Week 51 · 2025-12-15 → 2025-12-21 · 47 newsletters

Last Working Week Of The Year

year-end-retrospectives · ai-workflows-and-bottlenecks · product-and-thinking · macro-and-markets · nyc-and-grace-notes

The last full work week of 2025. Fifty-four emails across seven days, weighted toward year-end retrospectives and 2026 outlooks, with a cluster on AI coding workflows and a smaller spine on product thinking. No dominant news story. The writers who used the week to take stock did the best work.

Year-End Retrospectives: Honest Ledgers Beat Curated Highlights

The dominant form of the week was the year-in-review, and the spread between strong versions and boilerplate ones was wide. Nikunj Kothari at Balancing Act wrote the standout with "Plugged into the Matrix," a confessional on what two years of being extremely online, writing 72 Substack posts, and running the VC allocation game has done to his head. He is 35, has two kids, and admits much of his writing comes from anger at people wasting their time. His resolution for 2026: less dopamine, more depth, back to long form. Ben James at Ben by Fax ran the inventory version: a live 3D Tube map, a prototype solar data center, 75 Feathers McGraw replicas around London, a Claude-terminal typewriter, and a fractured jaw from a bike fall. The format is honest because it grades itself.

Ben Cmejla at TheGP ran the most operational year-end: a quarter-by-quarter walkthrough of TheGP's bi-weekly AI debates, with predictions that look obvious in hindsight. Figma's David Kossnick called the current moment "the Telnet days of AI interfaces" in January. Charlie Labs' Riley Tomasek predicted the death of editor-based coding in February. Alec Flett called that "MCP-first" would soon mean what "API-first" used to, with MCP now at over 10,000 public servers and 97 million SDK downloads. That receipts-based retrospective is the version of the genre that earns its slot.

Conventional best-of pieces came from Kyle Poyar at Growth Unhinged, Emily Kramer at MKT1, Yue Zhao at The Uncommon Executive, and Jared Blank's Gobbledy Marketing Awards. Kyle polled 130 readers on their top growth experiment; outbound/ABM and partner/ecosystem tied at 20%, well ahead of anything AI-specific. Eater NY ran its dishes of the year list. Wendy MacNaughton at DrawTogether was the most direct about how the week actually felt: "most of us are ready to wave a big fat goodbye to 2025."

The take: the retrospectives that worked were the ones where the writer had to confront something. Nikunj's anger, Ben James's ship list, Cmejla's predictions audit. The ones that curated highlights without grading them faded fast. If you are writing your own 2026 wrap, pull the receipts.

AI Workflows And Where The Bottleneck Moved

The largest single thread of the week was operators writing about how they actually use AI in production. Addy Osmani ran the most thorough version with "My LLM coding workflow going into 2026," opening with the data point that ~90% of Claude Code is now written by Claude Code itself, and walking through a spec-before-code workflow that treats the LLM as a pair programmer requiring direction and oversight. Ethan Mollick at One Useful Thing extended his "Jagged Frontier" framing, arguing LLMs lack permanent memory of new tasks and even supersmart AIs may never fully overlap with the surface of human work.

Daniel Pupius at TheGP wrote the cleanest essay of the cluster. His GitHub integration triages issues with Claude and either produces a mini spec or a working PR; his role is no longer writing code but deciding what gets merged. When code generation gets cheap, the constraint shifts from "what can we build" to "what should we release," and the apparatus built around story points and sprints stops making sense. Noah Brier at Alephic ran the lightweight version with Aesthete, a vibecoded site of nearly 200 aesthetically satisfying brands. Paul Stansik at Hello Operator launched an "AI 101" page focused on practical prompts, on the premise that too many people are playing with AI and not enough putting it to work.

Ben James ran the hardware-hack version: converting an old typewriter into a Claude terminal, with an Arduino and Raspberry Pi Zero so Claude can phantom-type back on the original keys. Seeing an LLM physically manifest as a clunking mechanical thing felt like rediscovering LLMs for the first time. Taylor Majewski's General Podcast episode with Merrill Lutsky of Graphite and Zach Lloyd of Warp pinned down why code review is becoming the real bottleneck as agents generate more code. Graphite was acquired by Cursor the day the episode dropped.

The take: the AI conversation in the operator class has fully shifted from "what can it do" to "where does the human judgment sit in the loop." Addy, Daniel, Ethan, and the Graphite-Warp episode converge on the same point. If you read one piece, Pupius on the bottleneck moving from execution to governance is the cleanest statement of what 2026 looks like inside a team.

Product, Procrastination, And The Long Game

Shreyas Doshi ran three posts, unusual for him, and together they form a coherent argument. "The Product Builder's True Journey" makes the case that every builder lives two lives: first you just build, then you internalize that you are going to be shackled by what you build, and the second life is when you ship more useful things more often. "Why Products Fail" was the audio companion. "On Beating Procrastination" closed the trio with a seven-step protocol for working through latent fear. The thread across all three: the work is not technical, it is emotional.

Nikunj Kothari opened the week with "Liar's Valuation," a primer on how bookings, revenue, and ARR get blurred at AI companies and why even sophisticated people sign offers without understanding what they were given. The line that lands: your time is the only asset you can't make more of, so you should know what you're trading it for. signull ran a short piece called "design collisions" arguing ideation is not an activity but an event, that real ideas come from collisions between your psychology and the world's need at the right time, and that scheduled ideation produces only a simulation of an idea.

The take: Doshi names the fear, Kothari names the trade, signull names the difference between performance and event. Read together in the slow week, they form a single argument.

Macro, Markets, And A Few Real Numbers

Matthew Klein at The Overshoot anchored the macro thread with the cleanest read of the week: typical American worker wages are still rising at about 4% annualized, vs. ~3% in 2018-2019, enough to explain why inflation is running about 1 point hotter than pre-pandemic. The BLS lost six weeks of data collection to the October shutdown, the job market still looks healthy, and the case for rate cuts is weaker than the consensus assumes. Citrini ran "26 Trades for 2026," framing the annual-outlook format as a thematic watchlist on the premise that the value is in surfacing blind spots rather than calling tops.

Craig Kennedy at Navigating Russia flagged the geopolitical signal of the week: new US, UK, and EU sanctions on Russian oil are the most significant since 2022, with estimates Russia could be forced to shut in 1.6 to 2.8 million barrels per day, the worst energy-sector crisis there since the 1990s. Stonebridge Capital followed its Vertiv writeup with a Modine pitch, arguing the 100-year-old auto-radiator company is quietly pivoting to hyperscale data-center liquid cooling and is mispriced because screening tools still classify it as Auto Components. Tech Buzz China ran "Temu Watch 11" arguing profitability has finally arrived. Charlie Liu at Fintechnize flagged Coinbase's System Update and the deepening Robinhood-Coinbase battle.

The take: the real macro story is sticky wage growth keeping inflation stickier than the rate-cut consensus wants to admit. Klein is the one to read on it. The Modine pitch is the cleanest example of the find-the-AI-thematic-without-paying-the-AI-multiple trade that will be a 2026 conversation.

NYC, Stand-Up, And Grace Notes

David Federico at East of the Bowery ran two LES dispatches. Bolzot, a Mongolian bar with vodka and comfort food, opened at Monroe Street in Two Bridges. SC103, an indie fashion brand from Pratt grads Claire McKinney and Sophie Andes-Gascon, opened December 12 at 25 Henry Street with a Vogue write-up. The bittersweet companion: Maryam Nassir Zadeh is closing its Norfolk Street store after 17 years, continuing only at the Paris location.

Alec McNayr wrote the most personal piece of the week on starting stand-up in his mid-40s, sparked by the Will Arnett movie "Is This Thing On?" His critique: the second open-mic set always goes worse than the first, because the first time you stumble into honesty and the second time you are trying. Two years in, he is still shaking the hard-to-watch state of trying on stage. George Mack at High Agency ran "The Busy Trap" as 11 quick thoughts, sharpest among them Taleb's "if you're busy at work, odds are you will eventually be replaced by a robot" and Tversky's "you waste years by not being able to waste hours."

Brianna Zuniga at Circular Architect wrote the most striking piece of the week with "nightlife taxonomy," on nightlife as a woman being two nature documentaries in parallel: prey animal tracking exit routes, and David Attenborough calmly observing mating displays. Both prosecutor and witness. The dual awareness arrives when the female prefrontal cortex finishes developing around 25, also peak fertility, the ultimate cosmic joke. Mishka Makes Food ran a ginger-scallion white bean salad. Liz Tran wrote on how long change takes. Peter Teague published Sally Vance-Trembath on summoning heroes when monsters are rampant.

The take: local writing and personal essays carried the back half of the week more than industry writing did. Zuniga, McNayr, and Federico ran three different formats and all earned their slot.

Three Takeaways from the Week

The best year-end posts of the week were the ones that had to confront something the writer could not gracefully edit around. Nikunj's anger. Ben James's ship list with the fractured jaw on it. Cmejla's quarterly receipts. If you are still drafting your own 2025 wrap, pull the predictions you actually made and grade them.

The largest operator conversation was where the bottleneck moved as code generation got cheap. Pupius, Osmani, Mollick, and the Graphite-Warp episode are running the same play: the constraint is no longer building, it is governance, judgment, and release. Carry that frame into Q1 and the org-design questions look very different.

If you revisit three pieces, I would suggest Daniel Pupius on going AI native for what changes inside a team when code stops being the bottleneck, Matthew Klein on US wages and sticky inflation for the cleanest read on the macro setup heading into 2026, and Brianna Zuniga's nightlife taxonomy for the piece most worth reading slowly. The slow week is the right time to read writers who do not write quickly.