(Preview) A Summer Break Mailbag: Memory Mania, Vibe Coding, Mafia PR, Caffeine Intake, Garages, and How to Fix Soccer
Summary
Thompson reads Apple’s mid-cycle price increases as a bearish signal because it missed the memory shock twice: first in procurement, then in March pricing. Jensen Huang was visibly securing supply while “we didn’t see Tim Cook drinking soju nine months ago”; Apple then launched new Macs without increases, only to raise prices by hundreds of dollars out of band. That breach of Apple’s stable-pricing promise suggests management “just doesn’t know what’s going on” with AI.
Apple may still deliver a good Siri in 2026, but Thompson argues it will embody what was possible in 2024. AI was already good enough then to support “tons of products,” so Apple can fulfill its old promise while remaining “nowhere close” to what AI is capable of in 2026 and beyond. Sharp’s framing is also harsh: Apple has become “a bunch of normies out there in Cupertino.”
The memory shortage could turn China into a dominant player by letting it climb the learning curve profitably. Thompson says Chinese producers cannot yet make current HBM, though they can make HBM1 or HBM2 while the frontier is HBM4. DRAM is fungible, and as the big three suppliers focus on HBM, scarcity lets China sell older DRAM at profitable prices, make a lot of it, and improve.
Export controls and attempts to capture scarce supply may accelerate the competitor they are meant to constrain. If Nvidia is not allowed to take Chinese memory, Huawei and others will take it. Thompson’s categorical conclusion is that “the die is cast” and China will become dominant in memory, while Sharp sees possible security-driven US investment over the next 5 or 10 years despite weak economics.
Vibe coding collapses the distance between an idiosyncratic need and purpose-built software. Thompson wanted a lookup system for household objects—not inventory or task-management software—designed around one assistant doing the cataloging and one owner who does not want to do the administrative work. “I don’t need permission to do it. I don’t need to hire a developer.”
The resulting app combines a physical-location hierarchy, AI vision, and QR codes into a narrowly optimized workflow. An assistant chooses where everything belongs; a photo gets roughly 95% of objects right; scanning a shelf’s QR code assigns a later arrival to its location. The specificity is the point: conventional software accumulates features to serve millions, while AI enables software tuned to one user.
The lighter mailbag answers expose contrasting personal operating systems rather than investment calls. Sharp consumes a Venti iced coffee, a Grande Americano, and on three of five workdays a 20-ounce yerba-mate drink, despite hitting “a point of diminishing returns.” Thompson usually drinks one morning pot and avoids caffeine past noon. His broader productivity maxim is sharper: “You don’t fix your weaknesses; you make more money so you can pay for someone else to fix them.”
Deep dive
1. Apple’s price reset says management missed the memory regime
Thompson calls Apple’s price increases “one of the more bearish indicators for Apple” he has seen in a long time. His first clue is procurement: he points to Jensen Huang’s public soju diplomacy with Samsung as evidence that Huang was securing supply, whereas “we didn’t see Tim Cook drinking soju nine months ago.”
Apple’s pricing promise is predictability, not cheapness: “My price is my price. It’s only going in one direction.” Unlike Dell’s constantly changing offers, Apple normally avoids altering a product’s price while it remains in market, removing any incentive to await a sale.
March offered a clean opportunity to price new Macs for higher memory costs. Apple declined, apparently hoping the cycle might “blow over,” then raised prices by hundreds of dollars out of band after warning that memory would affect the second quarter “a little bit” and the third more substantially.
Thompson still owes John Gruber a steak after betting on a March increase: “It’s not enough to be right, timing matters.” Yet the delayed move strengthens his substantive concern that Apple “just doesn’t know what’s going on” with AI.
2. Apple’s delayed AI can succeed while remaining years behind
Thompson expects Apple to ship in 2026 what it promised in 2024, and expects it to be “great” because the underlying AI was already powerful enough to support many useful products.
His distinction is between the quality of the forthcoming product and the pace of the technology: Siri may be good, but it is “2024 technology,” nowhere close to what AI can do in 2026 or going forward.
Sharp’s own framing is harsher: Apple has become “a bunch of normies out there in Cupertino.” He agrees the delayed Siri is “pretty good,” but does not offer a substantive counterargument to Thompson’s claim that Apple is behind.
3. Scarcity gives Chinese memory producers a profitable learning curve
Thompson’s framing starts with fungibility: the immediate need is DRAM, even if Chinese manufacturers cannot yet produce current HBM. He says they can make HBM1 or HBM2 while “we’re on HBM4 now,” leaving a real technology gap without eliminating demand for their output.
As the big three suppliers focus on HBM, China can fill the DRAM shortage with older or less advanced technology, though its DRAM may have bigger problems. Previously it had to “eat” the cost of low-end production to gain experience; now scarcity permits it to move down the learning curve profitably.
Cutting China off through export controls or commercial supply capture creates “the conditions for inevitable competition.” If Nvidia is not allowed to take Chinese memory, Huawei and others will take it; Thompson concludes, without hedging, that China “is going to be a dominant player in memory.”
Sharp notes the longer-term policy tension: American memory investment may not make economic sense, yet could make security sense over the next 5 or 10 years as Chinese capability rises.
4. Personal systems range from caffeine overload to outsourced discipline
Sharp describes himself as “deeply addicted”: a Venti Starbucks iced coffee every morning, a Grande Americano in early afternoon, and a 20-ounce yerba-mate drink on roughly three of five workdays. If he drinks too much, his brain “just shuts down at a certain point.”
Late-night recordings once found Sharp trying to solve that state with Diet Coke around 9:45 or 10:00 p.m. He says a daily Red Bull is probably a bad idea—“I’m not a doctor”—while acknowledging that he tries to moderate his own caffeine use.
Thompson makes one morning pot from freshly ground beans, occasionally gets a Diet Coke when writing at night, and generally avoids caffeine after noon. He says he can go a week without coffee and values Starbucks because “it’s the same every time,” not because it is exceptional.
Thompson’s broader weakness is execution: he buys networking gear or other components based on ideas for solving problems, then waits for a forcing function. “You don’t fix your weaknesses; you make more money so you can pay for someone else to fix them.”
5. Vibe coding makes bespoke software economically plausible
Thompson’s garage problem is a coordination loop: multiple shipments were hurried onto shelves, but he resists putting remaining objects away until every category has a durable home. An assistant could organize everything, yet Thompson would then have no idea where anything was.
Existing home-inventory products feel like “enterprise software for your house.” Thompson connects their clutter to software economics: Word and AWS need vast feature sets because products sold to a huge audience must cover many different use cases.
His app rejects that model. It is solely a “look up something and find out where it is system,” optimized for an assistant willing to perform the cataloging and an owner who “doesn’t wanna do any work at all.”
The house becomes property, room, structure, then storage spot. The assistant owns placement; AI gets roughly 95% of photographed objects right; and later arrivals are photographed, shelved, and attached to a location by scanning its QR code.
A bottle of specialized windshield-washer concentrate serves as the test object for checking whether an item is already in the system. The project restores Thompson’s 1990s feeling of being “a kid again” with computers, though his closing reality check is blunt: building it “takes longer than you think.”