February 2026 Random Ramblings

Summary

• Walker’s core tension this month: his instinct in a panic is to be “a gunslinger running into a firefight… spraying money everywhere,” but the AI-driven SaaS selloff is a panic where the earnings of companies with no tangible assets might evaporate. Some banks in 2023 had $1 billion of tangible equity while selling for $600–700 million; busted biotechs in early 2025 traded below cash; SaaS “earnings can evaporate quickly and when they do there’s kind of nothing left over for the equity holders.”
• The AI fear is no longer contained to software — on the day of recording (Feb. 12, 2026), office stocks were down 5–10%, RXO fell roughly 20% and ODFL about 5% on what Walker believes was an AI-for-LTL paper, and insurance and brokers were hit earlier in the week. The tell of true dislocation: whole-sector days where “every name is down 10%” across totally disparate businesses — enterprise Salesforce and consumer Duolingo down together.
• On claims that specialty insurers sold off because Anthropic rolled out a basic insurance product, Walker rejects the smug “how stupid is the market?” reaction: if an AI tool can price standardized life insurance better than anyone else today, “in three years they’re going to be able to do specialized insurance.” His anchor is the Will Smith video example — from melting wax-figure horror with eight fingers two years ago to “indistinguishable from A-level Hollywood movies” today — because “humans are really really bad at dealing with exponential improvement.”
• The bull-side caveat he keeps top of mind: college students in 2021 or 2022 emailed him Twitter-short pitches with clones “coded inside of a day” — and they were “exactly wrong,” because the moat is network effects, not the website. Claude vibe-coding something that looks like Salesforce is not a Salesforce competitor; no enterprise switches wholesale to a visual clone.
• He cautions generalists in this panic: unlike banks, where reading footnotes and buying below book worked, SaaS pits you against sector specialists running CIO calls and surveys — and he’s hearing real-time behavior shift, people who said “I’d never use AI for that” three months ago now “changing some habits around the margin.” A trailing 10-times-free-cash-flow multiple, even after taking out stock comp, is therefore still scary rather than conclusive. His preferred leading indicator: compare current SaaS adoption at 500-employee companies with whether smaller companies are delaying the adoption larger companies showed two years ago.
• The hard-asset flight-to-safety bid (coal, steel, cement) is “probably right to some extent,” but “be careful what you wish for”: if AI “cracks the code on efficient batteries,” solar-plus-storage gets “really bearish” for coal and natural gas. Cement is his concrete example — local, heavy, hard to ship, with an AI data-center construction tailwind.
• His most original frame: AI may do to SaaS what YouTube and the iPhone did to the cable bundle — drive distribution and scale costs down, replacing hundreds of engineers with “one unique software engineer—or three unique software engineers—and two great salespeople” custom-coding the last 5% for a couple of clients. The TikTok-era outcome: hundreds of well-paid stars nobody’s nationally famous like “a Jennifer Aniston or a Wolf Blitzer.”
• The closing meditation: investing demands arrogance and humility balanced “on a knife edge” — and he openly asks when his own 15–18 months of saying he is “a little confused by the markets” and somewhat bearish fails the update-your-priors test, lest he become the CNBC guy bearish since 2011. The 20%-position blow-up pattern: great for five years, then one year all the concentrated bets are wrong, down 80%, “and it’s all over.”

Deep dive

1. The AI panic has escaped software and is hitting everything with a story

• Recording Feb. 12, 2026, Walker catalogs the contagion: office stocks down 5–10% (“big moves for office buildings”) on fears AI layoffs empty buildings forever; RXO down roughly 20% and ODFL about 5% because, he believes, “some fly-by-night penny stock released a paper on using AI for LTL”; insurance and brokers hit earlier in the week. “It’s not just SaaS, it’s everything else blowing up.”
• His panic diagnostic: days where an entire sector drops 10% across names with nothing in common — enterprise CRM Salesforce alongside Duolingo, consumer “quote unquote” language learning (“I did Duolingo for two years and I think I learned 10 words of Polish”). “That’s generally when there’s a panic and that’s generally when there’s dislocation.”

2. Why this panic is harder to buy than banks 2023 or biotech 2025

• Walker’s regret runs one way: “I wish I had done nothing my career but run into panic” — banks after the SVB and First Republic crash and early-2025 busted biotechs both did incredibly well. But those had hard assets: some banks had “$1 billion of tangible equity” while selling for $600–700 million; if their bonds and loans were properly marked and the discount remained, the bank-run risk was gone. Biotechs were “trading so far below cash value that you no longer had to worry” — “they weren’t even science projects anymore.”
• The SaaS asymmetry that scares him: all the value sits in engineers, contracts, and cash flows — no tangible floor. “The earnings there can evaporate quickly and when they do there’s kind of nothing left over for the equity holders.”

3. Exponential improvement vs. network effects — the two poles of the debate

• The exponential case, as told: the Will Smith AI video that two years ago “looked like a horror movie where the character was made of wax or cheese… and he had like eight fingers” is now “indistinguishable from A-level Hollywood movies.” So “maybe it can’t right now” replace Salesforce — but at exponential improvement rates, parity comes “really effing fast.”
• On claims that specialty insurers sold off because Anthropic rolled out a basic insurance product, he rejects the smug take that the market is stupid. His argument is conditional: if an AI tool can price standardized life insurance better than anyone else today, “in three years they’re going to be able to do specialized insurance.”
• The counterweight he keeps top of mind: 2021–22 college students emailing Twitter-short pitches with clones built in a day were “exactly wrong” — “the Twitter website is not what is unique about Twitter. It is the network effects.” A Claude-coded Salesforce lookalike is not a working competitor; “no company is going to switch over whole hog.”

4. Generalists beware; hard assets aren’t automatically safe

• In banks, a generalist reading footnotes and buying below book made money. In SaaS, “you’re competing against sector specialists” doing CIO calls, expert panels, and surveys — and Walker’s own calls show behavior shifting in real time: someone who said “I’d never use AI for that” three months ago is now “changing some habits around the margin.” That makes relying on a trailing 10-times-free-cash-flow multiple, even after taking out stock comp, really scary rather than conclusive.
• The indicator he’d watch: survey 50 CIOs at roughly 500-employee companies on current SaaS use, then compare whether smaller companies are delaying the adoption larger companies showed two years ago. If that delay appears at the small-company end, it will start bleeding up to larger companies.
• On the rotation into hard assets — coal, steel, and cement as a roughly “AI-proof” flight-to-safety trade: “probably right to some extent,” and cement (local, heavy, data-center tailwind, “we’re going to be using cement 100 years from now”) is his concrete example. But “be careful what you wish for”: if AI “cracks the code on efficient batteries,” solar-plus-wind-plus-storage “gets really bearish” for coal and natural gas.

5. The YouTube-ification of SaaS — and the knife edge of arrogance and humility

• His media analogy: the 2010s cable bundle was “probably the greatest business that’s ever been invented” until YouTube and the iPhone drove distribution costs way down and damaged cable channels. AI might do the same to SaaS scale — instead of hundreds of engineers and salespeople, “one unique software engineer—or three unique software engineers—and two great salespeople” who vibe-code a CRM, custom-build “the last 5%,” and work with clients full-time. Fragmentation, TikTok-style: hundreds of well-paid stars, none famous like “a Jennifer Aniston or a Wolf Blitzer.”
• His honest self-interrogation on priors: friends have called AI a bubble for 18 months — “when would they say they’re wrong? When would they update?” — and he turns it on himself, having been “a little confused by the markets and a little bit more on the bearish side” for 15–18 months while AI stocks, power stocks, and the Magnificent Seven ripped. The failure mode: the CNBC guy “bearish since 2011.”
• The closing frame: investing requires arrogance (“I’m going to beat the most competitive of games”) balanced against humility on “a knife edge.” The blow-up pattern he’s watched: investors taking 20% positions do great for five years, “then one year all of them are wrong and they go down 80% and it’s all over.” His own evolution: he once idealized “Walter Schloss investing in a windowless room and just reading 10-Ks,” now finds that podcasting, talking to smart people, and discussing markets help him “think deeper, think harder.”