dispatch as a one-off aged into a trend. Gabby Lord's week-three dispatch from SXSW, that it was "not worth the money," was treated at the time as a single writer's gripe. Read against the rest of the month, it is the third or fourth data point that the conference economy is being quietly re-evaluated by the operators who pay to attend it. The one-off becomes a trend when the rest of the inbox is showing the same instinct from a dozen different rooms. The 2025 conference economy looks like one of those quiet structural stories that does not get its own news week but is reshaping a whole layer of how operators learn and hire.

What to Carry Into Next Month

The regime change is the frame to carry into April. The Last Bear Standing's "Vortex" is the timestamp, but the harder work is the bifurcation thesis. A wealth-effect recession driven by the top 10% is a different recession than the ones the playbook is calibrated for. The interesting trade in Q2 is not growth versus value. It is wealth-effect-vulnerable versus wealth-effect-insulated. The fintech and embedded-finance pieces ran by Zarik Khan at Fintech Compliance Chronicles all month, including the Lauren Kornhauser interview on interpretable AI in credit and the Lauren McCollom interview on Grasshopper Bank in week six, are the kind of structural compliance writing that gets more valuable as the macro deteriorates. Track the state capitols and the regulators, not the press conferences.

The proof-of-work cluster is the frame to carry into the inner-work reading. If Julie Zhuo, Henrik Werdelin, Steven Schlafman, Jack Butcher, Abby Falik, and Winning Therapy are all independently circling the same question without coordinating, the question is real. The reading list for April is the writers who can hold both halves at once: that AI is going to keep making things easier, and that the human texture of choosing the harder thing is the part worth protecting. Not slowness for its own sake, but the recognition that some weeks the work is what you are not doing. The AI conversation has crossed an interesting line. The reading list should follow it across.

If you only revisit three pieces from March, I would suggest The Last Bear Standing on "Regime Change" for the cleanest market call of the year so far, Julie Zhuo's "Our Souls Need Proof of Work" for the inner-work frame that compounded across the month, and Ben James on boiling the Mediterranean for the climate piece that names the trade-off honestly. The month told me three things in sequence: the macro tape was breaking before the news caught up, the writers I read most carefully were independently circling the same human question about effort, and the AI conversation was finally moving from what the models can do to what they do to the people who used to do the work. Those are the three frames I am carrying into April.