Darren

Developing an open framework for AI-mediated discovery — how AI systems build trust, resolve decisions and form defaults.

[Global Markets Investor @GlobalMktObserv]

Anthropic's finances are deteriorating ahead of its IPO:

Anthropic reported +$16.96 billion in Q3 2026 revenue, up +47% QoQ.

However, quarterly revenue growth slowed sharply from +144% in Q2, as price reductions and contractual discounts offset revenue by ~$600 million.

Meanwhile, Anthropic's operating profitability lasted just ONE quarter as the company swung back to a GAAP operating loss of -$567 million in Q3, following operating income of +$974 million in Q2.

This comes as compute and infrastructure expenses surged +77% QoQ to $9.9 billion, accounting for 58% of revenue, up from 48% in Q2.

In effect, every $1 spent on computing infrastructure generated just $1.72 in revenue, down from $2.08 in Q2, marking the first quarterly deterioration in this key efficiency metric.

Overall, Anthropic's net loss reached -$23.3 billion in Q3, bringing its cumulative net loss for the first 9 months of 2026 to a staggering -$104 billion.

While most of the quarterly net loss came from non-cash valuation adjustments, the underlying operating business also slipped back into the red.

Excluding the $21.9 billion in non-cash valuation adjustments, Anthropic's Q3 net loss was -$1.4 billion.

Anthropic's trillion-dollar AI ambitions are increasingly colliding with the financial reality of an infrastructure race where computing costs are rising faster than revenue.

[Darren @letsgoelsewhere]

Evidence #1018
The is no moat in AI

[Guillermo Flor @guilleflorvs]

BREAKING: Revolut moved ~2,000 engineers onto self-hosted GLM for coding and pushed Claude Code and Codex to the exception list.

Per David Abitbol, who leads AI ops there: 56 Nvidia B200s, a blended 30 cents per million tokens, roughly a tenth of what they paid Anthropic and OpenAI. Engineering lead time improved 70%.

Here's the part that matters: the gains showed up when tokens stopped being rationed. Engineers quit saving AI for the hard tickets and pointed it at everything.

Frontier models become the exception, open weights become the default. That's a pricing problem for every lab heading to IPO.

h/t @OliverMolander

Would you move your team off frontier APIs if the bill dropped 90%?

[Darren @letsgoelsewhere]

AI token prices just hit another record low. And the next wave hasn’t even landed yet.

The price of AI intelligence is collapsing.

And it’s about to get worse for anyone betting on scarcity.

DeepSeek and Qwen have new models coming. More capable models. Lower inference costs. More tokens for every dollar.

The direction of travel is becoming impossible to ignore:

Better.
Smaller.
Cheaper.
More efficient.
More open.

This is fantastic news for AI adoption.

But it creates a very uncomfortable question for the companies that have built their valuations around selling intelligence at a premium.

What happens when the product you’re selling becomes a commodity?

The market is already questioning the economics of the infrastructure buildout.

Now the economics of the intelligence itself are moving against them.

The AI opportunity is enormous. But the moat isn’t intelligence. It’s what you can reliably resolve with it.

That’s where the Elsewhere Systems Framework comes in.

https://t.co/tdb1X5Fif4

[Ed Zitron @edzitron]

Ramp estimates Anthropic Q4 business spend growth is trending toward 15% to 25% growth quarter over quarter, down from 47% from Q2 to Q3. Seems like a sign that revenue is decelerating

[Darren @letsgoelsewhere]

@GeorgeGammon Hyper-failures! Very funny. That could catch on.

[Darren @letsgoelsewhere]

@MauiBoyMacro What’s $900bn between friends?

[CG @cgtwts]

holy shit, deepseek just open-sourced a super-agent that gets work done almost twice as fast, and they’re giving it away for free.

[Darren @letsgoelsewhere]

Where are the moats now?

[Andrej Karpathy @karpathy]

Weekend hack: HackerTube. I was annoyed with YouTube's recommendations so I had Claude filter down 43M comments over 20 years of HN to 6,454 videos that keep being recommended. *Significantly* better viewing experience, ~2 hours with Opus 5.5:
https://t.co/JIhXN7OSh1

- Excellent recs without any logins (state is kept in local browser session)
- Thumbnail is a representative frame from middle of the video instead of the super duper A/B test hyper-optimized user uploads
- Underlying data from the comment analysis as one zip file, point your Claude at it and ask for your own version
- HackerTube Longs ✊

A bit more on the process:
I use Claude Code (desktop app for Mac), Opus 5.5 on Medium effort as my default. I only talk to Claude and have it do everything, the only "manual work" I helped with in tandem was in the browser (reminiscent of MenuGen): 1) I had to create some kind of API key from Google Cloud UI to query YouTube for video metadata, 2) I booted up a VPS box for serving and 3) I bought the domain name. Once the box was up I gave Claude the IP and told it update, secure, harden, deploy and optimize the code, and it did the whole thing.

Many awesome projects that would have normally taken weeks/months even can now be few-shotted for breakfast. What else are we not thinking of.

[Darren @letsgoelsewhere]

@EdgeCGroup @Barchart There’s so much complacency in retail. I’ve seen that over on Reddit first hand

[andrew jennings @andrew_jennings]

SEO is dead, and Reddit's CEO just confirmed it on their earnings call.

He said AI search doesn't send traffic anymore, to Reddit or anyone else. If Reddit can't get clicks out of Google, a brand blog with 40 posts on "best gifts for dad" has no chance.

52% of orders on one of my brands start with a Google search. Brand name search is safe. But searching "best gift under $200" now gets an answer in the AI chat, and the AI read my article to write it. Nobody clicks through.

So any SEO content line in a 2027 budget is gone. That money moves to the traffic you own - your email list, your SMS list, and creators who send people to search for your brand name. The product page gives all facts in plain text, so when the AI names 3 brands, one of them is yours.

People are calling that AEO. Same job as SEO, written for an agent.

[Benjamin De Kraker @BenjaminDEKR]

Two weeks ago Grok Bot was an incredible success story powering SpaceXai's revenue growth

Now it's unsustainable and loses money (???)

[Darren @letsgoelsewhere]

@StinsonJed @KyleReidhead it

[Darren @letsgoelsewhere]

@heyshrutimishra This 660× efficiency gap challenges the economics of the AI scaling boom!!

[Darren @letsgoelsewhere]

@GavinSBaker Yes there was huge over hiring during Covid. But aren’t you just highlighting all
that sweet VC money going round and round registering revenue with everyone. It’s beautiful.

Apollo have a different set of charts.

[Darren @letsgoelsewhere]

Even if you believe it (I really don’t, it’s meme like) you have to go through 12 months of extreme pain. Proper head in a vice pain. Are we just supposed to skip past that and dream of all these wonderful layoffs. It’s all nonsense. Markets aren’t stupid, they’ve locked onto it now.

[Darren @letsgoelsewhere]

@samongaro_ It’s the new moat! I’ve built a business on it. It’s actually transferable in the AI era. Trust has to be earnt but it can be networked to help you accelerate building it. ✌️

[Darren @letsgoelsewhere]

@ChizuruVolpe No it’s the other way around. It’s a discounting mechanism that continually tries to get ahead of what’s coming. Otherwise it would be a free money machine. That would be nice but sadly not.

[Darren @letsgoelsewhere]

@puckrin That’s the scariest chart in the whole wide world. 👻