Week 52 · 2025-12-22 → 2025-12-28 · 20 newsletters

The Haze Between Years

year-end-reflection · payments-and-ai-agents · thinking-about-thinking · lower-east-side-notes

The week between Christmas and New Year's. Paul Stansik at Hello Operator calls it "the haze," and the inbox confirmed it: 23 emails across seven days, most of them year-end reflections, one big payments retrospective, two food recipes, two NYC neighborhood dispatches, and a small cluster of writers thinking out loud about thinking. No dominant news story, no breaking macro thread. The writing that landed best this week was the writing that admitted what week it was.

Year-End Reflection: The Best-Of Genre, Done Two Ways

The dominant form of the week was the year-in-review post, and the spread between the good ones and the perfunctory ones was wide. Nikhil Basu Trivedi at next big thing did the version with teeth: he pulled his seven 2025 predictions from a year ago and graded them honestly. AI-Native Applications got an A, with his observation that we have "almost become numb" to twenty-person teams hitting $10M ARR being the line that stuck. AI Workers also got a high mark. The grading exercise is the format at its best, because the writer has to confront where they were wrong, not just curate where they were right.

Paul Stansik at Hello Operator did the other strong version: pulling his Substack stats and asking which of his own pieces deserved a second read. Most Liked was "What My Dad Taught Me About Work," which he says took longer to write than anything he has published. Most Shared was "How To Make ChatGPT Sound Like You." The honesty about which of his own pieces still hold up, and the framing of the haze itself ("lunch is a combination plate of leftovers from three different meals"), made it the most readable holiday post of the bunch.

Clara Ma at Ask a Chief of Staff ran the more conventional best-of for her 13,500-subscriber community. The All American ran a one-paragraph "we're on break, see you in January" note, which is its own kind of honesty. Steph Mui admitted upfront that her newsletter would be shorter for two weeks because she wanted less screen time over the holidays. Liz Prueitt at Have Your Cake announced she was taking January off after eighteen months of weekly posting to "reset."

The take: the writers who admitted out loud that this was a low-output week wrote better posts than the writers who pretended otherwise. The genre rewards specificity (Nikhil's grading, Paul's stats) and rewards saying nothing when there is nothing to say. The middle ground, the dutiful holiday wrap-up, is where the genre dies.

Payments and AI Agents: One Real Retrospective, One Real Argument

The week's only substantive industry piece was Dwayne Gefferie's Payments Strategy Breakdown "Future Proof," a long retrospective arguing that 2025 was the year four parallel trends in payments matured at once: AI agents moved from demos to production, tokenization became default infrastructure, stablecoins crossed from crypto trading into real settlement, and cloud processors replaced legacy batch systems. The numbers he chose are the ones to remember: Visa provisioned 12.6 billion network tokens, stablecoins processed $26 trillion in on-chain volume, Mastercard committed to 100% European e-commerce tokenization by 2030. The October Agentic Commerce Protocol launch, with OpenAI partnering with Stripe and ChatGPT users completing Etsy purchases inside the chat interface, was the inflection event he flags as the start of agentic commerce proper. Google's Agent Payments Protocol followed.

The companion piece came from Alephic's Forward Deployed podcast, Episode 3 on context engineering. Noah and Lance dug into Karpathy's framing of context engineering as the discipline of "filling context windows with just the right information for agents to take the next action." The three buckets they offered (reducing context, isolating context, offloading context to file systems) are the right primitives. The Claude Diary concept (agent memory through reflection and evolving CLAUDE.md files) and the case for skills over MCPs through progressive disclosure are the operational takeaways. The line that stuck: "treat agents like humans works, because of dual-use tools and new coworker onboarding."

The take: payments and AI agents are converging faster than either industry's press is treating it. If you take Dwayne and the Alephic podcast together, the picture is that the infrastructure layer beneath agentic commerce got built in 2025 in the quiet, and 2026 is the year the consumer-facing applications either find product-market fit or do not. The bottleneck has moved from "can the agent transact" to "where does the human judgment live in the loop." That is the operator-level question to carry into Q1.

Thinking About Thinking: Three Writers, One Argument

Shreyas Doshi ran three pieces this week, which is unusual for him, and together they form a single argument about epistemics. The first, "On the Confusion Between Writing and Thinking," pushed back on the popular "writing is thinking" mantra by pointing out that for many people writing is the only time they think clearly, and so they overgeneralize from personal experience to claim that all clear thinking requires writing. The three errors he names (exclusivity, supremacy, definitional) are clean. The second piece was the audio deep-dive companion built on NotebookLM. The third, "The Problem with Peer-reviewed Studies on Human Behavior," made the harder claim: when it comes to behavioral research, "all that should matter to you as an individual is what works for you," because anyone sufficiently smart can design a study to prove any behavioral claim they want.

The companion piece was George Milton at Gross to Net on "You Don't Understand Fiduciary Duty," which is the same kind of argument applied to a different received wisdom. The widely-held belief that corporations are legally required to maximize shareholder profits, Milton argues, is not actually true under Delaware law or federal securities law. The DGCL purpose clause says corporations may pursue "any lawful act or activity," and the statute is deliberately written to give boards maximum flexibility. The misreading has been so pervasive that even people who should know better, including lawyers, repeat it as gospel.

Aditya Bhargava at Ducktyped ran the soft version of the same instinct in "One year of keeping a tada list," about replacing the to-do list with a to-done list and discovering that the practice surfaced not just accomplishments but the chain of dependencies between them ("I was able to do this thing, because I did this other thing earlier"). Dan Koe ran the more populist version with "How to fix your entire life in 1 day," arguing that 80-90% of resolution failures happen because people pursue change for status reasons rather than internal ones.

The take: Doshi and Milton are running the same play in different domains. Both are arguing that smart people anchor on received frameworks and stop questioning them, and the cost of that anchoring is invisible until you back into it from your own lived experience. The "what does the research say" reflex Doshi attacks and the "shareholder primacy" reflex Milton attacks are the same epistemological tic. If you only take one frame from the week into 2026, it is that one.

Lower East Side Notes and Other Grace Notes

David Federico at East of the Bowery ran two posts this week, both worth saving. The Sunday Dispatch is a new format he is trying, aggregating lesser-known LES sources (Lo-Down, EV Grieve, Hell Gate, Crain's, Gothamist) into a weekly neighborhood roundup. The grim notes: Frank Arroyo of Frank's Bike Shop passed at 81 in early December, Bluestockings closed December 27 with the landlord asking $14K a month for 2,500 square feet, and Morgan Stanley bought the 202 Broome NYU Langone space for $56 million. The brighter note came in the Wednesday post on Chin Up Bar, a new gin bar from Brian Gummert (owner of Subject) and Blake Walker (formerly of Subject and the pandemic-shuttered Nitecap) that opened December 16 at 171 Chrystie Street, with sky-mural ceilings by lifelong LES artist Ori Carino.

Brick at farmers market girl made the winter case for radicchio (tardivo, castelfranco, treviso), parsnips, and "daikons of every color." The argument that winter produce is "massively underrated" is a defensible one, and the sautéed-radicchio-with-pasta move is sound. Mishka Makes Food ran a smoky harissa chickpea stew built around two cans of chickpeas, a can of coconut milk, and pantry harissa, which is the right kind of recipe for a haze week.

Abby Falik at Taking Flight ran the week's most reflective piece, "Why do you walk so fast?", on her seven-year-old host cousin Carlitos asking her that question in Nicaragua in 1996 and the same question returning thirty years later on the Kumano Kodo pilgrimage in Japan. It is a long meditation on speed as conditioned virtue and slowness as the harder practice, and it is the kind of post that lands in the haze precisely because the haze is when you might actually slow down to read it.

Molly Graham announced a paid tier on a Substack that started "as an accident" four years ago and has reached 680,000 reads and 10,000 subscribers. Justin Mares at The Next announced Truemed's $34M Series A led by a16z and made the chronic disease argument that 50% of Americans have diabetes or prediabetes and 73% are obese or overweight. Ben Thomas at Roots of Progress ran the Progress Conference 2025 dispatch, with Afra Wang's "Physical dynamism and the immigrant's edge" essay on immigrant founders driving America's new era of physical building as the standout. Stonebridge Capital wrapped its three-week "Thermal Moat" basket on Vertiv and Modine as AI-data-center cooling plays.

The take: the local writing, the food writing, and the reflective writing carried the week more than the industry writing did. That is the right balance for a week between Christmas and New Year's. The newsletters that earned their slot were the ones that knew what week it was.

Three Takeaways from the Week

The best year-end posts were the ones that admitted what they were. Nikhil's graded predictions and Paul Stansik's "what did you actually read of mine" pull worked because they replaced the genre's default (curated highlights) with something the writer had to think about. The perfunctory best-ofs faded fast. If you are running an annual recap in 2026, the rule is to either grade yourself or step away from the keyboard.

The single live industry conversation in a 23-email week was payments meeting AI agents, and Dwayne Gefferie's "Future Proof" plus the Alephic context-engineering podcast are the cleanest reads on where the infrastructure actually sits at the start of 2026. The agentic commerce protocols launched in October are the wedge; the context-engineering primitives are the operator-level question. Both are quiet right now and will not be in six months.

If you only revisit three pieces from the week, I would suggest Shreyas Doshi on the confusion between writing and thinking for the cleanest epistemic frame, Dwayne Gefferie's "Future Proof" for the cleanest payments retrospective of the year, and Abby Falik's "Why do you walk so fast?" for the post most worth reading slowly in a week that finally lets you. The haze is short. The pieces that survive it are the ones that earn the slowness.