Instagram | Sponsor this newsletter | Let’s Connect!

My fellow AI explorers

Happy May Day. And yes, I did just open a newsletter about AI with a labor holiday.

That's not a coincidence.

This week, workers are being replaced by AI systems they were forced to train. The Pentagon is signing classified contracts with the biggest names in tech. Mark Cuban is handing out warning labels like candy on Halloween. And Mistral just shipped the kind of model that lets you start a coding session, walk away, and come back to a finished pull request.

In today’s edition:

• 🪖 The Pentagon's AI Deal: seven companies, classified networks, and one very uncomfortable letter from Google employees

• 💔 Oracle's Gut Punch: 30,000 people fired after training the AI that replaced them

• ⚡ Mistral Goes Remote: async coding agents, a 128B open-weights model, and a serious moat for regulated industries

Speak messy. Prompt clean.

Go on tangents. Change your mind mid-sentence. Say "um" twelve times. Wispr Flow doesn't care — it takes everything you say, strips the filler, and gives you clean, structured text ready to paste into any AI tool.

The result: prompts with the full context your AI tools need to give you useful answers. Not the abbreviated version you'd type because typing is slow.

Works inside ChatGPT, Claude, Cursor, and every app on your screen. Millions of users worldwide, including teams at OpenAI, Vercel, and Clay.

Start flowing free

AI in Governments

🪖 The Pentagon Just Signed AI Into the Chain of Command

Seven tech giants — including Microsoft, Amazon, Google, Nvidia, OpenAI, and SpaceX — have been cleared to deploy their AI directly into classified U.S. military networks.

This is not a pilot program. This is not a study. The Department of Defense formalized agreements this Friday to give these companies access to Impact Level 6 and Impact Level 7 networks… the military's most sensitive classified environments, covering everything from secret-level data to the intelligence community's most restricted systems.

Here's what's happening:

• IL6 and IL7 clearance means these AI systems will be running on the same infrastructure that handles highly classified national security data

• 1.3 million DoD personnel have already used GenAI.mil, the Pentagon's official AI platform, generating tens of millions of prompts and deploying hundreds of thousands of agents in just five months

• Google's own employees sent an internal letter to leadership this week urging them to refuse the deal. "We want to see AI benefit humanity," they wrote, "not be used in inhumane or extremely harmful ways." The letter made its way to the Washington Post. Leadership signed anyway.

Let's sit with that last bullet for a second.

Google employees, from inside one of the most powerful companies in the world, publicly opposed a decision their employer had already made. And the company proceeded anyway. This is the new normal: the interests of the workforce and the interests of the boardroom have never been more visibly misaligned.

🔮 Prediction: This is the beginning of a permanent restructuring of the defense-tech relationship. Historically, companies like IBM and Lockheed defined the military-industrial complex. Going forward, it'll be the same logos you see on your phone and laptop. The AI arms race between the U.S. and China is now a lot more official. And Silicon Valley is now conscripted.

Read the full DoD announcement →

AI replacing employees

💔 Oracle Employees Trained the AI That Fired Them

Oracle laid off between 20,000 and 30,000 workers in March, including people with 30+ years of tenure. Many had even spent the previous months documenting their workflows to train the AI systems that replaced them.

This is one of the most human stories to come out of the AI boom so far, and I don't think it's getting nearly enough attention.

Here's what makes this different from a typical round of corporate layoffs:

• Workers were explicitly asked to document and train AI on their workflows before the layoffs hit. One former Oracle employee told TIME she was "scared about the outcome of this training" but felt trapped. "We were training AI to replace us," she said, "but the AI is the only way we can get through our workload."

• The terminations came via early morning email with same-day effect. No manager warning. Just a message from "Oracle Leadership" before most of the country had finished coffee.

• Long-tenured employees lost unvested stock options overnight, and in some cases, hundreds of thousands of dollars in compensation were gone instantly.

• Oracle reported its best growth quarter in 15 years the same period. Revenue hit $17.2 billion, up 22%. This was not a company struggling. This was a company choosing between people and data centers… and data centers won.

One former senior director with 19 years at Oracle said: "This is a job that I was so dedicated to for 19 years and gave everything to — and none of it matters."

I'm not going to editorialize too heavily here. But I will say this: we are going to have a reckoning with what we've collectively agreed "progress" means. Because when that progress requires workers to dig their own professional graves, as a company celebrates a good quarter, it’s a story that doesn't end well.

🔮 Prediction: Oracle is the canary. The model will be replicated: identify AI-replaceable roles, ask workers to train the replacements, and then cut. The question isn't whether other companies are watching. They are. The question is whether any regulatory guardrails arrive before it becomes industry standard.

Read the full TIME investigation →

AI Coding

⚡ Mistral Goes Remote: Agents in the Cloud, Finally

Mistral just launched remote coding agents in Vibe and debuted Mistral Medium 3.5: a 128B model that can run on as few as four GPUs and think on its own while you sleep.

The big shift here is that coding agents have historically lived on your machine. Mistral is moving them to the cloud. It means that you can kick off a complex coding task, go to a meeting, and come back to a finished session notification.

Key moves:

• Mistral Medium 3.5 is a dense 128B model with a 256k context window: instruction-following, reasoning, and coding in a single model, released open weights under a modified MIT license

• Remote agents in Vibe let you start a session from the CLI or directly in Le Chat, with the work running in parallel on Mistral's cloud while you do other things

• Work mode in Le Chat brings in a multi-step agent that calls tools in parallel for complex tasks like research, analysis, and cross-tool workflows

The open-weights release under a modified MIT license is significant. This is Mistral playing the long game on developer trust. If your model runs self-hosted on four GPUs, you don't need to rely on anyone's API, and your data never leaves your servers.

🔮 Prediction: The async coding agent space is about to get crowded. Anthropic, OpenAI, and Google are all heading there. But Mistral's self-hosted angle gives it a strong lane with regulated industries (healthcare, finance, defense) that cannot send code to a third-party cloud. That's a real moat.

Try Mistral Vibe →

30-Second AI Play

Turn Your Job Into an AI-Proof One (Before Your Company Does It For You)

Mark Cuban said this week that five job categories are most at risk: entry-level white-collar roles, routine coding, customer service, data analysis, and legal/finance support.

Here's a quick exercise to run on your own role. Takes about five minutes:

Step 1: Write down your five most time-consuming weekly tasks.

Step 2: For each one, ask: "Could a well-prompted AI agent complete this if given the right data?" Be honest. Yes/No.

Step 3: For every "Yes" task, ask: "What is the higher-order skill required to direct this task? Is it to set the goal, verify the output, and course correct?" That skill is your new job description.

Step 4: Identify one tool (Claude, Gemini, Mistral, Perplexity) that can handle your most time-consuming "Yes" task, and spend 30 minutes this week learning how to delegate it.

Step 5: Write down what you did with the time you saved. That's your value-add. That's your argument for staying employed.

Cuban's actual advice was blunter than most: "There are only two types of companies. Those who are great at AI and everybody else." The same applies to individuals. The workers who adapt quickly, who use AI to go deeper and not shallower, are the ones who come out ahead.

💡 Pro tip: The goal isn't to automate yourself out of a job. It's to make yourself so clearly the strategic layer on top of the AI that removing you means breaking the system.

Other Relevant AI News!

🖥️ Apple's Q2 earnings also dropped a quiet signal that a major AI acquisition may be coming. Tim Cook told analysts the company is "clearly investing more" in AI R&D and is actively "open to pursuing M&A if we think it will advance our roadmap." For a company sitting on $130B+ in cash, that's not a throwaway line.

🤖 The Pentagon's AI platform, GenAI.mil, is already seeing massive adoption. 1.3 million DoD personnel have generated tens of millions of prompts and deployed hundreds of thousands of AI agents in just five months. The transition to an "AI-first fighting force" isn't a future plan. It's already happening.

💊 The White House released a new drug abuse strategy this week that includes AI-powered wastewater surveillance and AI-assisted treatment recommendations. It’s a reminder that government AI adoption is quietly expanding into public health infrastructure too.

📉 Mark Cuban's full list of at-risk jobs this week: entry-level white-collar, routine coding, customer service, data analysis, and legal/finance support. But his real warning was for companies: those that don't redesign their workflows for AI-native operations won't survive competition from startups that are.

Golden Nuggets

• 🪖 The U.S. military is now officially an AI customer of Google, Microsoft, Amazon, OpenAI, Nvidia, and SpaceX, the most significant convergence of Silicon Valley and the Pentagon in a generation.

• 💀 Oracle's layoff story isn't about a struggling company. It's about a profitable one choosing data centers over people. That template will be copied.

• ⚡ Mistral's self-hosted angle isn't just a developer flex: it's a real moat for healthcare, finance, and defense teams that can't send code to a third-party cloud.

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