Welcome back to your daily memorandum talking tech, business, AI, markets, and more. 🗞️

In today’s edition, we are tackling the following:

⚡ xAI launches Grok Code Fast 1, a compact coding model.
🖥️ Microsoft debuts MAI-Voice-1 and MAI-1-preview for Copilot.
💰 Dell reports $29.8B Q2 revenue but issues weak guidance.
🔐 TransUnion breach exposes personal data of 4.4 million customers.
☢️ Russia supplies North Korea with weapons tech, boosting its nuclear program.

TECHNOLOGY

• Microsoft unveils powerful new home-grown AI models

  • Microsoft introduced MAI-Voice-1, a speech model that generates a minute of audio in under a second on a single GPU.

  • It also revealed MAI-1-preview, a large language model trained on 15,000 Nvidia H100 GPUs aimed at consumer Copilot features.

  • The move signals a shift toward self-reliance in AI innovation, reducing dependency on external models like those from OpenAI.

• Elon Musk’s xAI enters agentic coding with new ‘grok-code-fast-1’

  • xAI released grok-code-fast-1, a compact and cost-efficient autonomous coding model, now free for select partners.

  • The launch places xAI alongside Microsoft and OpenAI in the growing domain of AI tools that write and manage code by themselves.

  • This comes at a strategic moment—xAI has also filed an antitrust lawsuit citing unfair competition by Apple and OpenAI.

• Anthropic begins training AI on consumer chat data unless users opt out

  • Starting September 28, user chat and coding transcripts will be used to train Anthropic’s models unless they actively opt out.

  • Data retention extends to five years for consenting users; older inactive sessions won’t be used unless resumed.

  • The company emphasizes automated filtering, no data sales, and user control—but critics warn of privacy risks and UI nudges.

BUSINESS

•  Dell beats expectations with record Q2 results

  • Dell posted $29.8 billion in revenue, up 19% year-over-year, driven by booming AI server sales.

  • Despite strong earnings, the stock fell over 5% as guidance for Q3 missed analyst forecasts and raised caution.

  • The company now expects $20 billion in AI server shipments this year and raised its full-year revenue outlook.

• U.S. begins publishing GDP data on public blockchains

  • The Commerce Department started releasing GDP figures through blockchains like Bitcoin, Ethereum, Solana, and six others.

  • Officials describe it as an extra publication channel, not a replacement, aimed at promoting data transparency and crypto adoption.

  • Critics say the initiative is symbolic and lacks clear benefits, highlighting the tension between centralized and decentralized technology.

• TransUnion confirms breach affecting 4.4 million customers’ data

  • A third-party application breach exposed personal information of 4.4 million customers—but no credit or financial data.

  • The incident, which occurred on July 28, prompted investigation, notifications, and efforts to strengthen data protection protocols.

  • The breach raises fresh concerns about the cybersecurity risks in using external service providers in sensitive industries.

MARKETS

WORLD

• North Korea advances nuclear and missile capabilities with Russian backing

  • Russia supplies North Korea with air-defense systems and weapons tech in exchange for Pyongyang's support in Ukraine.

  • North Korea has unveiled tactical nuclear warheads, MIRVs, hypersonic glide vehicles, and strike drones, accelerating its weapons progress.

  • Analysts warn that the alliance diminishes denuclearization and shifts the strategic balance in Northeast Asia toward a more dangerous rivalry.

• Protesters pelt Argentina’s Milei with rocks amid corruption backlash

  • Demonstrators attacked President Milei's motorcade with stones at a campaign rally as corruption allegations linked to his sister intensified.

  • Milei was evacuated safely; no injuries were reported, but the incident underscores mounting political tension ahead of key elections.

  • The attack comes amid a federal probe and could further erode public trust in his administration’s anti-corruption platform.

• U.S., U.K., France, and Germany to reimpose UN sanctions on Iran

  • European powers are moving to activate the ‘snapback’ mechanism under the Iran nuclear deal to reinstate UN sanctions.

  • The shift follows failed diplomacy efforts, with a 30-day window offered for Iran to resume meaningful negotiations.

  • If ignored, sanctions could automatically resume by mid-October, resetting pressure on Iran's nuclear program without new UN approval.

CONTENT

• Anthropic’s $10B round, Klarna’s IPO, onside a16z’s 72 deal

  • Marc Benioff weighs Salesforce’s AI strategy, talent wars, and lessons from Palantir on forward-deployed engineering.

  • Discussion covers Anthropic’s $10B raise, Klarna’s valuation swing from $45B to $6B to $15B IPO.

  • a16z’s ‘seed machine’ investing approach—72 bets compared to Sequoia’s 27—sparks debate about consensus versus conviction.

• How 80,000 companies build with AI

  • Microsoft’s Asha Sharma outlines AI product trends, from “products as organisms” to the shift toward agentic task networks.

  • She explains Microsoft’s ‘seasons’ planning framework, post-training investment trends, and the rise of code-native interfaces.

  • Conversation highlights predictions of an ‘agentic society’ where AI agents outnumber humans in future corporate structures.

• How to reinvent yourself in your 30s and 40s

  • Robert Greene emphasizes deep work, silence, and boldness as keys to reinvention in midlife.

  • He discusses lessons from childhood, Buffett’s inner scorecard, and the power game dynamics shaping success.

  • Greene shares practical advice on finding purpose, embracing authenticity strategically, and turning setbacks into opportunities.

FUTURISM

• Microsoft launches its first in-house AI models

  • Microsoft unveiled MAI-Voice-1, a speech model generating one minute of audio in under a second on one GPU.

  • It also introduced MAI-1-preview, a large language model trained on 15,000 Nvidia H100 GPUs for consumer applications.

  • These models will integrate into Copilot, signaling Microsoft’s shift away from dependence on OpenAI partnerships.

• MathGPT expands to over 50 colleges offering ‘cheat-proof’ AI tutoring

  • MathGPT avoids giving direct answers, instead guiding students with Socratic questioning while supporting instructors with monitoring tools.

  • Now used by 50+ universities, including Penn State and Tufts, it integrates seamlessly with Canvas, Blackboard, and Brightspace.

  • Professors control access and features such as attempt limits, unlimited practice questions, and accessibility customizations.

• Taco Bell rethinks AI voice drive-thrus after glitches and trolling

  • Taco Bell’s AI voice assistant faced glitches, long delays, and trolling, including prank orders like 18,000 water cups.

  • CTO Dane Mathews said the chain is reassessing AI’s role, especially at high-volume locations needing human backup.

  • The company remains committed to AI but plans hybrid systems blending automation with staff support.

EXTRAS

• Israeli protesters push harder for hostage deal as divide with Netanyahu deepens

  • Hundreds of thousands rallied in Tel Aviv, demanding a ceasefire and hostage release, clashing with Netanyahu’s escalating war stance.

  • Families of hostages spearheaded strikes and protests, insisting negotiations are the only path to secure loved ones’ return.

  • Netanyahu dismissed the demonstrations as aiding Hamas, fueling a widening disconnect between government leadership and opinion.

• Nvidia CEO says the AI boom may be ‘far over’ after disappointing sales forecast

  • Nvidia warned investors of softer demand, suggesting the AI hardware frenzy may be cooling following its cautious sales projection.

  • Despite mixed signals, long-term AI interest remains, though investors are watching for sustained enterprise and consumer adoption.

  • The company’s cautious tone follows explosive growth in recent years, raising questions about how deep and lasting the AI boom is.

• Meta is racing to launch its Llama 4.X model by year-end to revive AI momentum

  • Meta’s new Llama 4.X, developed by its elite Superintelligence Labs team, aims to launch by year-end to regain AI traction.

  • The iterative release responds to the underwhelming reception for April’s Llama 4, addressing bugs as Meta pushes innovation.

  • Amid intense internal pressure, Meta is betting on next-gen breakthroughs to restore confidence and stay competitive in the AI race.

AND MORE

• Polish F-16 crashes during airshow rehearsal in Radom, pilot dies.

• S&P 500 and Dow close at record highs amid resilient AI-driven rallies.

• Explore the full catalog of bourbon investment offerings. Download your copy of the 2025 CaskX Distillery Catalog now.*

• United States declines to participate in UN human rights review, drawing criticism.

• Russian missile and drone strikes kill at least 21 in Kyiv; EU and UK protest.

• U.S.—GDP revised upward to 3.3% in Q2, boosted by AI investment.

• TransUnion confirms data breach affecting over 4 million consumers.

• China extends anti-dumping duties on phenol imports from multiple major exporters.

• You get high-impact people. No upfront costs. Just proven talent, delivered fast.