As I wrote this, I saw X go into meltdown over tokens.
You've seen the headlines: “Uber blows yearly AI budget in just one quarter.” “Meta employee burns 281 billion tokens in April.”
But, the problem isn't spending. Spending works. Since 2023, the top quartile of our AI spenders doubled their revenue. The bottom quartile? Flat.
It's blind spending. We don’t know which spend worked.
A sales team has qualified leads. A support team has resolved conversations. These are units you can measure against. All a token tells you is the meter ran, not whether the work was worth it or not.
Finance says, “half the budget,” engineering says, “double it” and you don’t know who’s right because there is no shared language of value. There’s no attribution, and no attribution means no allocation.
For example, right now, all work, no matter the size or shape, defaults to frontier models. But meeting summaries and calendar updates don’t require GPT-5.5 Pro.
In isolation this seems trivial, but re-route just 10% of a $10M AI bill from frontier to GPT-4 level intelligence you’ve saved nearly one million dollars. This sounds like a made-up stat — it’s not. It truly is that much cheaper.
This is the future of finance: not blindly rubber-stamping or rejecting AI spend, but allocating it with the same rigor companies apply to headcount.
1/ Sharpest breakdown of the token vs labor question I’ve heard anywhere, on @HarryStebbings’s @20VC.
@rodriscoll framed the one number that decides everything: for every dollar you pay an engineer, how many dollars go to tokens? 10%? 13% like EDA tooling? 33%? The whole AI valuation stack balances on the answer.
2/ The wild data point: Mercor already spends more on tokens for its internal agents than on salaries. A $10B company, 300 people, and inference has already crossed payroll. That’s not a 2027 projection. That’s now.
3/ @jasonlk takes it further: by year-end, leaders stop getting a headcount budget and start getting a dollar budget. Faced with that choice, he picks tokens over humans. The cruel layoff email writes itself: “it’s not you, we needed the tokens.”
4/Why it’s live now: everyone cranked in Q1, CFOs ran the accruals mid-quarter and found they were 10x over. Uber’s patch is to cap engineers at ~$1.5k/month. Transitory fix or the new normal? That’s the trillion dollar question, literally.
5/ But here’s the trap in the “tokens replace humans” story: orgs re-bloat. When you can ship 3x the products, you need humans to manage 3x the products. One portfolio co is ending the year with triple the product lines and a bigger EPD team than they’d like. Fewer humans per product, far more products.
6/And the contradiction at the heart of it all: a whole industry building to automate white-collar work and predict mass unemployment in 3 years, and every person you talk to says “I’ve never worked this hard, I’m at it 24/7.” We’ll automate all work AND we can’t hire fast enough. Both, at once.
(Token/budget segment ~43m–1hr) https://t.co/i8CJtImnp3
It’s tough but we used Squarespace for all our non-core websites until Fall 2025. And no longer have any use for it or Wix.
Since then, we’ve vibed everything, and those websites are so, so much better
The open question is Wordpress. It’s still an epic CMS and “headless” content platform.
But I’d so much rather vibe the front-end …
One round. $65 billion. That’s on the order of a full year of European VC across every country and stage, and well over 10x Australia’s total annual venture spend.
We've raised $65 billion in Series H funding at a $965 billion post-money valuation, led by @AltimeterCap, Dragoneer, @Greenoaks, and @sequoia.
This investment will help us advance our research and expand our capacity to meet growing demand for Claude.
Shoutout to @OpenAI and @gabrielchua for putting on a cracker of a Sydney hackathon.
The part I love isn't the polished demos. It's watching builders on the front lines hit something that used to be a hard blocker and just walk through it.
One team built a Game Boy emulator from scratch over the weekend. CPU cores, PPU scanlines, framebuffers, the lot. Outside the room, same pattern with people playing in robotics, hardware, embedded. The blocker was never the idea (trust me, I can relate). It was the domain tax. Those barriers are collapsing with tools like Codex, the agent stack, they don't lower the ceiling. They collapse the time between 'I wonder if I could' and 'oh shit it works.'
Bet on the front-line builders.
Budget night, big structural changes are a swing without much evidence they’ll work. Would have nice to see:
1.Gas tax
2.Public sector headcount reduction
3.Tighter immigration (cools demand)
4.Serious housing build-out (more supply)
5.Trimmed NDIS and programs spend
You will flip up to a US topco, and when you exit for $1 billion+ you won’t pay CGT here. The Australian tax payer, and talent pool will be worse off. Also think of hiring talent from larger companies with high pay, the ESOP shares now look 50% lower, meaning smaller companies growing faster and helping productivity will be hurt.
Ghostty is leaving GitHub. I'm GitHub user 1299, joined Feb 2008. I've visited GitHub almost every single day for over 18 years. It's never been a question for me where I'd put my projects: always GitHub. I'm super sad to say this, but its time to go. https://t.co/DQDemHdytV