The Cogito Brief

Good morning,

If you blinked this week, you missed a ton. Here's what happened.

Three stories caught my attention because they each reveal a different piece of where AI is headed. We saw a new AI model get banned by Anthropic, conversations around a potential OpenAI IPO started to emerge, and 60 Minutes give us a glimpse into humanoid robots.

Together, they paint a much bigger picture about where AI may be heading next.

1st ad:

Marketers are using AI for everything from research to reporting. The challenge is getting reliable data into those workflows.

Agent A by Ahrefs combines AI with Ahrefs' marketing intelligence to help you move faster. Ask it to analyze competitors, find keyword opportunities, identify content gaps, investigate ranking drops, audit technical issues, track your brand's visibility in AI search, or build content plans and reports.

Agent A comes with pre-built marketing skills, so you don't need to master prompt engineering to get useful results. It can also connect with tools like Slack, Notion, HubSpot, WordPress, and more, helping teams turn insights into action without constantly switching between platforms.

If you want an AI assistant that does more than generate text, see what Agent A can do for your marketing workflow at ahrefs.com/agent-a.

Explore Agent A

On June 9th, Anthropic released Claude Fable 5, a new AI model designed to push agentic AI and coding workflows to the next level. According to my X feed, many believed it could become one of the biggest AI releases of the year, and for a few days, that looked like it might actually happen.

To keep it simple, I asked Claude to explain Fable 5 like I was a teenager, and here’s what it said:

“Fable 5 is like a super-smart robot that lives inside a computer and can help create things, solve problems, and build projects with very little human guidance.”

Shortly after its release, however, concerns began to emerge. Government officials and security researchers worried that certain capabilities could be exploited to identify weaknesses in software and cybersecurity systems, which ultimately led to restrictions being placed on the model.

In my opinion, this was probably the right decision, even if some people in the AI community disagreed. As these models become more powerful, safety concerns are going to become much more common.

At the same time, this story highlights a much bigger conversation that we will have to face over the next several years. AI is becoming so powerful that people with little to no technical background will eventually have access to tools that rival expert-level knowledge.

The difficult question will no longer be whether AI is capable enough. Instead, it will become: How much access is too much access?

Because eventually, society will have to find a balance between innovation and safety, and that may become one of the biggest technology debates of the next decade.

2nd ad:

PEFT. Human-in-the-loop. Software factories. Seven days to ship. Every Wednesday, Gauntlet AI professors go deep on real AI engineering — live and free. Browse Sessions

There has been a lot of discussion surrounding a potential OpenAI IPO, and for good reason.

If OpenAI eventually goes public, it could become one of the biggest technology IPOs (behind SpaceX) in recent history.

You may ask, “Why should I care?”

My answer is simple: it would be another sign that AI is no longer just a cool technology trend. AI is becoming a real industry with real money, real investors, and real pressure to grow.

Public companies are expected to grow, release new products, increase revenue, and prove to investors that they deserve their valuation. If OpenAI eventually goes public, that could push the company to move faster, expand into more industries, and turn AI into an even bigger part of everyday business.

In my opinion, an OpenAI IPO would be less about the stock market and more about what it represents: AI becoming mainstream.

3rd ad:

Agents share a staging database. One bad migration breaks every downstream task. ghost forks a fresh Postgres database per task. 58ms. Delete when done. Start building for free.

60 Minutes recently covered Atlas, the humanoid robot Boston Dynamics has been developing, and it immediately caught my attention. Here’s a link to watch the YouTube video.

Atlas is currently being developed with the goal of joining the factory floor at Hyundai, which gives you a glimpse into where this technology may be heading. Instead of robots only being used for repetitive assembly-line tasks, the next generation of robots is slowly being trained to understand human motion, repeat physical tasks, and eventually work in real-world environments.

One way companies are doing this is by studying human movement. They can use sensors and other equipment to track how a person walks, lifts, reaches, balances, and completes a task. Over time, that information can help train robots to copy certain movements and perform physical work more effectively.

One thing I’ve learned while following AI is that this technology rarely moves in one direction.

Seeing these developments happen in real time is nothing short of incredible. One week it’s model safety, the next it’s IPO rumors, and then it’s humanoid robots.

Tomorrow, we’ll take a closer look at Boston Dynamics, Atlas, and the future of humanoid robots.

Zack Wright