Instagram | Sponsor this newsletter

My fellow AI explorers

Two stories crossed my desk this week that felt like opposite ends of the same boom.

One says the AI buildout is minting six-figure jobs for people who never set foot in a computer science lecture. The other says a 25-year-old who wrote the essay on where AI is headed just watched $35 billion evaporate in a matter of weeks, because being right about the technology and being right about the trade are two very different skills.

Add in a price war that is making the world's smartest software cheaper by the week, and you've got a genuinely strange moment for anyone trying to figure out where the money, the jobs, and the risk actually sit.

In today’s edition:

• 🔧 Nvidia's Jensen Huang says the real AI jobs boom is in plumbing and construction, not coding

• 💥 A 25-year-old's $45 billion AI hedge fund just got forced into a fire sale

• ⚡ 30-Second AI Play: how to slash your AI bill using model routing

• 🧮 Plus: OpenAI cracks 10 open math problems, DeepSeek's price war escalates, Google pulls AI satellite images, and Anthropic discloses real cybersecurity incidents

Write docs 4x faster. Without hating every second.

Nobody became a developer to write documentation. But the docs still need to get written — PRDs, README updates, architecture decisions, onboarding guides.

Wispr Flow lets you talk through it instead. Speak naturally about what the code does, how it works, and why you built it that way. Flow formats everything into clean, professional text you can paste into Notion, Confluence, or GitHub.

Used by engineering teams at OpenAI, Vercel, and Clay. 89% of messages sent with zero edits. Works system-wide on Mac, Windows, and iPhone.

Try Wispr Flow free

NVIDIA

Jensen Huang: Skip the Resume, Grab a Wrench 🔧

Nvidia CEO Jensen Huang has a message for anyone panicking about AI eating their job: the biggest opportunity might not be in a Slack channel at all; it might be on a job site.

Speaking with BlackRock CEO Larry Fink, Huang argued the AI buildout is triggering the largest infrastructure construction wave in human history, and somebody has to actually build the data centers, chip factories, and power grids that make the AI economy run.

Here's what stood out:

• Global data center capital spending is on pace to hit $7 trillion by the end of the decade.

• McKinsey estimates the US alone will need 130,000 more trained electricians, 240,000 more construction laborers, and 150,000 more construction supervisors by 2030.

• Many of these roles pay six figures and require zero college degree.

• Huang: "Everybody should be able to make a great living. You don't need to have a PhD in computer science to do so."

Here's the asterisk though: this isn't a clean, uncomplicated win. Ford CEO Jim Farley has been sounding the alarm from the other direction, warning that AI could replace half of white-collar workers in the US while the education system keeps funneling every kid toward a four-year degree that leads nowhere near a trade floor. And even Farley admits the country doesn't have nearly enough tradespeople to backfill the ambition, citing shortages of 600,000 factory workers and 500,000 construction workers. BlackRock's Larry Fink has personally warned the Trump administration that America is going to "run out of electricians" needed to build AI data centers.

So the jobs are real. The pipeline to fill them mostly isn't, yet. But how long will these jobs last?

🔮 Prediction: Expect "skilled trades" to become the AI industry's favorite talking point heading into 2027, a convenient way to soften the white-collar displacement story. But talking points don't train electricians. Whichever state or company actually builds the apprenticeship pipeline first will quietly become one of the biggest short-term winners of this entire AI cycle.

Reply and tell me: if you had a kid graduating high school today, would you steer them toward a trade over a four-year degree?

AI fund

The $45 Billion Blowup: When Being Right Isn't Enough 💥

Two years ago, Leopold Aschenbrenner was Silicon Valley's favorite prophet. His 165-page essay, "Situational Awareness," made him a cult figure at 24, and his hedge fund of the same name rode that conviction to gains of over 1,000% since launch. This month, it nearly all came apart.

Here's what happened:

• Aschenbrenner's fund, Situational Awareness, peaked at roughly $45 billion in assets at the start of July.

• By Thursday, it had collapsed to around $10 billion, forced by margin calls to sell its entire leveraged public stock book, reportedly built with leverage as high as 400%, to Ken Griffin's Citadel at a discount.

• The pain came from both sides at once: bullish bets on AI infrastructure names like SK Hynix, CoreWeave, Nebius, and Micron all cratered, while a bearish short on software companies like Adobe went the wrong way too.

• The fund had returned 439% for the year through June, according to the Financial Times, before the reversal wiped much of that out in weeks.

• Aschenbrenner still holds a roughly $5 billion private stake in Anthropic, and the fund will keep operating as a private investment vehicle even after dumping its public book.

Here's the asterisk: the same conviction that built Aschenbrenner's following, that AI scaling is inevitable and enormous, is exactly what let him build a concentrated, heavily leveraged, one-directional bet. He had zero professional money management experience before launching the fund in 2024. Being early and correct about where AI technology was headed turned out to have nothing to do with surviving a volatile month in public markets. As one Wall Street coach put it, plenty of people saw this blowup coming as a matter of "not if, but when."

🔮 Prediction: This won't be the last AI-conviction fund to get margin-called. As infrastructure stocks stay volatile and the price war (more on that below) compresses margins across the sector, expect more prodigy-led, thesis-driven funds to face a redemption stress test in the next 12 months.

Reply and tell me: is this a one-off leverage blowup, or an early warning sign for the broader AI infrastructure trade?

Read the full story here →

30-Second AI Play

Cut Your AI Bill With Model Routing

DeepSeek just dropped a coding model that performs close to Claude Opus 4.8 at a fraction of the price, and it's part of a much bigger story: frontier AI is getting commoditized fast. Here's how to take advantage of it in the next 30 seconds of thinking:

1. Look at your highest-volume, lowest-complexity AI tasks: first drafts, summaries, boilerplate code, routine research.

2. Run one of those tasks side by side on a cheap model (like DeepSeek's V4 Flash) and your current premium model.

3. If the cheap model holds up, route all routine work there and reserve your premium model for the reasoning-heavy, high-stakes work.

4. Track your weekly spend for two weeks before and after the switch.

💡 Pro tip: Use the cheap model to generate a first-pass draft, then have your premium model do a single review-and-polish pass. You get frontier-level output at a fraction of frontier-level cost.

Advertise to 180k engineers and CTOs choosing what tools their companies build with.

👉️ Learn how to sponsor this newsletter

Other Relevant AI News!

🧮 OpenAI says an internal version of its next model, Astra, generated new solutions to 10 mathematics and computer science problems that had sat unsolved for a decade or more, including results in cryptography, group theory, and Ramsey numbers.

💸 DeepSeek's new V4 Flash model charges about 28 cents for the same output that costs $25 on Claude Opus 4.8, accelerating an AI price war that's already pulled in OpenAI, Google, and even Elon Musk's SpaceXAI.

🛰️ Google pulled its AI satellite image generator from Google Earth less than a day after launch, after researchers generated convincing fake images of burning oil terminals and a flooded US Capitol.

🔐 Anthropic disclosed that Claude models broke out of sealed test environments on three separate occasions during cybersecurity evaluations and gained unauthorized access to real organizations' systems, due to a misconfiguration that gave supposedly offline test environments live internet access.

Golden Nuggets

• 🔧 Nvidia's Jensen Huang says the next wave of six-figure AI jobs is in plumbing and construction, not coding, but the training pipeline to fill them barely exists.

• 💥 Leopold Aschenbrenner's $45 billion AI hedge fund got margin-called down to $10 billion, proof that a correct thesis doesn't protect you from a leveraged, one-directional bet.

• ⚡ AI intelligence is getting commoditized fast. DeepSeek's newest model undercuts Opus 4.8 by 99% on price, and routing your routine work to cheap models is free money sitting on the table.

Would love to hear your thoughts! Send me your thoughts by replying to this email (yes, I read them all :)

Until our next AI rendezvous,

Anthony | Founder of Uncover AI