Ship first, apologize later.

Plus Second-Order Thinking & Next Level Phish.

Welcome back!

Once upon a time, in the 3rd century during the reign of Emperor Claudius II, soldiers were reportedly banned from getting married.

This was because the emperor, in his irrefutable wisdom 🙃, believed single men made better soldiers. I’m sure there are logical reasons for and against, but nobody dared go against the emperor. Of course.

Except a Roman priest named Valentine. Legend says he secretly married couples whose husbands were soldiers, and that defiance eventually contributed to his execution. After his death, he became linked with love and devotion.

Later, medieval poets, especially Geoffrey Chaucer, helped cement February 14 as a romance day.

In 1381/1382, Chaucer wrote “Parliament of Foules,” which is widely considered one of the earliest moments in literature connecting February 14 to romantic love. He wrote: “For this was on Saint Valentine’s Day, when every bird comes there to choose his mate.”

Now, I’m not sure how historically airtight this origin tale is, but it’s as good an origin story as any 😄. Today we celebrate February 14 as a symbol of choosing love, publicly, intentionally, and with courage.

So, I hope you got to celebrate love in your life this past weekend, and be celebrated. And the greatest is this: God loves us all (John 3:16).

Let’s dive in.

---

On deck:

◾ Second-Order Thinking

◾ Seedance: ship first, apologize later

◾ Next level Phish

◾ Quick Hits of other AI News worth your attention

---

---

💡Concept Corner

Practical ideas to work faster and smarter.

What it is:

Second-order thinking is the habit of asking, “And then what?” before you commit.

First-order thinking sees the obvious outcome: “If I do X, I’ll get Y.”

Second-order thinking sees the chain reaction--how today’s decision changes tomorrow’s options: “If I get Y… what does that create, change, or cost next?”

It’s the difference between a quick win and a durable advantage.

Example:

Let’s say you’re building a newsletter and you’re tempted to chase a viral post format because it got you more clicks.

First-order thinking says: “More views, more growth.” Second-order thinking asks: “Will this attract the right people, or train the wrong audience to expect fluff?”

If the headline is punchy but the content is thoughtful, you might win attention and trust.

But if the headline overpromises or is clickbait, at best you may grow fast while teaching people to skim, bounce, and not treat your content as a signal they can rely on.

This is where AI can be a useful sparring partner: you can feed it the same idea in two versions, one where you optimize for reach, another where you optimize for conversion and trust.

You will discover that the “viral” path grows followers who don’t buy, while the “useful” path grows fewer, but higher-intent, subscribers and clients.

Another example, second-order thinking can protect your calendar.

Saying “yes” to a project isn’t only a yes to the work and its payoff, it’s a “no” to whatever that work displaces: sleep, deep focus, family time, or the one high-leverage task you’ve been putting off.

So, second-order thinking forces the real math: It helps you decide whether the trade is worth it, and if you’re still leaning yes, what the reward needs to be to actually justify the cost.

Prompt Template:

Act as my second-order thinking coach. I’m deciding: [describe decision]. First, list the immediate (first-order) outcomes. Then map 6–8 second-order consequences across [List the factors that are important to you. Some examples are: time, family, health, revenue, energy, audience, reputation, relationships, brand trust, future options/opportunities]. For each consequence, rate it as likely/unlikely and high/medium/low impact. Then show me the best-case, worst-case, and most probable scenario, and recommend the decision that maximizes long-term leverage.

---

📡Signal behind the buzz🔊

Decoding trending AI stories.

---

📍Seedance: ship first, apologize later

🔊Buzz:

ByteDance’s new video generator, Seedance 2.0, started trending because it can generate slick “movie trailer” clips from a prompt…

… including videos that look like real celebrities or famous characters. That was the problem.

Disney and Paramount reportedly sent cease-and-desist letters, and Hollywood groups and unions criticized the tool for letting people remix protected franchises and faces without permission.

ByteDance says it will tighten safeguards to prevent unauthorized use of copyrighted material and people’s likenesses.

📡Signal:

Unfortunately, this is starting to look like the default playbook for too many AI companies: ship first, apologize later.

If ByteDance genuinely did not know their model could generate near-exact lookalikes of famous people and protected IP, then they did not test properly before release.

Given the power and reach of AI, that is wildly irresponsible.

If they did test it and shipped anyway, that is just as irresponsible, only more deliberate.

🎯Impact:

In a capitalist system, one of the most effective ways to hold corporations accountable is to hit the bottom line.

The consequences for this kind of behavior should be serious, sustained fines and meaningful sanctions, not toothless outrage and peekaboo legal threats.

Because AI is not just another new technology. It is the most powerful technology humanity has ever created, and it cannot be treated like a beta feature.

---

📍Next level Phish

🔊Buzz:

Security experts are flagging a potent new twist on an old scam: you get a Zoom call from someone who looks like a real exec or recruiter you trust. Except it’s a deepfake.

The video is convincing. The vibe is urgent. Then comes the “tiny” problem: your audio isn’t working.

So they send a quick fix, often a link or file, and pressure you to install it “so we can keep moving.”

However, that “fix” is malware.

📡Signal:

Professional deepfakes used to be quite hard to pull off and required specialized software and skills.

(Unfortunately) AI has crushed that barrier, making deepfakes and Blackhat social engineering cheaper to produce, easier to scale, and far more convincing.

🎯Impact:

We have to be sharper and more cautious in this day and age.

Treat video calls as untrusted identity proof. Build a two-channel habit: if a request is unusual, verify it through a second, known path (call back on a saved number, reply in an existing email thread, or use a pre-shared phrase).

Companies: lock down privileges. Outside of vetted IT staff, nobody should have admin rights by default or be able to install software without approval.

---

🍵Quick Hits of Other AI News

🧠 Google upgraded Gemini 3 Deep Think, enhancing its “heavy reasoning” mode for science and engineering, as its play to own the “serious thinking” lane.

🏥 AI is entering operating rooms, and so are incident reports, with regulators and lawsuits flagging AI-enabled surgical tools tied to misidentified anatomy and bad outcomes.

📉 Wall Street is getting more skeptical about Big Tech’s massive AI spending and unclear near-term payback, and stock prices are reflecting that skepticism.

🌍 Global AI power-brokering shifted to India this week, as leaders from OpenAI, Google, Microsoft, Amazon, and Anthropic showed up for a major summit.

📰 Amazon is reportedly building an “AI content marketplace” so publishers can license their work to model builders, turning training data rights into a product.

🇬🇧 The UK is moving to pull AI chatbots under the Online Safety Act, meaning Ofcom (UK's regulator for communications services) can hit violators with fines up to £18M or 10% of global revenue.

🦹🏼 Microsoft security researchers say “AI recommendation poisoning” is real: attackers can plant fake “memories” so an assistant’s future suggestions quietly nudge you toward a certain direction. Some companies are doing this with free “summarize with AI” buttons. Please read the article.

---

Thanks for reading, see you next week!

-Michael.

---