The blatant dishonesty of our elected representatives is galling. Let’s replace this with honest, real-world numbers. Elon’s net worth is ~$1.14T as of 5 pm ET today. A 5% tax on that would raise* ~$57B. The best estimated cost for universal free college & trade school is ~$75B/year, which was provided by PWBM & Urban Institute- imperfectly priced in the demand side response & ignores the supply side reality. Nothing in any of those has modeled the interaction between surging demand plus institutions with no price discipline and a federal payer. A more realistic scenario is a $150B first year price tag, slowly falling in real terms over the next 10 years as new supply (seats) comes online and people realize free degrees don’t have the same value but still have a significant time cost. So: no, a 5% tax does not pay for this. It doesn’t come close.

Major Shopify outage…again. Unreal.

I watched Bernie's video

The 5th Amendment exists. The real effect of this: it would destroy ~20% to ~45% of the value of every investment in your brokerage account, 401(k) or IRA. It would make the US government (in whatever form/fashion this bill decides) the controlling shareholders in the entire technology sector. And - most critically - it would relegate every entrepreneur who spends his/her life building something from an owner to a temporary custodian. The fact that this was even entertained sickens me as someone who has come from relatively little. It's the most Anti-American idea I've ever heard. @BernieSanders is not fit to be a sitting United States Senator. We're well past the point where his antics are those of a kooky old wanna-be socialist.

If my goal was to destroy the US' global domination in finance, innovation & technology, I could not run a better playbook than @BernieSanders. This is the most insane thing I've read in months. Nationalizing 50% of some of the largest companies in the world would immediately decimate millions of investors beneficiaries (including teachers, nurses, government workers, retirees). These companies are owned by individuals. Bernie Sanders is proposing that half your NVIDIA shares or half your Alphabet shares or half your Microsoft shares go to the US government. That's the end state. And if you say, "Well I don't own those in my retirement, I only own $VOO [or whatever ETF]" - guess what accounts for 40%-45% of that ETF's total holdings? The fact that a sitting US Senator wrote this should shake every American to his/her core. There is nothing more anti-American than stealing from teachers, doctors, social workers, retirees & ordinary men and women so Sanders can give out [whatever] to his preferred constituencies.

Collin is exactly right. It is actually cheaper for us to build maintain several tools in-house than it is to continue paying thousands per month in costs. Yes, that includes maintenance hosting updates. We are finalizing development of one tool for under $10k all-in that will *upgrade* a $1,599/mo subscription. Hard to beat that from an investment standpoint.

Because of Griggs v. Duke Power Company. SCOTUS ruled unanimously that under Title VII, any hiring screen producing “disparate impact” on a protected class must be demonstrably job-related and a business necessity…and the burden of proof sits on the employer, not the plaintiff. Companies responded (predictably) by moving to less rigid hiring assessments & expanding the use of unstructured interviews, as those are legally safer despite being objectively less predictive. In essence, companies traded more accurate, more consistent prediction of applicant success for fewer lawsuits.

This is a categorically insane statement. At $400k/yr as a *salary*, you can live a wonderful life with relatively little stress. Save $8k/mo. After taxes, you’re living on $15k-20k/mo. By 45-50, you have a wonderful nest egg of $4M-$8M. You’ve lived a life better than 99% of the world on your way to it. And, since Meta does love to give out options & incentive units, they likely have another $1M-$2M in stock, which is just the cherry on top of the rich life sundae. For 99% of the population, that is a better life than what they’ll ever experience. And, the best part of the deal is that this high W2 salary provides them optionality for anything else. If they want to build a side project? No problem. If they want to angel invest or advise startups? They have the resources & a resume tailor-made for it. If they want to travel, enjoy life, raise a family? Can do. The notion that every job is a jail sentence simply must go. The reality is that 98% of the population is better off working for someone else.

The Microsoft and Uber stories are being misread by most people because the reading skips the part where neither company actually scaled back. MSFT is migrating 100k engineers from Claude Code to Copilot (which they own capture every prompt bit of user feedback to improve their own product). That's a different thing vs. cost capitulation - it's MSFT electing to build internally, where they can use frontier capacity at cost vs. at cost-plus-markup AND make that investment do double work (improve Copilot save $$$). Uber's CTO said the company was "back to the drawing board." That doesn't entail that they're no longer using the tools - it means they're rebuilding their forecasts/models/FinOps around consumption they underestimated by an order of magnitude. Engineers spending $2k/mo on tokens aren't doing it for fun; they're doing it because the productivity differential justifies it. Both examples point to the same conclusion (and it isn't the one being reported). It's "this got valuable enough that usage exceeded projections in a fraction of the time" aka Jevons paradox showing up in real time - not the end of the subsidy era. The market structure emerging from this is bifurcation, which is what is missed in the entire conversation. On one end: Frontier inference (reasoning, long context, agentic workflows) is consolidating rents at the top. Demand at that tier is supply-constrained, not price-sensitive. On the other: Commodity inference (the GPT-4-class workloads that defined the category 18 months ago) is collapsing toward marginal cost. Equivalent-capability pricing is now ~1/600th of what it was 6 years ago. Those 2 curves moving in opposite directions are the entire pricing landscape. They have almost nothing to do with each other. Treating this as a single number is an analytical error that's leading to most of the bad takes. The implication for procurement is straightforward: the cost differential between running frontier on everything vs. routing intelligently across frontier commodity tiers is now 100x to 500x on equivalent workloads. That delta is the rent (and likely, the most valuable infrastructure business that has yet to be built). Whoever builds it will capture more durable margin than either the layer above or the layer below. The labs know this, which is why they're racing to wrap routing into their own platforms. Most enterprise buyers haven't bothered to figure it out (they're too busy building stuff for their business to think about how to optimize the cost of doing it), which is why their spend looks categorically insane. The Uber analogy gets reached for a lot in this conversation. It doesn't hold. Uber subsidized rides on the way to a clean rent-extraction end state (i.e. own the network, raise prices, let the unit economics work). AI labs can't run that playbook, because the terminal state is multiple breakthroughs away and the "extraction" time for a frontier model is 12-18 months. Fortunately for them, their strategic capital is patient in ways venture money isn't, and their end state is bifurcated, not unitary. The right-sizing pressure people are looking for isn't going to land at the model layer. It's going to land at the application layer, on every company that raised on the assumption that commodity inference would stay subsidized long enough to build a moat. That subsidy is ending, but the labs aren't the ones who have to absorb the cost; it's their customers' customers who will end up holding the bag. The question that follows naturally is: how do the labs absorb training costs that increase 10x per generation if commodity inference falls to zero. It has an answer the labs say out loud and most analysts ignore. They don't absorb it through token margin. They absorb it through a combination of (1) enterprise bundle contracts, (2) (eventually) through agentic outcome pricing & (3) ultimately through strategic capital treating training CapEx as optionality on capability scaling as opposed to cost-of-goods. The right comparable for an AI lab is late-stage pharma. Both have decade-long burn, binary terminal values & a capital base sized to the option vs the current P&L. That model survives commodity compression in a way SaaS economics can't. It also fails differently when it fails, which is the part the writedown discourse hasn't started pricing in yet. The "subsidy era ending" framing sounds good, but it's analytically wrong. The subsidy era didn't end. The undifferentiated consumption era is ending.

The disclosure included 3,600 trades in 3 months That’s around 40 trades a day Many of them are $1,000 - $15,000 a trade 1,000 of different stocks were traded Does anyone really think he’s managing the account?

If the argument here is that ringing up stuff that customers selected from the store you work in = "making the company money"....I think we need to take a step back. You didn't design or build the store You didn't vet/source the goods sold You didn't pay for the inventory You didn't pay the OpEx (electricity isn't free!) You didn't advertise/market the store You took no risk. Also: if you "make" the company $2,000 in gross, for most retail that's ~$200 net. Your $71 "made" is actually ~$100 to the company (FICA, fringe, compliance). So, you got ~25% of the net value you "created". That's pretty fair, wouldn't you say?

The structural advantage in any system with multiplicative dynamics is (almost) always lower variance at acceptable mean, not higher mean at any variance.

The more I look at consumer behavior, the more convinced I become that the sum of our vices remains constant - the only thing that changes is the underlying distribution.

I've been saying this for 2 years. Interest stacks / LALs for niche-y products make mathematical sense. Instead of asking the algorithm to surface patterns across an audience of ~270M people, you compress the search space to 10M people. End result? (1) Better prior (if real buyers are ~0.5% of the broad audience better targeting increases density to 5%, you've raised the probability of a conversion by 10x before learning); (2) faster signal acquisition (Meta needs ~50 conversions/week/ad set to exit learning. Higher buyer density = more stable customer acquisition = less wasted spend faster learning exit. This is particularly true with LALs, because they do something IS can't: they cluster in Meta's latent feature space capture unique, non-intuitive buyer/prospect signatures that no IS could possibly surface. End of the day: broad can be great if you have a product in a known category with massive TAM (i.e. buyer density in the general population is sufficiently high that the costs of segmentation are higher than the benefits). But if you have a product that targets a niche buyer (or if you have multiple buyer segments), LALs/IS can make a lot of mathematical sense.

Remind me who Beyoncé, LeBron, MJ, Taylor Swift & Oprah “extracted” If you genuinely believe that “billionaires are inherently extractive” then I’m curious at what level that stops. Is it $500M? Because that person is ~7 years away from being a billionaire, assuming decent returns & no colossal mistakes. What about $250M? Only 15 years away. Is the SMB owner throwing off ~$5M a year, living on $500k and optimizing taxes to invest $2.5M a year in equities and some real estate (exactly what has been preached by people on all sides of the political spectrum - save more than you spend! Be fiscally responsible. Invest in your future) inherently extractive? Because the person that does that for 30-40 years, and generates 1 standard deviation above median returns, is either at or damn close to “billionaire” status. There’s a reason virtually every study of billionaires admits they undercount precisely this “silent” segment. I personally know multiple people who did exactly this. They aren’t flashy. You’d think they were well-off, but never guess they were closer to 10-figures. Are they inherently extractive?

The path that produced the billionaire is the same path that produced the millionaire next door, the same path that produced the comfortable retirement of every disciplined saver you know (like the school teacher who saves for 40 years and leaves $3M to her kids). AOC’s argument doesn’t actually target billionaires - it targets the mechanism that produces wealth at every level, applied at a threshold that seems rhetorically & politically convenient. That threshold won’t stay where she put it, because the moral framing doesn’t have a stopping point. Today, it’s billionaires. Tomorrow, it’s centi-millionaires. Then it’s “potential billionaires” (I.e. anyone with the capacity to earn & invest substantial money ($2.5M a year) over a sufficiently long time horizon). The doctor with 3 offices & a net worth of $20M is wealthy beyond what most Americans will ever see. So is BigLaw lawyer with $30M saved after 30 years of practice. So is SMB owner with $50M. Every one of them is a ‘billionaire in waiting’ - the only question is investment returns & time. Every one of them becomes a target once the precedent is set that majority preference can override individual wealth accumulation. The honest version of this debate isn’t ‘should billionaires exist.’ It’s ‘should disciplined wealth accumulation be permitted at all.’ If the answer to the above question is yes, then the billionaires are legitimate by extension. If the answer is no, then we’re not arguing about billionaires - we’re arguing about whether the savings account, the IRA, the family business & the patient stock portfolio should exist as a path to anything meaningful. The only differences between the disciplined saver and the billionaire are: (1) how much they invest, (2) where they choose to invest their capital and (3) how long they’ve been doing it. If we’re going to debate this, let’s have an honest debate @AOC @RobertMSterling figured you might have a thought about this.

For those who don't know about Congressman Khanna: He's a sitting member of Congress & one of the ~600 most powerful people in the world - including other heads of state, members of the US Congress, CEOs, POTUS/VPOTUS, Cabinet Members, etc. His net worth is far in excess of $100M (one of ~29,000 people world wide with a 9-figure net worth - or Top-0.0004% of the population). That seems pretty wealthy. He is - definitionally - one of the "wealthy & powerful"

Let’s address some things here: 1. ~$65M of this is one-time comp to make Nichol whole for the loss of his unvested Chipotle stock. 2. When you compare steady-state FTE:CEO pay, you end up with a ~$25M:$40k ratio, or 625:1. The S&P 500 average is 285:1. The delta there is explained by relative size/complexity (Starbucks is a top ~25% F500 company, so CEO comp should be higher than the mean). 3. Any residual difference is bundled pricing of (a) base comp for the role, (b) lock-in premium against poaching, (c) insurance against unplanned transition cost & (d) signaling premium that affects succession candidate quality. But let’s make the bigger point: the CEO of Starbucks’ pay should not be compared to the median worker. You wouldn’t compare LeBron’s comp to be median stadium usher. The relevant question is whether Nicholas is paid fairly relative to the other CEOs capable of executing a turnaround of a $120B company. Given his record at Chipotle ( 700% gain), I’d argue Nichols’ steady-state comp is fair or even below market.