token optimization is the new performance engineering: biggest model, trimmest context.
maxdosing on intelligence, microdosing on memory.
every token you leave in the window pays rent on every turn. the best engineers in 2026 are basically landlords auditing their own leases.
He’ll still tell you gold is a relic of the past, but the drive he sparked in generations of students is unforgettable. Few leave a mark like that. https://t.co/fO5YuPX5qc
Nvidia is the world’s most valuable company, generating $120 billion of annualized net income selling highly engineered chips and software that have required decades to perfect.
So why have they felt it necessary to issue a rebuttal to short sellers like Michael Burry?
Three reasons:
1. It is difficult to sustain 75% gross margins when some of your biggest customers are unprofitable, and your chips are 80% of their costs.
2. Those unprofitable customers are dependent on investors to fund their losses. Investors can move in herds, and their confidence about AI has been made fragile by OpenAI’s "manifestly ridiculous" $1.4 trillion of compute commitments.
3. More broadly, Nvidia is unsure where value will land in the AI business. Anyone using the latest models knows how they can lower the cost of legal advice or reading an MRI. Our own work finds that reading an MRI costs $150 for a human doctor compared to $0.15 for an AI model. This 99.9% spread is the biggest we have ever seen in business. But who gets it? The chip maker? The model maker? An application maker? The doctor? The patient?
Nvidia’s real competitor is Google, which generates $140 billion of annualized net income and owns important assets across the full stack of AI: TPU chips; Gemini models; data from 3.5 billion humans using its services; and distribution across multiple surfaces like Search, YouTube and Cloud.
Nvidia’s strategy is to support its customers’ efforts to raise investor capital until their businesses capture enough of that 99.9% spread to be self-sustaining, and to hope that its own 75% gross margins are sustainable. A loss of investor confidence impairs this strategy, which why the rebuttal was issued.
Google’s strategy is to drive costs down across the full stack, to protect Search's advertising business and to generate profits wherever they end up landing. Its founders have said they are willing to bankrupt the company rather than lose in AI.
Investors chasing winner-take-all outcomes across the AI stack may be disappointed. Based on how unit economics look today, much of AI looks like a commodity business in which low cost wins.
Nvidia’s and Google’s strategies are in obvious conflict. Some investors will make a lot of money and others will lose a lot. More will be revealed as this unfolds. Pass the popcorn.
Gold ripping past $4,000 and Poland buying more of it than China is pure Fujiwhara effect — two monetary storms circling the same center. When central banks start dancing, the system’s about to change.
the long-run asymmetric upside in AI probably lies in these “distant” plays, and the current hype is misallocating too much capital to already mature (or semi-mature) frontiers (20 min mark) $idna $smh
One of the best policy changes in the last fifty years. With congestion pricing you allocate using price ("How much am I willing to pay to drive into Manhattan?"); otherwise you allocate using time ("How much time am I willing to spend driving into Manhattan?")
A is dramatically more efficient than B.
@JonErlichman The U.S. venture success story is deeply intertwined with the dollar’s reserve status and Fed monetary capabilities. These factors create an environment of abundant capital, low borrowing costs, and systemic confidence that fuels risk-taking.
@profplum99@chamath Sufficiently red-pilled. Even more disturbing is how much concentration risk has poured over the whole passive ETF complex https://t.co/Qkpm3THceC
@jamesschwartz Not saying it's desirable or likely but I do think there are real fracturing risks to be taken seriously, starting with acknowledgement that domestic dysfunction is principally the reason this was mainstreamed and not just MAGA/Oligarch opportunism.
I suspect many Canadians, especially those feeling politically homeless right now, share a deep anxiety about our country’s future.
On Nov 21, while discussing US vs Canadian business dynamics over lunch...
Do you think replacing Trudeau will permanently close this chapter, or is there more required to fundamentally alter Canada’s trajectory?
Leadership matters, but I wonder if we’re grappling with deeper structural and cultural issues that one election cycle can’t fix.
This wouldn’t have entered the realm of discussion but 10 years of Trudeau leaves some Canadians willing to entertain what is trolling by Trump as a real possibility. Instead of responding, let’s get rid of the Trudeau-lead govt and start repairing the damage and rebuilding from the all of the carnage and division.
I’m not endorsing it, but in a decade of tectonic shifts, this possibility feels more like a symptom of deeper fractures than simple MAGA theatrics.
Would love to hear what others think about this.