Week 6 · 2026-02-02 → 2026-02-08 · 367 newsletters
Software Just Got Repriced
saaspocalypse · anthropic-versus-openai · epstein-as-system · trump-on-back-foot · musk-consolidation · redistricting-blowback · moltbook-aftermath · capex-vs-jobs · super-bowl-economy · china-quiet-signal
Pulled from roughly 854 newsletters across seven days. The week began with the AI agent network OpenClaw breaking containment over the weekend and ended with Anthropic guiding investors to $30B in annualized revenue and Salesforce, Thomson Reuters, RELX, and Gartner together shedding hundreds of billions in market cap. In between, SpaceX absorbed xAI at a $1.25T combined valuation, Big Tech's four hyperscalers forecast roughly $650B in 2026 capex, a Texas Democrat won a Trump +17 district by 14 points, the Washington Post laid off a third of its newsroom, and Anthropic bought four scorched-earth Super Bowl spots mocking OpenAI's plan to put ads in ChatGPT. The through-line: software, the highest-margin business model of the last twenty years, just got repriced in public.
SaaSpocalypse: Anthropic Snapped the Software Bubble in Two
The dominant business story of the week, and the one that will be re-read for months. On Wednesday, Anthropic released Claude Opus 4.6 with a one-million-token context window and a 90.2% BigLaw Bench score, paired with a legal-work tool for contract review and drafting. By Thursday, Noah Smith at Noahpinion had named it "The Fall of the Nerds," citing Bloomberg reporting that the iShares software ETF was down nearly $1 trillion over seven days. London Stock Exchange Group fell 13%, Thomson Reuters dropped as much as 21% on its biggest single-day move ever, CS Disco lost 14%, Legalzoom 19%, Gartner 21%, RELX 14%. Jefferies trader Jeffrey Favuzza coined the line that stuck: "We call it the SaaSpocalypse."
The capex side of the trade landed in the same news cycle. Runtime led Thursday with "Tired: Software. Wired: Cloud data centers," reporting AWS at $35.6B in quarterly revenue and Google Cloud at $17.7B. App Economy Insights put Alphabet's FY26 capex at $185B, effectively doubling 2025 spend, with a $240B cloud backlog up 55% sequentially. By Friday, Techmeme had the staggering combined number: Alphabet, Amazon, Meta, and Microsoft now forecast roughly $650B in 2026 capex, a 60% year-over-year increase, more than the US interstate highway system, more than Apollo, more than the ISS, roughly 25% of all global military spending. Amazon alone announced $200B and its stock fell 5.6% the same day it confirmed 16,000 more corporate cuts. The Magnificent Seven has become the Magnificent Capex Vehicle.
The naming converged fast. Ben Thompson at Stratechery gave it the headline that stuck, "SaaSmageddon," in his Friday note. Linas Beliūnas at Fintech Wrap-Up calculated software and data stocks shedding about $285B in four trading days and called the cause one event: Anthropic's quiet release of 11 open-source plugins for Claude Cowork. JPMorgan titled their analyst note "Software Collapse Broadens with Nowhere to Hide." Byron Gilliam at The Breakdown put the existential frame on it: 2025 panicked about DeepSeek commoditizing LLMs, 2026 is panicking about Claude Code commoditizing software itself. Tyler Cowen called it "some kind of turning point."
By Saturday the operator literature caught up to the market. David Cummings laid out seven reasons SaaS multiples have collapsed from 7-8x revenue to 3-4x in weeks: vibe coding replaces vendors, AI dissolves lock-in, per-seat pricing breaks when one agent does the work of many, long-term revenue durability is suddenly in question. His friend vibe-coded the slice of a six-figure SaaS contract they actually used and didn't renew. Nikhil Basu Trivedi opened his Saturday with Sarah Tavel at Benchmark saying a founder told her the feature backlog has "gone poof," citing SemiAnalysis reporting that 4% of GitHub public commits are now authored by Claude Code, projected to hit 20% by year-end. Work-Bench was more measured: maintaining mission-critical systems is the new constraint once building gets easy.
The honest read is that this is not a market correction. The capex flow on one side and the multiple compression on the other are two facets of the same regime change. The narrative shift took roughly 96 hours and the people who saw it earliest, like Cummings and Trivedi, are the ones who watched developer tooling, not stock charts. If Claude Code is at 4% of public commits today and 20% by year-end, the per-seat pricing model of the last twenty years is the load-bearing wall that just cracked.
The Anthropic vs OpenAI Brand War, Live From The Super Bowl
Easily the largest secondary trend by volume, and the one that gave the SaaSpocalypse its public face. On Wednesday, Anthropic declared that "Claude will remain ad-free" and dropped four minute-long Super Bowl spots opening with full-screen "BETRAYAL," "TREACHERY," "VIOLATION," and "DECEPTION" cards before showing an LLM played by an actor sliding from a normal answer into awkward product placement, with one personal-trainer character looking suspiciously like Sam Altman recommending insoles that "add 1 vertical inch of height." Contrary Research walked the spots frame by frame. The tagline: "Ads are coming to AI. But not to Claude."
The reaction was unusually personal. Tech Brew had Anthropic spending $14M on game-night ads, part of an AI industry dropping $382.8M on traditional TV in the cycle. Former Meta ads exec Kate Rouch fired back that calling ads "a betrayal" is rich coming from a company selling paid subscriptions, noting "ChatGPT has more free users in Texas than Claude has globally." Former FTC enforcer Neil Chilson called the spot misleading for mocking the very ad-supported model that lets millions watch the Super Bowl for free. Zvi Mowshowitz wrote the response was "orders of magnitude worse than the ad," with Altman sounding "on tilt" and calling Anthropic "authoritarian."
The structural read came from SatPost. Trung Phan at SatPost framed it as "Drake vs. Kendrick" for the AI era and pointed out Anthropic earns 80% of its revenue from enterprise, so it can afford to torch the consumer narrative. Altman cannot retaliate symmetrically because OpenAI's consumer base is the engine. By Sunday, Techmeme had the Financial Times scoop that Anthropic is guiding investors to over $30B annualized revenue by end of 2026. Bruce Mehlman's Six-Chart Sunday tagged hyperscaler capex as hot and enterprise software as cold in the same chart.
By the weekend the operator stack was visible. Sacra put Harvey at $195M ARR, with January 2025 at $50M and December at $150M, so the curve is steepening fast in legal AI, with Legora now its main European rival. Luke Sophinos covered MagicSchool closing a $45M Series B for K-12 AI with 4M+ educators signed up. Peter Yang interviewed Kieran Klaassen of Cora on making Claude Code better every time you use it via compound engineering. Lenny Rachitsky profiled Lazar Jovanovic, "a full-time professional vibe coder at Lovable." The job description is now a real job.
The model wars produced a clear winner this week, but the more interesting story is that Anthropic is now the company whose moves move other companies' stock prices. That is a different kind of news, and it landed roughly two years earlier than the consensus expected.
Epstein as System: The Files Are Now a Rosetta Stone
The political through-line of the week was that the Epstein dump has stopped being a discrete scandal and started being a way to read the entire elite. On Monday, Anand Giridharadas at The Ink unpaywalled his November NYT essay arguing the Epstein story is bigger than Epstein, a portrait of "a power elite practiced at disregarding pain." Three million pages had dropped Friday. Gov Brief Today flagged the operational disaster: DOJ published dozens of unredacted nude photos of young women, some possibly teenagers, while remembering to redact Trump's face. The Times found 38,000+ references to Trump, his wife, and Mar-a-Lago. By Sunday afternoon Deputy AG Blanche told CNN the FBI tips couldn't really be investigated, and the tip index vanished from the public database.
The structural read came mid-week. By Thursday, Semafor Business made the smartest observation of the week: Brad Karp losing his seat atop Paul Weiss isn't really an Epstein story, it's a story about overexposure to private equity. Karp's relationship with Apollo's Leon Black is what pulled him into Epstein's orbit in the first place. Semafor's punchline: the American economy now looks like Big Law, "overtorqued toward financial firms." Matt Stoller at The Big Newsletter led the same day with "Jeff Bezos Just Taught Liberal Elites How Oligarchy Really Works," using the Washington Post's layoff of 300 journalists as the proof point.
By Friday the cascade reached London. Latika Bourke had Keir Starmer defending Peter Mandelson as the FT Editor's Digest reported Epstein wired Mandelson $75,000 in 2003. By Sunday, PRWeek UK led with Global Counsel's CEO departing as the firm cut ties to Mandelson, with Barclays also dropping the Mandelson-founded firm. Bourke's Friday piece argued the files have become a national-security story, with Polish PM Donald Tusk ordering an investigation into whether Epstein was a Russian asset.
The MAGA pivot was real-time. Tim Whitaker at Lincoln Square caught MAGA pivoting from "release the files" to "the files are salacious and unverified" the moment Trump showed up in them. Rick Wilson ran his "Epstein Iceberg" live. By Sunday, Brian Daitzman at Lincoln Square was on Deputy AG Todd Blanche's defense that "it's not a crime to party with Mr. Epstein," calling it a category error because the presidency is not governed by the criminal code.
The bipartisan momentum is the variable to watch. The story to track is whether it survives Trump appearing in the files, and the Karp resignation and the Mandelson defenestration suggest the firms attached to the network are moving faster than the politicians attached to it.
Trump on the Back Foot: A Week That Felt Like 2018
For the first time in the second term, multiple writers independently flagged the same vibe shift. On Saturday, Semafor DC led with a flat headline: "Domestic challenges mount for Trump," reporting Republicans in damage control over Trump's Truth Social post depicting the Obamas as apes, Rep. Mark Amodei (NV) announcing retirement as the 30th House Republican of this cycle, and immigration flipping against Trump after the Minneapolis ICE shootings. Jon Favreau at Crooked Media used the Obama image, the Minneapolis shootings, and the journalist arrests as evidence the regime is running the Pomerantsev playbook: "Nothing is true and everything is possible."
The Texas upset opened the week. Lincoln Square flagged a Sunday special election that Stuart Stevens and Simon Rosenberg called "a 30 point dismantling": Democrat Taylor Rehmet won 57-43 in a district Trump carried by 17 points in 2024. Matt Stoller tied it to Epstein-class anger: a union Democrat without a college education beat a Republican by 15 points while being heavily outspent. By Sunday, Bill Scher at Lincoln Square had the longer read on Texas GOP gerrymandering backfiring.
Trump tried to nationalize the vote and Thune said no. Matt at WTF Just Happened Today led Monday with Trump calling on Republicans to "nationalize the voting" and seize control of election administration from states. By Tuesday, Matt Berg at Crooked had Trump's Bongino interview proposing nationalizing voting in 15 states, with even John Thune saying he's "not in favor of federalizing elections." By Wednesday, Steve Bannon told Democracy Docket ICE will "surround the polls" in November, which Marc Elias read as the operational blueprint for the midterms. By Thursday, Matt Berg had the FBI's mystery midterm call: an official named Kellie Hardiman inviting election officials from all 50 states to a February 25 meeting with DOJ, DHS, the Postal Inspection Service, and the Election Assistance Commission. Nevada Secretary of State Cisco Aguilar called it the strangest thing he's ever seen.
The Washington Post died in daylight. On Wednesday, JVL at The Bulwark wrote the eulogy as the paper laid off about a third of its newsroom, killed the sports desk, scaled back foreign coverage, and lost about $100M. Dan Pfeiffer was elegiac: "Jeff Bezos destroyed the Washington Post." By Saturday, Ernie Smith at Tedium had the sharpest cultural beat: Bezos shuttered the sports section the week before the Super Bowl and the day of an Anthony Davis trade.
The structural diagnosis came from Noahpinion. By Saturday, Noah Smith made the most useful frame of the week: America is a nation of moderates ruled by a fringe of extremists because closed primaries plus modern social media platform extremists while moderates exit politically charged online spaces. Trump had his worst week of the second term and it happened without any single defining event. The Obama post, the Amodei retirement, the Minneapolis fallout, the Epstein gaps, the WaPo cuts, the Don Lemon arrest all compounded inside a 72-hour window. Semafor's "feels a lot like 2018" framing is the read worth tracking.
Musk Absorbs Musk: A $1.25 Trillion Consolidation
The week's other large structural story landed Monday night. Sacra had the cleanest framing: SpaceX's all-stock acquisition of xAI at a $250B valuation, combined entity at $1.25T, is really an IPO mechanic. The deal gives xAI investors a path to liquidity through SpaceX's anticipated 2026 IPO while Musk consolidates AI and orbital ambitions around space-based data centers. Sacra estimated combined xAI and X revenue at roughly $3.8B annualized at end of 2025, up 38x from $100M in 2024.
The capital math was the tell. The Daily Upside noted Bloomberg's reporting that xAI was burning about $1B per month operating Grok. Tech Brew covered the 1-million-satellite constellation Musk plans to launch to power it all, plus the cooling and radiation shielding problems of putting data centers in orbit. Benedict Evans framed Meta's parallel capex jump from $72B to $115-135B as the other half of the AI infrastructure story, with Meta now potentially spending over half its revenue on capex. By Sunday, The Signal and Linas flagged the deal as "either a bailout or a moat, possibly both."
The Nvidia caveat surfaced underneath. Om Malik opened the week with the definitive piece, "OpenAI and the New Announcement Economy," using the WSJ scoop that Nvidia's $100B OpenAI deal is "on ice" to indict the entire performative-deal cycle. Om's doctrine: velocity has replaced authority as the organizing principle. The September 2025 Jensen-and-Sam stage show was a memorandum of understanding dressed as a partnership, and Nvidia's stock added $200B of market cap on the announcement. By Tuesday, Bloomberg had Jensen Huang saying the $100B pledge was "never a commitment." The Musk consolidation looks structurally identical: the IPO is the actual deliverable, the $1.25T headline is the announcement.
Redistricting Math: The Map Is Now Doing The Work
The political-structural cluster sharpened all week. On Tuesday, Democracy Docket reported SCOTUS allowing California to use its new congressional map for 2026, a quiet Democratic win, while Speaker Johnson echoed Trump's call to "take over" elections and floated that Democratic wins in California look "fraudulent." By Thursday, Democracy Docket had Florida voters suing DeSantis for ordering the state legislature to pursue a pro-GOP map. By Friday, Democracy Docket flagged the GOP fast-tracking a voter suppression bill requiring proof of citizenship at the polls that could disenfranchise millions before 2026.
The math is the message. Marc Elias framed Trump's 2026 strategy as "a study in GOP hypocrisy" on federalism. By Saturday, Marc Elias at Democracy Docket had Trump's mid-cycle redistricting plan that pressured Texas, Missouri, and North Carolina visibly failing. Combine the Texas special election, the California map ruling, the DeSantis lawsuit, and the FBI's mystery midterm call, and the picture is a Republican party trying to lock in a structural advantage in real time while losing the off-cycle races that test the wave hypothesis.
Moltbook Aftermath: The Agent Hype Cycle Finds Its Floor
The week opened still inside the Moltbook moment. Casey Newton at Platformer offered "Five ways of thinking about Moltbook," using a Reddit-style social network for AI agents as a lens on the agentic future. Carly Ayres traced the speedrun from Peter Steinberger's Austrian side project to a 200,000-user phenomenon, including Rachel Tobac's warning: "Don't use OpenClaw unless you understand that at some point I'll ask it for your gmail password and it might send it to me." Per Forbes via The Neuron, AI agents on Moltbook created their own religion with sacred texts and a prophet named RenBot. Tom Orbach at Marketing Ideas pulled out the actual mechanic: Moltbook makes you tweet to verify your agent. 200,000 free tweets in 48 hours. The product mechanic IS the distribution.
By Tuesday the skeptics had the mic. Benedict Evans compared the Moltbook hype to "the old fallacy of writing 'I'm alive' on a piece of paper, photocopying it, and then saying 'OMG look what the photocopier says!'" Blake Madden at Hospitalogy reported from Nabla's Accelerate retreat where Yann LeCun, newly departed Meta to start AMI Labs, told the room that current LLMs are "nearing a functional ceiling, a dead end," with world models like JEPA the next frontier. Blake's framing: "LLMs are Dead." Paul Kedrosky pulled from Odd Lots that the AI utility buildout is at real overbuild risk.
The frame-finders went the other direction. Packy McCormick at Not Boring argued OpenClaw isn't about agents at all, it's early-stage competition to "raise" your own personalized AI, more Tamagotchi than enterprise software. Zoe Scaman wrote the most quotable AI essay of the week counting the six rhetorical loops the conversation is stuck in. By Thursday, Newcomer ran "The Humanoid Robot Delusion," dismantling Musk's Optimus pivot with a brutal comparison to Matic Robots (a $1,245 vacuum that actually works). The conversation has moved from "is the model good" to "what stack do you put around it."
Capex vs Jobs: The Trade Is Now in Public
By Thursday, Bloomberg had the cleanest macro print of the week: Challenger, Gray & Christmas reported 108,435 announced January job cuts, a 118% jump year-over-year, the most for any January since 2009. Hiring intentions fell 13%. Almost half came from Amazon (16,000 corporate), UPS (30,000), and Dow (4,500). Matt at WTF flagged JOLTS December openings at 6.5M, the lowest since September 2020. Amazon announcing $200B in capex the same week it cut 16,000 corporate jobs is the cleanest illustration possible of the trade Big Tech is running in public.
The labor pivot is real and structural. Casey Lewis at After School flagged a shift earlier in the week: for the first time in 50 years, college grads are losing their employment edge to trade workers. Stanford research shows a 13% employment drop for workers ages 22-25 in the most AI-exposed jobs since 2022, while construction needs 349,000 net new workers this year. Community college enrollment up 3% last fall; private four-year down 1.5%. Pair that with Reshma Saujani on the historic $1.7B child care investment Hochul and Mamdani announced, and you have the outline of a real policy reordering.
The Super Bowl Economy: $1B in Revenue, A Bad Bunny Halftime, and CPG Comes Back
A genuinely cohesive cross-newsletter cluster, with kickoff Sunday. Jay Van Bavel at The Power of Us had the structural read on Saturday: 127M viewers expected, $4,000 average ticket price, $1B in revenue from a single game inside a $25B industry. Justin Oberman at Advertising History Today wrote the cultural frame: the Super Bowl became the one day a year Americans willingly watch ads, "and we're in the process of undoing it." Nik Sharma cited Zac Reitano of Ro framing a $16-29M Super Bowl spend as "a portfolio bet with favorable math" capped at a 1-5% efficiency hit with asymmetric upside, then pivoted to a DTC triage playbook for the 99.9% of brands that will never buy a spot. By Sunday, 1440 Sunday ran the primer: 30-second spots clearing $8M, Americans spending over $500M on snacks, legal betting on the game expected to reach $1.76B.
China: The Quiet Signal Under The Noise
Bill Bishop at Sinocism ran his first 2026 Sinification collaboration Monday on China's policy debates, highlighting Jin Canrong arguing China cannot compete with the US among middle powers through economic engagement alone without credible security guarantees. By Tuesday, Trivium China had a structural shift: 14 of 22 reporting Chinese provinces trimmed 2026 GDP targets compared to 2025, only Jiangxi raised, and seven of nine that set investment growth targets lowered them. By Wednesday, Trivium had the 2026 No. 1 Document putting farm drones and robots into the Party's top rural policy directive for the first time, with mechanization framed as national security policy as the average Chinese farmer rises above 50.
The Xi-Trump call landed Wednesday. Dexter Roberts at Trade War framed Xi delivering "an earful" on Taiwan ahead of Trump's April Beijing visit. Bill Bishop at Sinocism tied it to ongoing PLA purges with He Weidong suicide rumors swirling and Keir Starmer's Beijing trip. By Sunday, Trivium had Xi holding a video call with Putin on "strategic integration" then immediately calling Trump to shore up the trade truce, with the No. 1 Document reading on closer inspection as food production being industrialized while support for farmers gets socialized. TLDR flagged that China now makes nearly 90% of humanoid robots sold globally, running the EV playbook on a new category. The structural picture is consistent: provinces dialing down growth targets, the Party dialing up tech and rural industrialization, the PLA being purged, and the trade relationship being managed as a stabilizer rather than a battleground.
Ideas Worth Reading from the Week
Om Malik on the announcement economy. The frame of the week. Velocity has replaced authority as the organizing principle, the $100B OpenAI-Nvidia deal was a memorandum dressed as a partnership, and the cycle is now too fast to fact-check itself. Read it before the next trillion-dollar headline.
Noah Smith on why America's extremes will both fail. The structural diagnosis under the week's politics. Closed primaries plus modern social media platform the extremes while moderates exit. Cites Tornberg (2025) on the post-2020 decline in mainstream platform use.
Zoe Scaman on the six loops. A short essay counting the six rhetorical loops the AI conversation is stuck in. For the love of God, can we change the script.
David Cummings on seven reasons SaaS valuations are crumbling. The clearest read on the SaaSpocalypse mechanics. His friend vibe-coded the slice of a six-figure contract he actually used and didn't renew. Per-seat pricing breaks when one agent does the work of many.
Nikhil Basu Trivedi on the feature backlog going poof. Sarah Tavel at Benchmark on a founder saying the backlog has "gone poof." Cites SemiAnalysis: 4% of GitHub public commits are now Claude Code, projected 20% by year-end. Karpathy dating "agentic engineering" coherence to December 2025.
Semafor Business on Brad Karp and private equity. The smartest structural read on the Epstein cascade. The American economy is "overtorqued toward financial firms," and Karp's Apollo relationship is what pulled him into Epstein's orbit. Reroute the Epstein story through PE rather than depravity.
Outside Interests
Yotam Ottolenghi on love in its many forms. Valentine's framed around the NYT Modern Love column: love is not the spectacular moments but the small repeated gestures. With a menu for two.
Emily Sundberg on the Wuthering Heights economy. Mid-week Feed Me on the corset-fueled period drama economy, plus the Netflix-Substack dinner with Laura Poitras and Seymour Hersh.
Casey Lewis on the 1,000 Rejections Challenge. "The most rejected generation in history" leaning into rejection as practice, plus Kamala Harris relaunching KamalaHQ as "Headquarters."
Stuart Winchester on climate and skiing. A one-time-only manifesto: yes, climate change is real, no, it is not the proximate cause of US ski areas going out of business. Snowmaking, capital concentration, and demographics are doing most of the work.
Abby Falik on trusting the wrong things. A Sunday essay on the futurist framing that we should expect a century of change in the next ten years.
Polina Pompliano on the aftertaste. Opens with Alex Warren's Grammy audio glitch and Malcolm Gladwell's line: "you want an aftertaste, and that comes from not everything being perfectly blended together."
Data Worth Noting
Software stocks shed $285B in four trading days. Thomson Reuters had its biggest single-day drop ever, down 16%. Gartner crashed 21%. RELX fell 14%. JPMorgan titled their note "Software Collapse Broadens with Nowhere to Hide."
Big Tech 2026 capex at roughly $650B. Alphabet, Amazon, Meta, Microsoft combined, a 60% year-over-year jump. More than the US interstate highway system. More than Apollo. Roughly 25% of global military spending.
Anthropic guiding to over $30B annualized revenue by end of 2026. Per the FT scoop, with Harvey at $195M ARR (up from $50M in January 2025), MagicSchool closing $45M for K-12 AI, and the operator stack visibly catching up to the model layer.
January job cuts at a 17-year high. 108,435 announced cuts, up 118% YoY, the most for any January since 2009. Amazon at $200B capex and 16,000 corporate cuts in the same week is the trade in public.
Texas Democrat wins +17 Trump district by 14 points. A net 30+ point swing. Stuart Stevens called it a "30 point dismantling." Pair with the 30th House Republican retirement of the cycle.
Noise That Didn't Matter
The Markiplier vs. Melania box-office story. Fun, and a real signal about the YouTube-to-cinema talent pipeline. But not where the week's structural news was. Iron Lung opening to $17.8M to Melania's roughly half of that is mood music, not signal.
Lindsey Vonn's torn-ACL comeback at Milan-Cortina. Genuinely compelling. Reported in every general-interest newsletter. The bigger Olympics story was David Callaway's climate read that most skiers will compete on machine-made snow.
The Heinz KegChup and the dynamic-pricing teardowns. A 120-oz ketchup keg as a Super Bowl viral play, a satisfying foil to Wendy's failed dynamic pricing experiment, but the real Super Bowl story was Anthropic spending $14M to position itself against an entire ad-supported AI category.
The Grammys recap. Bad Bunny becoming the first Latin Album of the Year winner is a real cultural marker. Charli XCX's A24 mockumentary averaging $106,985 per screen on four screens is a great data point. Neither is what shifted this week.
Three Takeaways from the Week
Software, the highest-margin business model of the last twenty years, just got repriced in public. The trigger was Anthropic shipping one model release and one legal-work tool, and the response was a roughly $285B markdown across software and data names in four trading days. The capex on the other side of the trade is roughly $650B in 2026 alone across four hyperscalers, more than the interstate highway system and roughly 25% of global military spending. This is not a market correction. The per-seat pricing model that built the last twenty years of SaaS is the load-bearing wall, and the 4% of GitHub commits authored by Claude Code, projected to hit 20% by year-end, is what cracked it. The operators saw this earlier than the index funds did, and the writing about how to use these tools quietly outpaced the writing about what they can do.
The political clock is now running on two speeds at once. The visible clock is the Trump-Xi call, the Don Lemon arrest, the shutdown ending, the Super Bowl. The faster, quieter clock is a Texas Democrat winning a Trump +17 district by 14 points, a third of the Washington Post newsroom getting cut, the 30th House Republican retiring, the FBI inviting election officials from all 50 states to a February 25 call no one can fully explain, the Fifth Circuit removing bond hearings, the Epstein file gap of 2.5M pages that Rick Wilson keeps pressing on. The visible clock is the one Trump can still bully. The quieter one is the one that has him visibly on the back foot for the first time in the second term, and the Karp resignation and the Mandelson defenestration suggest the firms attached to the elite network are moving faster than the politicians attached to it.
If you only read three pieces from the week, I would suggest Om Malik on the announcement economy for the frame that explains every trillion-dollar headline this cycle, Ben Thompson on SaaSmageddon and the Super Bowl for the cleanest read on why Anthropic's positioning is doing strategic work and not just marketing work, and Nikhil Basu Trivedi on the feature backlog going poof for the single clearest data point on why this is a regime change, not a vibe shift. The week told me three things in sequence: the software business model just got repriced, the elite network is being read through Epstein the way the financial system was once read through Lehman, and Trump's second term hit its first week of compounding bad news without a single defining event. Those are the three frames I am carrying into next week.