Sirsystems

Hypothesis: Most mediocre mid-career people do not talk *enough* about their problems, they are often just giving gyaan to younger folks (yes including me 😂)

Tom Brown, a retired engineer, dedicated 25 years to preserving approximately 1,200 apple varieties from extinction.

This is very true. I have seen average developers grow very fast once they got the right mentor, and very smart developers stay stuck for years without one. A software engineer friend of mine was trying to learn distributed systems on his own. He would watch long videos, read blogs, open Kafka docs, close them, then jump to database replication, then CAP theorem, then back to coding APIs. He was putting in effort, real effort, but the loop was broken. Too much input, very little correction, and almost no signal on what actually mattered. Then he joined a team with a very good senior. That senior did not just dump resources on him. He would say things like: Do not learn all of Kafka today. First understand why your service needs a queue. Then understand at least once delivery. Then build a small consumer. Then break it. Then tell me what went wrong. That changed everything. In 2 weeks, my friend understood more than he did in 6 months alone, cause now every rep had feedback attached to it. He wrote code, got corrected early, learned what was important, and stopped wasting energy on random complexity. This is why a good mentor matters so much in software engineering. They compress the feedback loop. They help you avoid fake learning. They tell you when you are confused. They show you what good looks like. And sometimes one sentence from the right mentor can save you 6 months of going in circles. Talent matters, yes. But for most devs, progress really starts when someone experienced helps turn vague struggle into tight feedback and real reps. It can be a good manager or a good team lead or even the VP if he is hands on.

Engineers that don't know their ass from a vector space aren't going to invent the Levi-Civita connection or discover Riemann curvature. They aren't David Hilbert or Einstein, the philosophy you need for this is from a culture that wouldn't exist.

Today, SanDisk replaced Atlassian in the Nasdaq-100. A NAND manufacturer took the seat of a Jira vendor. That is the regime change in one headline. Read it twice. For 15 years, the Nasdaq-100 was a monument to asset-light software. High gross margins, negative working capital, zero cleanroom capex, Rule of 40 as scripture. SanDisk is the anti-Atlassian. Fabs, fluorine chemistry, EUV-adjacent lithography, 232-layer stacks, wafer yields measured in basis points. Actual physics. And here is the part nobody on fintwit is pricing in. The US graduated 24,547 electrical engineering bachelor’s in 1986-87. In 2020-21, we graduated 16,914. Four decades later. Still below the Reagan-era peak. Not flat. Down. Computer science over the same window went from roughly 39k to 109k. Call it 2.8x. The two lines on the chart do not just diverge, they mock each other. I say this as an electrical engineer. UCSD, Fainman’s ultrafast nanoscale optics lab, silicon photonics, micro-ring resonators. I know exactly what it takes to train someone who can actually move a process node, close timing on a mixed-signal die, or debug a yield problem at 3am. It is not a weekend cohort. It is a decade minimum, and most of that decade happens inside a fab or a tape-out cycle, not a classroom. An entire generation of smart kids was correctly told to chase software. That is where the returns were. TAM expansion, zero marginal cost, stock comp that prints. Nobody was writing Substacks about NAND process engineers in 2015. Nobody was telling their kid to go learn III-V epitaxy instead of React. Now the regime has flipped and the pipeline is a ghost town. You cannot bootcamp a device physicist. You cannot GPT your way into an analog layout. You cannot vibe-code a 232-layer charge trap stack. The training loop is a PhD plus a decade of tribal knowledge locked inside TSMC, Micron, Hynix, Samsung, Applied Materials, and maybe a dozen labs that actually still teach this stuff. When SanDisk wants to double enterprise SSD output, the binding constraint is not capital. The market will fund it at any multiple you can type into a DCF right now. The constraint is humans who know how to make the stack yield. And those humans are already employed, already vested, and already being counter-offered. This is the Memory Wars thesis with a labor-market overlay. The software guys spent 15 years telling us hardware was a commodity. Now the commodity has pricing power and the “real engineers” are the ones walking into comp negotiations with leverage for the first time in a generation. Pricing power accrues to whoever already has the talent locked up. That is the incumbents, and anyone with a serious university pipeline. Everyone else is about to learn that CHIPS Act money does not manufacture device physicists. It just bids up the ones who already exist. SanDisk entering the NDX is not the story. SanDisk entering the NDX while the US graduates fewer EEs than it did under Reagan, that is the story. Source: NCES Digest of Education Statistics, Tables 325.35 and 325.47. Bachelor’s degrees only.

has anyone ever seen someone go from “low agency,” like Really low agency, to being able to substantially advance their own life/make better decisions? and I don’t mean “they couldn’t do anything because of a severe depressive episode”

You floss. You bleed. That bacterium doesn’t stay in your mouth. It crosses into your bloodstream. Then it crosses into your brain. It was found in 96% of Alzheimer’s tissue. P. gingivalis. You’ve watched someone lose their memory. Slowly. Everyone said it was age. You told yourself it wouldn’t be you. It lives in your gum pockets. Feeds on blood. Produces enzymes called gingipains that destroy tissue. Not just in your mouth. When your gums bleed, that’s a broken barrier. P. gingivalis is in your blood. I spent six months reading gut research before I found this study. Higher gingipain load correlated with worse Alzheimer’s pathology. Post-mortem. Published in Science Advances. PMID: 30675515. Human brains. Now you know something your dentist doesn’t. If someone you love is losing their memory — screenshot this and send it to them. Your dentist has seen your bleeding gums hundreds of times. Never once mentioned your brain. P. gingivalis lives BELOW the gumline. In pockets standard brushing doesn’t reach. Bleeding when you floss isn’t normal. The barrier is already broken. One organism. Two barriers. Both broken. P. gingivalis can’t survive without iron from blood. Stop the bleeding and you STARVE it. Bleeding gums aren’t a cosmetic problem. They’re a supply line. Keep flossing and ignoring the blood. Every time, P. gingivalis enters your bloodstream. That’s been happening for years.

Without Einstein's general relativity from 1915, your GPS would drift about 10 km per day, and you'd have no idea why.