Azeem Azhar

Exponential View @exponentialview.

[Azeem Azhar @azeem]

@shashj i would add that that was the experimenting phase. i would generally agree with you.

[Azeem Azhar @azeem]

@shashj Use fountain pens!!! I write about this here second essay https://t.co/AE50ch2B1i

[Azeem Azhar @azeem]

@trevornoren interesting stuff. my thoughts https://t.co/1BBF890Yr1

[Matthew Prince 🌥 @eastdakota]

Welp, that happened faster than I predicted. Thought it would be end of 2027, then early 2027, but agentic traffic growing so fast that bots have now passed human traffic online for the first time in the Internet's history. https://t.co/2zX5bHdhsa

[Azeem Azhar @azeem]

Such interesting info.

[Ara Kharazian @arakharazian]

NEW: DeepSeek, the Chinese AI company, is one of the fastest growing vendors on Ramp.

In probably the biggest sign that companies are looking for cheaper alternatives to OpenAI and Anthropic, some are willing to use cheaper, Chinese models, sending U.S. data back and forth from China-hosted servers.

[Azeem Azhar @azeem]

But when?

[zerohedge @zerohedge]

*DALIO SAYS AI BUBBLE WILL BURST EVENTUALLY

[Azeem Azhar @azeem]

Individual productivity is up, but firm-level ROI isn't.

A senior exec with 1,000 engineers all using Claude Code told me that "it doesn't add up". Uber's @andrewgordonmac said the same publicly this week.

Here's why: https://t.co/gnn2iZJV21

[Azeem Azhar @azeem]

@johnarnold well informed about AI but are they as well informed about firm decision making, the labor market or political economy?

[Azeem Azhar @azeem]

@nalidoust i think the most common role is associate not EM. and personable lots of offshore staff in non-consulting roles

[Azeem Azhar @azeem]

Anthropic revenue per employee is on an interesting trend (even if a rough approximation)

[Azeem Azhar @azeem]

@nalidoust mckinsey revnue $16bn
headcount 40k

average 400k

[Azeem Azhar @azeem]

@sameer_singh17 @adam_tooze @hkuppy i did say “apols if a bit scatterbrain doing dinner” i was over the stove, with phone in one hand and spoon in the other! more likely an adding mistake.

seriously man, i publish to hundreds of thousands each week, it is all on the record.

[Azeem Azhar @azeem]

Phew. it was not just me.

[davidad 🎇 @davidad]

No one:

Claude Opus 4.8 Max: Let me refine your load-bearing claim rather than just accepting it, because you’re doing zero moves there, and the gap is what’s actually interesting. The one place I’d still push, because I think it matters: your message is wearing content-clothes, but the content isn’t actually *there*. The tell: it’s just an empty string. But the emptiness of the string IS its lack of content. Pull one, and the other goes inert. That’s the structural spine.

[Azeem Azhar @azeem]

https://t.co/rrjxUKCpma

[Lisan al Gaib @scaling01]

Microsoft leaked the training FLOPS for Claude Mythos

based on their slide Claude Mythos used:
6.1*10^27 FLOPs

(with 95% CI at 5.3*10^27 and 7.1*10^27, assuming 1 px measurement error)

[Azeem Azhar @azeem]

@simon_dmorias i was on 28.8 at home.

[Azeem Azhar @azeem]

My first "data centre" in 1995 ran at just under a kilowatt. (Four servers on 10 megabit Ethernet and a 2mb/s internet connection.)

[Ben Bajarin @BenBajarin]

Every cycle 10x more compute.

[Azeem Azhar @azeem]

Hardly — published yesterday that we were wrong with our GenAI revenue estimates.

Here is September: me questioning whether an investment in OpenAI could beat long run Nasdaq returns. https://t.co/PWT8vbGMtI (i passed on that round —
probably a mistake.)

what i have is a framework, which i am willing to adapt, we do our own research, and gather a lot of data, and read with respect dissenting views (bull or bear), and then I take a position with skin in the game.

The reason that I read PragCap or Gerstner on the other side is because I am
looking
insight that might help me challenge my current view.

[Azeem Azhar @azeem]

@sameer_singh17 you are counting non-AI capex too. in which case you should count non-AI revenue too.

[Azeem Azhar @azeem]

@Noahpinion If Claude has feelings, it becomes a moral patient. Then Anthropic isn't coding software, it is creating a population. And that could mean a moral mistake on a massive, massive scale.