Paul Marino

[Rohan Paul @rohanpaul_ai]

Google released Gemini Southeast Asia Report.

- 900M+ monthly users globally by Apr-2026, while Gemini’s Southeast Asian user base more than doubled in last 12 months.

- Singapore has the world’s highest Gemini adoption per capita, while its users prompt Gemini roughly 10 times a day—the highest engagement in Southeast Asia.

- 42% of prompts already involve voice, images, videos, documents or other non-text inputs, meaning nearly half of Gemini interactions are multimodal.

- Gemini became the fastest-growing major Google app in Southeast Asia.

- Prompts per Southeast Asian user more than doubled

- Under-25s have more frequent and longer Gemini conversations—and give more detailed spoken prompts—than every other age group in Southeast Asia.

- Southeast Asian users create over 5 billion images, while almost 1 million songs were generated in March 2026 alone after Lyria 3 arrived.

- 7 in 10 Gemini prompts are written in native languages—rising to 89% in Vietnam, 87% in Thailand and 84% in Indonesia.

- Indonesia is overwhelmingly mobile-first: 82% of its prompts come from phones, and one in every two prompts includes an input beyond text.

- Vietnam ranks first in Southeast Asia for academic usage: over 160,000 students use Exam Prep every month, and more than 55,000 teaching-related prompts are submitted daily.

[Stock Talk @stocktalkweekly]

*STRIPE AND ADVENT EQUITY FIRM MAKE JOINT OFFER TO ACQUIRE PAYPAL FOR $53 BILLION, OR $60.50 PER SHARE (+28% MARKET PREMIUM) $PYPL

[Warren Pies @WarrenPies]

JUNE CPI

Waller gets his wish.

Cold inflation across the board.

Supercore -.20% month-over-month...Largest decline since May 2020 (Covid lockdowns).

[Mike @Doranimated]

Dutch intelligence agencies published an official warning: Russian state hackers are hacking the security cameras of private companies in the Netherlands, other NATO countries, and Ukraine

The goal: spying on weapons shipments to Ukraine. Hackers tap into cameras along Dutch military transport routes.

How it works: Many internet-connected cameras are poorly secured. Hackers use simple online scanning tools to find them and break in.

The Netherlands, as a transit hub supporting Ukraine, is a prime target. The agencies are publishing security guidance to help organizations lock down their cameras.

[Mike Zaccardi, CFA, CMT 🍖 @MikeZaccardi]

30yr yield just 3bps from the highest weekly close in 19 years

[Mike Zaccardi, CFA, CMT 🍖 @MikeZaccardi]

The Dollar’s Hidden Dependence on the AI Trade... Net foreign inflows into US equities have surged to a record high, driven in large part by overseas investors seeking AI exposure

Torsten at Apollo

[JT @lokoyacap]

Oh

https://t.co/MgocwL5QOG

[Paul Marino @pmarin0]

@HayekAndKeynes Heart attack?

[Petr Pinkhasov @pinkhasov]

Monitor your usage, straight in CLI

https://t.co/smNlrg6eUc

[Craig Fuller 🛩🚛🚂⚓️ @FreightAlley]

The industrial economy is ramping significantly as we can see in the rail volume data that comes out each week

[leo 🐾 @synthwavedd]

🚨 SCOOP(s):

- GPT-5.6 will be the final model in the 5.x series. GPT-6 is slated to launch in about a month, earlier than expected, and possibly even later this month
- GPT-6 will be based on a new, significantly larger pretrain (versus the ~4T 5.5/5.6 'Spud' base)
- There is lots of excitement at OpenAI over this new base, which they believe will be much better able to compete with both Fable 5 and upcoming 5.1, targeting a similar release window. OpenAI initially intended to continue with Spud through GPT-6, but decided against it
- On the topic of Fable 5.1, it is in the late stages of the pipeline at Anthropic and a release is expected "in the coming weeks"
- On the other side of the globe, DeepSeek are preparing for an imminent launch of V4 GA, which seems likely to be on par with or better than GLM-5.2, and have begun work on a new, larger model that will compete with the upcoming 2.7T MiniMax Pro

[Paul Marino @pmarin0]

@BrettHarrison @AgustinLebron3 @ComputeDesk Congrats. Needed innovation. A forward curve for depreciation would be very helpful.

[Brett Harrison @BrettHarrison]

Announcing ComputeConnect, the financial industry’s first exchange-for-physical (EFP) network for compute, coming soon from Architect and @ComputeDesk. ComputeConnect links US exchange-traded compute futures to compute capacity delivery.

Exchange-listed cash-settled compute futures are entering US markets to correct course on the current AI economy, reorienting debt to long-term growth:

• Creating price discovery and transparency independent of any single capacity provider.
• Establishing a forward curve for measuring deprecation and forecasting supply and demand.
• Providing financial hedges for compute consumers and producers.
• Enabling hedge funds, ETF companies, and traders to gain long and short financial exposure to compute.

US cash-settled compute futures lack a physical delivery mechanism, and ComputeConnect fills this gap.

Existing physically settled futures such as energy and agriculturals require their clearing house (DCO) to set a uniform standard for the grade and delivery method for the underlying commodity. Compute, by contrast, is highly fragmented, heterogeneous, and rapidly evolving, making it infeasible for any single DCO to define and enforce comparable standards.

ComputeConnect establishes a network of compute capacity providers and links the network with Architect’s US futures products using exchange-for-physicals (EFPs), OTC contracts in which futures positions are exchanged for the assets the futures track. EFPs allow counterparties to negotiate the grade, timing, location, and other characteristics of the commodity along with a basis tied to the futures settlement price.

ComputeConnect will

• Build a network of capacity providers and capacity marketplaces.
• Establish an open protocol for members of the network to receive delivery requests and advertise available GPUs.
• Publish standard basis tables for different SKUs, memory configurations, and locations for GPUs.
• Book the futures legs of the transactions to Architect’s DCM, the American Innovation Exchange.
• Facilitate and guarantee delivery of capacity using Compute Desk’s ComputeClear platform.

The advancement of US AI is constrained at every link in the supply chain: materials, power, chips, capital… The American Innovation Exchange, ComputeConnect, and our industry partners aim to secure compute’s dominance as an American asset class.

[Claude @claudeai]

We're extending access to Claude Fable 5 on all paid plans through July 12.

[Daily Chartbook @dailychartbook]

"59% of S&P 500 technology stocks are now trading at least 20% below their 252-day highs, a common definition of a bear market."

@DeanChristians @TPMRSignals

[Paul Marino @pmarin0]

@AgustinLebron3 Follow the money. We need to not let foreign financing move social arguments to this extent.

[clem 🤗 @ClementDelangue]

A study from @Stanford showed that 71.3% of chatgpt queries could be accurately answered by a local model. I suspect a major part of enterprise AI workloads could be run locally too for free (compared to the massive costs of frontier API cost).

Also, it reduces the risk of these workloads being taken away from you because you own the models instead of renting them - which sounds like a good idea these days haha.

That's why we're introducing the ability for everyone to filter AI models on @huggingface based on your local hardware.

For me, there are 800k+ public models that fit on my M5 24GB and that I can use easily thanks to llamacpp.

Let's go local AI!

[Paul Marino @pmarin0]

@TheLoopyBlgger @23savage____ Knicks weren’t going to resign him after signing Alvarado & Shamet and Dolan wanting to stay under the cap. Good for him to make his own decision.

[Paul Marino @pmarin0]

@danshep55 Knicks weren’t going to keep him anyway after signing Alvarado & Shamet and Dolan wanting to stay under the cap. Good for Mitch to get best for him.

[Linda Raschke @LindaRaschke]

Fugly market award: Bonds - new momentum low and new price swing low. More to come eventually.