🍭 Guardrails, Grief, and “AI Privilege” 🤝

Good morning. I taught ChatGPT my love language and now it only texts me passive-aggressive memes.

Let’s dig in.

🍭 What’s Cookin’:

• Altman <> Tucker

• Oracle’s giving ‘dot-com bubble’

• How To Build Your Own AI Co-workers

• And Everything Else

Sam Altman <> Tucker Carlson:
🎙️ Guardrails, Grief, and “AI Privilege”

The Bite: Sam Altman used Tucker Carlson’s show to draw a hard line:

AI isn’t alive, it’s just probabilistic math.

Altman says hallucinations have fallen “in the GPT-5 era,” but the hard part isn’t accuracy, it’s values.

OpenAI’s behavior is shaped by published model specs and global feedback.

For baseline ChatGPT behavior, Altman said him and the board are to blame one for better or worse.

On safety, he floated the idea of alerting authorities for minors in acute suicide cases and argued for a new legal privilege protecting AI chats from subpoenas.

The privilege would be similar to conversations with certain professions, like lawyers or doctors, that are protected to a certain point.

He also pushed back on Carlson’s repeated murder insinuations about a former employee, saying records he reviewed point to suicide and calling the whole thing a tragedy.

Snacks:

• ChatGPT’s ‘red lines’: no bioweapon help. Even for “curiosity” or quasi-research.

• Crisis policy: today = answer refusals + providing hotline #’s

• Privacy: He suggests “AI privilege” laws similar to doctors and lawyers

• Ex-employee death: Altman says reports point to suicide and treats the murder claims as unproven.

• Power & jobs: expects broad diffusion; customer support hit hard, nurses safer; programming = TBD.

• Deepfakes: push for code-words in families and cryptographic signatures for leaders; no mandatory biometrics.

Why it Bites: Every time Altman does a podcast or interview, every single word is hyper analyzed.

And let’s be honest, this is one of the men shaping the future of humanity.
Whether they want to admit it or not.

When Tucker pressed him on a former employee’s death, Altman pushed back on the murder angle and pointed to suicide reports.

It was raw, defensive, and heavy.
And a reminder that CEOs in this space aren’t just setting guardrails, they’re getting dragged into the darkest corners of public suspicion.

Then there’s the kid question.

Altman floated the idea of alerting authorities if a minor tells ChatGPT they’re in crisis.

That’s a big shift: suddenly the chatbot isn’t just a tool, it’s a lifeguard.

Parents will cheer it, privacy folks will yell “Nineteen Eighty-Four!”,

Either way it puts every product design choice under a
new kind of moral spotlight.

Stop pretending you read those email attachments

Just forward or Cc the offending emails to [email protected].

We’ll reply with smart answers.

(You can even try forwarding this newsletter).

Steal This Prompt
🦊 3D Kawaii Icons

Simple. Smooth. 3D.

Perfect for avatar remixes, UI icons, or just for fun.

Prompt Credit: AZIZ

1. Click this link (prompt)

2. Pick any subject to icon-ize

3. Click copy

4. Paste into ChatGPT

🍭 Try the prompt on Snack Prompt 🍭

_{Everything Else}
🧠 You Need to Know

🎓 Anchor co-founders launch an AI-powered learning app

→ The team behind Anchor (sold to Spotify) is launching Oboe, where anyone can spin up mini-courses using prompts.

Think lessons across text, audio, quizzes, and even games.

🐧 Virtual pet Pengu scores $15M for its next act

→ Born, the studio behind the TikTok-favorite AI penguin, wants to make “social AI companions” a thing.

Users team up to co-parent virtual pets and play mini-games.

🧪 People are asking AI to explain their lab results

→ While waiting to hear back from doctors, more patients are dropping test results into chatbots for instant feedback.

It’s convenient, but experts warn about handing over sensitive medical info.

🧠 Google’s August AI drop is a big one

→ Gemini gains “AI Mode” in Search for 180+ countries, plus “Deep Think” for paid users.

Pixel 10 lands with a new Tensor chip and there’s a free AI Pro plan for college students, and launches like Genie 3 (world modeling), Jules (AI coding), and Nano Banana for image editing.

📦 Box bets on AI workflows with new Automate tool

→ At BoxWorks, CEO Aaron Levie said the next wave of AI is all about context.

Box Automate lets companies stitch together agentic AI workflows across messy documents and data with enterprise guardrails built in.

_ToolBox™
🧰 Build Your Own AI Co-Workers

🎙️ Deepgram — Voice OS for developers

→ Speech‑to‑text, TTS, and speech‑to‑speech APIs to build voice‑native apps fast.

📞 Vapi — Voice AI for developers

→ Build, test, and deploy phone agents in minutes with a production‑ready stack.

📊 Hex — Agentic analytics notebook

→ Blend SQL/Python with AI to explore, explain, and ship insights collaboratively.

🏢 Den — “Cursor for knowledge workers”

→ AI‑first Notion style workspace. Allegedly used by YC internally.
Spaces and docs with memory. Your org can chat, plan, and automate in one place.

🧠 Wordware — Web-hosted IDE for building AI agents

→ Orchestrate agents with a web‑hosted IDE.
Write logic in natural language and ship fast.

Cloud Computing
💾 Oracle’s AI Glow-Up: 1999 Energy

The Bite: Oracle stock just went full rocket ship up 40%+ in a day, the biggest jump in company history.

You’d think it was because they crushed earnings.

Nope. It actually missed big time on earnings.

Because it promised an AI-fueled future:
$18B in cloud revenue by 2026,
$144B by 2030,

And a $455B backlog from a handful of mega contracts.

The rally minted Larry Ellison $100B in under an hour, briefly making him the richest man alive.

Peak dot-com deja vu.

Snacks:

• Earnings were meh, outlook did the heavy lifting.

• $455B Contracts: Up 359% YoY, from just four big deals.

• Spending $35B this year (half its revenue), way up from 13% last year.

• All-In on OpenAI: The bet leans hard on one customer that still hasn’t proven it can make money.

Why it bites: This is the AI gold rush:
shovel billions into servers now, hope profits show up later.

If OpenAI doesn’t figure out how to print money, Oracle’s mega-spend starts looking very 1999 dot-com crash.

If they figure out how to be profitable, Oracle just bought its ticket into the AI Big Leagues.